xref: /openbsd/gnu/usr.bin/perl/pod/perlguts.pod (revision 3d61058a)

=for comment
The part of this file between =for mg_vtable.pl markers is auto
generated by mg_vtable.pl; any changes there need to be made instead to
mg_vtable.pl

=head1 NAME

perlguts - Introduction to the Perl API

=head1 DESCRIPTION

This document attempts to describe how to use the Perl API, as well as
to provide some info on the basic workings of the Perl core.  It is far
from complete and probably contains many errors.  Please refer any
questions or comments to the author below.

=head1 Variables

=head2 Datatypes

Perl has three typedefs that handle Perl's three main data types:

    SV  Scalar Value
    AV  Array Value
    HV  Hash Value

Each typedef has specific routines that manipulate the various data types.

=for apidoc_section $AV
=for apidoc Ayh||AV
=for apidoc_section $HV
=for apidoc Ayh||HV
=for apidoc_section $SV
=for apidoc Ayh||SV

=head2 What is an "IV"?

Perl uses a special typedef IV which is a simple signed integer type that is
guaranteed to be large enough to hold a pointer (as well as an integer).
Additionally, there is the UV, which is simply an unsigned IV.

Perl also uses several special typedefs to declare variables to hold
integers of (at least) a given size.
Use I8, I16, I32, and I64 to declare a signed integer variable which has
at least as many bits as the number in its name.  These all evaluate to
the native C type that is closest to the given number of bits, but no
smaller than that number.  For example, on many platforms, a C<short> is
16 bits long, and if so, I16 will evaluate to a C<short>.  But on
platforms where a C<short> isn't exactly 16 bits, Perl will use the
smallest type that contains 16 bits or more.

U8, U16, U32, and U64 are to declare the corresponding unsigned integer
types.

If the platform doesn't support 64-bit integers, both I64 and U64 will
be undefined.  Use IV and UV to declare the largest practicable, and
C<L<perlapi/WIDEST_UTYPE>> for the absolute maximum unsigned, but which
may not be usable in all circumstances.

A numeric constant can be specified with L<perlapi/C<INT16_C>>,
L<perlapi/C<UINTMAX_C>>, and similar.

=for apidoc_section $integer
=for apidoc  Ayh ||IV
=for apidoc_item ||I8
=for apidoc_item ||I16
=for apidoc_item ||I32
=for apidoc_item ||I64

=for apidoc  Ayh ||UV
=for apidoc_item ||U8
=for apidoc_item ||U16
=for apidoc_item ||U32
=for apidoc_item ||U64

=head2 Working with SVs

An SV can be created and loaded with one command.  There are five types of
values that can be loaded: an integer value (IV), an unsigned integer
value (UV), a double (NV), a string (PV), and another scalar (SV).
("PV" stands for "Pointer Value".  You might think that it is misnamed
because it is described as pointing only to strings.  However, it is
possible to have it point to other things.  For example, it could point
to an array of UVs.  But,
using it for non-strings requires care, as the underlying assumption of
much of the internals is that PVs are just for strings.  Often, for
example, a trailing C<NUL> is tacked on automatically.  The non-string use
is documented only in this paragraph.)

=for apidoc_section $floating
=for apidoc Ayh||NV

The seven routines are:

    SV*  newSViv(IV);
    SV*  newSVuv(UV);
    SV*  newSVnv(double);
    SV*  newSVpv(const char*, STRLEN);
    SV*  newSVpvn(const char*, STRLEN);
    SV*  newSVpvf(const char*, ...);
    SV*  newSVsv(SV*);

C<STRLEN> is an integer type (C<Size_t>, usually defined as C<size_t> in
F<config.h>) guaranteed to be large enough to represent the size of
any string that perl can handle.

=for apidoc_section $string
=for apidoc Ayh||STRLEN

In the unlikely case of a SV requiring more complex initialization, you
can create an empty SV with newSV(len).  If C<len> is 0 an empty SV of
type NULL is returned, else an SV of type PV is returned with len + 1 (for
the C<NUL>) bytes of storage allocated, accessible via SvPVX.  In both cases
the SV has the undef value.

    SV *sv = newSV(0);   /* no storage allocated  */
    SV *sv = newSV(10);  /* 10 (+1) bytes of uninitialised storage
                          * allocated */

To change the value of an I<already-existing> SV, there are eight routines:

    void  sv_setiv(SV*, IV);
    void  sv_setuv(SV*, UV);
    void  sv_setnv(SV*, double);
    void  sv_setpv(SV*, const char*);
    void  sv_setpvn(SV*, const char*, STRLEN)
    void  sv_setpvf(SV*, const char*, ...);
    void  sv_vsetpvfn(SV*, const char*, STRLEN, va_list *,
                                        SV **, Size_t, bool *);
    void  sv_setsv(SV*, SV*);

Notice that you can choose to specify the length of the string to be
assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
allow Perl to calculate the length by using C<sv_setpv> or by specifying
0 as the second argument to C<newSVpv>.  Be warned, though, that Perl will
determine the string's length by using C<strlen>, which depends on the
string terminating with a C<NUL> character, and not otherwise containing
NULs.

The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
formatted output becomes the value.

C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
either a pointer to a variable argument list or the address and length of
an array of SVs.  The last argument points to a boolean; on return, if that
boolean is true, then locale-specific information has been used to format
the string, and the string's contents are therefore untrustworthy (see
L<perlsec>).  This pointer may be NULL if that information is not
important.  Note that this function requires you to specify the length of
the format.

The C<sv_set*()> functions are not generic enough to operate on values
that have "magic".  See L</Magic Virtual Tables> later in this document.

All SVs that contain strings should be terminated with a C<NUL> character.
If it is not C<NUL>-terminated there is a risk of
core dumps and corruptions from code which passes the string to C
functions or system calls which expect a C<NUL>-terminated string.
Perl's own functions typically add a trailing C<NUL> for this reason.
Nevertheless, you should be very careful when you pass a string stored
in an SV to a C function or system call.

To access the actual value that an SV points to, Perl's API exposes
several macros that coerce the actual scalar type into an IV, UV, double,
or string:

=over

=item * C<SvIV(SV*)> (C<IV>) and C<SvUV(SV*)> (C<UV>)

=item * C<SvNV(SV*)> (C<double>)

=item * Strings are a bit complicated:

=over

=item * Byte string: C<SvPVbyte(SV*, STRLEN len)> or C<SvPVbyte_nolen(SV*)>

If the Perl string is C<"\xff\xff">, then this returns a 2-byte C<char*>.

This is suitable for Perl strings that represent bytes.

=item * UTF-8 string: C<SvPVutf8(SV*, STRLEN len)> or C<SvPVutf8_nolen(SV*)>

If the Perl string is C<"\xff\xff">, then this returns a 4-byte C<char*>.

This is suitable for Perl strings that represent characters.

B<CAVEAT>: That C<char*> will be encoded via Perl's internal UTF-8 variant,
which means that if the SV contains non-Unicode code points (e.g.,
0x110000), then the result may contain extensions over valid UTF-8.
See L<perlapi/is_strict_utf8_string> for some methods Perl gives
you to check the UTF-8 validity of these macros' returns.

=item * You can also use C<SvPV(SV*, STRLEN len)> or C<SvPV_nolen(SV*)>
to fetch the SV's raw internal buffer. This is tricky, though; if your Perl
string
is C<"\xff\xff">, then depending on the SV's internal encoding you might get
back a 2-byte B<OR> a 4-byte C<char*>.
Moreover, if it's the 4-byte string, that could come from either Perl
C<"\xff\xff"> stored UTF-8 encoded, or Perl C<"\xc3\xbf\xc3\xbf"> stored
as raw octets. To differentiate between these you B<MUST> look up the
SV's UTF8 bit (cf. C<SvUTF8>) to know whether the source Perl string
is 2 characters (C<SvUTF8> would be on) or 4 characters (C<SvUTF8> would be
off).

B<IMPORTANT:> Use of C<SvPV>, C<SvPV_nolen>, or
similarly-named macros I<without> looking up the SV's UTF8 bit is
almost certainly a bug if non-ASCII input is allowed.

When the UTF8 bit is on, the same B<CAVEAT> about UTF-8 validity applies
here as for C<SvPVutf8>.

=back

(See L</How do I pass a Perl string to a C library?> for more details.)

In C<SvPVbyte>, C<SvPVutf8>, and C<SvPV>, the length of the C<char*> returned
is placed into the
variable C<len> (these are macros, so you do I<not> use C<&len>). If you do
not care what the length of the data is, use C<SvPVbyte_nolen>,
C<SvPVutf8_nolen>, or C<SvPV_nolen> instead.
The global variable C<PL_na> can also be given to
C<SvPVbyte>/C<SvPVutf8>/C<SvPV>
in this case.  But that can be quite inefficient because C<PL_na> must
be accessed in thread-local storage in threaded Perl.  In any case, remember
that Perl allows arbitrary strings of data that may both contain NULs and
might not be terminated by a C<NUL>.

Also remember that C doesn't allow you to safely say C<foo(SvPVbyte(s, len),
len);>.  It might work with your
compiler, but it won't work for everyone.
Break this sort of statement up into separate assignments:

    SV *s;
    STRLEN len;
    char *ptr;
    ptr = SvPVbyte(s, len);
    foo(ptr, len);

=back

If you want to know if the scalar value is TRUE, you can use:

    SvTRUE(SV*)

Although Perl will automatically grow strings for you, if you need to force
Perl to allocate more memory for your SV, you can use the macro

    SvGROW(SV*, STRLEN newlen)

which will determine if more memory needs to be allocated.  If so, it will
call the function C<sv_grow>.  Note that C<SvGROW> can only increase, not
decrease, the allocated memory of an SV and that it does not automatically
add space for the trailing C<NUL> byte (perl's own string functions typically do
C<SvGROW(sv, len + 1)>).

If you want to write to an existing SV's buffer and set its value to a
string, use SvPVbyte_force() or one of its variants to force the SV to be
a PV.  This will remove any of various types of non-stringness from
the SV while preserving the content of the SV in the PV.  This can be
used, for example, to append data from an API function to a buffer
without extra copying:

    (void)SvPVbyte_force(sv, len);
    s = SvGROW(sv, len + needlen + 1);
    /* something that modifies up to needlen bytes at s+len, but
       modifies newlen bytes
         eg. newlen = read(fd, s + len, needlen);
       ignoring errors for these examples
     */
    s[len + newlen] = '\0';
    SvCUR_set(sv, len + newlen);
    SvUTF8_off(sv);
    SvSETMAGIC(sv);

If you already have the data in memory or if you want to keep your
code simple, you can use one of the sv_cat*() variants, such as
sv_catpvn().  If you want to insert anywhere in the string you can use
sv_insert() or sv_insert_flags().

If you don't need the existing content of the SV, you can avoid some
copying with:

    SvPVCLEAR(sv);
    s = SvGROW(sv, needlen + 1);
    /* something that modifies up to needlen bytes at s, but modifies
       newlen bytes
         eg. newlen = read(fd, s, needlen);
     */
    s[newlen] = '\0';
    SvCUR_set(sv, newlen);
    SvPOK_only(sv); /* also clears SVf_UTF8 */
    SvSETMAGIC(sv);

Again, if you already have the data in memory or want to avoid the
complexity of the above, you can use sv_setpvn().

If you have a buffer allocated with Newx() and want to set that as the
SV's value, you can use sv_usepvn_flags().  That has some requirements
if you want to avoid perl re-allocating the buffer to fit the trailing
NUL:

   Newx(buf, somesize+1, char);
   /* ... fill in buf ... */
   buf[somesize] = '\0';
   sv_usepvn_flags(sv, buf, somesize, SV_SMAGIC | SV_HAS_TRAILING_NUL);
   /* buf now belongs to perl, don't release it */

If you have an SV and want to know what kind of data Perl thinks is stored
in it, you can use the following macros to check the type of SV you have.

    SvIOK(SV*)
    SvNOK(SV*)
    SvPOK(SV*)

Be aware that retrieving the numeric value of an SV can set IOK or NOK
on that SV, even when the SV started as a string.  Prior to Perl
5.36.0 retrieving the string value of an integer could set POK, but
this can no longer occur.  From 5.36.0 this can be used to distinguish
the original representation of an SV and is intended to make life
simpler for serializers:

    /* references handled elsewhere */
    if (SvIsBOOL(sv)) {
        /* originally boolean */
        ...
    }
    else if (SvPOK(sv)) {
        /* originally a string */
        ...
    }
    else if (SvNIOK(sv)) {
        /* originally numeric */
        ...
    }
    else {
        /* something special or undef */
    }

You can get and set the current length of the string stored in an SV with
the following macros:

    SvCUR(SV*)
    SvCUR_set(SV*, I32 val)

You can also get a pointer to the end of the string stored in the SV
with the macro:

    SvEND(SV*)

But note that these last three macros are valid only if C<SvPOK()> is true.

If you want to append something to the end of string stored in an C<SV*>,
you can use the following functions:

    void  sv_catpv(SV*, const char*);
    void  sv_catpvn(SV*, const char*, STRLEN);
    void  sv_catpvf(SV*, const char*, ...);
    void  sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **,
                                                             I32, bool);
    void  sv_catsv(SV*, SV*);

The first function calculates the length of the string to be appended by
using C<strlen>.  In the second, you specify the length of the string
yourself.  The third function processes its arguments like C<sprintf> and
appends the formatted output.  The fourth function works like C<vsprintf>.
You can specify the address and length of an array of SVs instead of the
va_list argument.  The fifth function
extends the string stored in the first
SV with the string stored in the second SV.  It also forces the second SV
to be interpreted as a string.

The C<sv_cat*()> functions are not generic enough to operate on values that
have "magic".  See L</Magic Virtual Tables> later in this document.

If you know the name of a scalar variable, you can get a pointer to its SV
by using the following:

    SV*  get_sv("package::varname", 0);

This returns NULL if the variable does not exist.

If you want to know if this variable (or any other SV) is actually C<defined>,
you can call:

    SvOK(SV*)

The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>.

Its address can be used whenever an C<SV*> is needed.  Make sure that
you don't try to compare a random sv with C<&PL_sv_undef>.  For example
when interfacing Perl code, it'll work correctly for:

  foo(undef);

But won't work when called as:

  $x = undef;
  foo($x);

So to repeat always use SvOK() to check whether an sv is defined.

Also you have to be careful when using C<&PL_sv_undef> as a value in
AVs or HVs (see L</AVs, HVs and undefined values>).

There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain
boolean TRUE and FALSE values, respectively.  Like C<PL_sv_undef>, their
addresses can be used whenever an C<SV*> is needed.

Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
Take this code:

    SV* sv = (SV*) 0;
    if (I-am-to-return-a-real-value) {
            sv = sv_2mortal(newSViv(42));
    }
    sv_setsv(ST(0), sv);

This code tries to return a new SV (which contains the value 42) if it should
return a real value, or undef otherwise.  Instead it has returned a NULL
pointer which, somewhere down the line, will cause a segmentation violation,
bus error, or just weird results.  Change the zero to C<&PL_sv_undef> in the
first line and all will be well.

To free an SV that you've created, call C<SvREFCNT_dec(SV*)>.  Normally this
call is not necessary (see L</Reference Counts and Mortality>).

=head2 Offsets

Perl provides the function C<sv_chop> to efficiently remove characters
from the beginning of a string; you give it an SV and a pointer to
somewhere inside the PV, and it discards everything before the
pointer.  The efficiency comes by means of a little hack: instead of
actually removing the characters, C<sv_chop> sets the flag C<OOK>
(offset OK) to signal to other functions that the offset hack is in
effect, and it moves the PV pointer (called C<SvPVX>) forward
by the number of bytes chopped off, and adjusts C<SvCUR> and C<SvLEN>
accordingly.  (A portion of the space between the old and new PV
pointers is used to store the count of chopped bytes.)

Hence, at this point, the start of the buffer that we allocated lives
at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
into the middle of this allocated storage.

This is best demonstrated by example.  Normally copy-on-write will prevent
the substitution from operator from using this hack, but if you can craft a
string for which copy-on-write is not possible, you can see it in play.  In
the current implementation, the final byte of a string buffer is used as a
copy-on-write reference count.  If the buffer is not big enough, then
copy-on-write is skipped.  First have a look at an empty string:

  % ./perl -Ilib -MDevel::Peek -le '$a=""; $a .= ""; Dump $a'
  SV = PV(0x7ffb7c008a70) at 0x7ffb7c030390
    REFCNT = 1
    FLAGS = (POK,pPOK)
    PV = 0x7ffb7bc05b50 ""\0
    CUR = 0
    LEN = 10

Notice here the LEN is 10.  (It may differ on your platform.)  Extend the
length of the string to one less than 10, and do a substitution:

 % ./perl -Ilib -MDevel::Peek -le '$a=""; $a.="123456789"; $a=~s/.//; \
                                                            Dump($a)'
 SV = PV(0x7ffa04008a70) at 0x7ffa04030390
   REFCNT = 1
   FLAGS = (POK,OOK,pPOK)
   OFFSET = 1
   PV = 0x7ffa03c05b61 ( "\1" . ) "23456789"\0
   CUR = 8
   LEN = 9

Here the number of bytes chopped off (1) is shown next as the OFFSET.  The
portion of the string between the "real" and the "fake" beginnings is
shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
the fake beginning, not the real one.  (The first character of the string
buffer happens to have changed to "\1" here, not "1", because the current
implementation stores the offset count in the string buffer.  This is
subject to change.)

Something similar to the offset hack is performed on AVs to enable
efficient shifting and splicing off the beginning of the array; while
C<AvARRAY> points to the first element in the array that is visible from
Perl, C<AvALLOC> points to the real start of the C array.  These are
usually the same, but a C<shift> operation can be carried out by
increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvMAX>.
Again, the location of the real start of the C array only comes into
play when freeing the array.  See C<av_shift> in F<av.c>.

=for apidoc_section $AV
=for apidoc Amh||AvALLOC|AV* av

=head2 What's Really Stored in an SV?

Recall that the usual method of determining the type of scalar you have is
to use C<Sv*OK> macros.  Because a scalar can be both a number and a string,
usually these macros will always return TRUE and calling the C<Sv*V>
macros will do the appropriate conversion of string to integer/double or
integer/double to string.

If you I<really> need to know if you have an integer, double, or string
pointer in an SV, you can use the following three macros instead:

    SvIOKp(SV*)
    SvNOKp(SV*)
    SvPOKp(SV*)

These will tell you if you truly have an integer, double, or string pointer
stored in your SV.  The "p" stands for private.

There are various ways in which the private and public flags may differ.
For example, in perl 5.16 and earlier a tied SV may have a valid
underlying value in the IV slot (so SvIOKp is true), but the data
should be accessed via the FETCH routine rather than directly,
so SvIOK is false.  (In perl 5.18 onwards, tied scalars use
the flags the same way as untied scalars.)  Another is when
numeric conversion has occurred and precision has been lost: only the
private flag is set on 'lossy' values.  So when an NV is converted to an
IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.

In general, though, it's best to use the C<Sv*V> macros.

=head2 Working with AVs

There are two main, longstanding ways to create and load an AV.  The first
method creates an empty AV:

    AV*  newAV();

The second method both creates the AV and initially populates it with SVs:

    AV*  av_make(SSize_t num, SV **ptr);

The second argument points to an array containing C<num> C<SV*>'s.  Once the
AV has been created, the SVs can be destroyed, if so desired.

Perl v5.36 added two new ways to create an AV and allocate a SV** array
without populating it. These are more efficient than a newAV() followed by an
av_extend().

    /* Creates but does not initialize (Zero) the SV** array */
    AV *av = newAV_alloc_x(1);
    /* Creates and does initialize (Zero) the SV** array */
    AV *av = newAV_alloc_xz(1);

The numerical argument refers to the number of array elements to allocate, not
an array index, and must be >0. The first form must only ever be used when all
elements will be initialized before any read occurs. Reading a non-initialized
SV* - i.e. treating a random memory address as a SV* - is a serious bug.

Once the AV has been created, the following operations are possible on it:

    void  av_push(AV*, SV*);
    SV*   av_pop(AV*);
    SV*   av_shift(AV*);
    void  av_unshift(AV*, SSize_t num);

These should be familiar operations, with the exception of C<av_unshift>.
This routine adds C<num> elements at the front of the array with the C<undef>
value.  You must then use C<av_store> (described below) to assign values
to these new elements.

Here are some other functions:

    Size_t  av_count(AV*);
    SSize_t av_top_index(AV*);
    SV**    av_fetch(AV*, SSize_t key, I32 lval);
    SV**    av_store(AV*, SSize_t key, SV* val);

C<av_count> returns the number of elements in the array (including
any empty slots (undefined ones) that are intermixed with filled-in ones).
The C<av_top_index> function returns the highest index value in an array (just
like $#array in Perl).  If the array is empty, -1 is returned.  It is
always equal to S<C<av_count() - 1>>.  The
C<av_fetch> function returns the value at index C<key>, but if C<lval>
is non-zero, then C<av_fetch> will store an undef value at that index.
The C<av_store> function stores the value C<val> at index C<key>, and does
not increment the reference count of C<val>.  Thus the caller is responsible
for taking care of that, and if C<av_store> returns NULL, the caller will
have to decrement the reference count to avoid a memory leak.  Note that
C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
return value.

A few more:

    void  av_clear(AV*);
    void  av_undef(AV*);
    void  av_extend(AV*, SSize_t key);

The C<av_clear> function deletes all the elements in the AV* array, but
does not actually delete the array itself.  The C<av_undef> function will
delete all the elements in the array plus the array itself.  The
C<av_extend> function extends the array so that it contains at least C<key+1>
elements.  If C<key+1> is less than the currently allocated length of the array,
then nothing is done.

If you know the name of an array variable, you can get a pointer to its AV
by using the following:

    AV*  get_av("package::varname", 0);

This returns NULL if the variable does not exist.

See L</Understanding the Magic of Tied Hashes and Arrays> for more
information on how to use the array access functions on tied arrays.

=head3 More efficient working with new or vanilla AVs

Perl v5.36 and v5.38 introduced streamlined, inlined versions of some
functions:

=over

=item * C<av_store_simple>

=item * C<av_fetch_simple>

=item * C<av_push_simple>

=back

These are drop-in replacements, but can only be used on straightforward
AVs that meet the following criteria:

=over

=item * are not magical

=item * are not readonly

=item * are "real" (refcounted) AVs

=item * have an av_top_index value > -2

=back

AVs created using C<newAV()>, C<av_make>, C<newAV_alloc_x>, and
C<newAV_alloc_xz> are all compatible at the time of creation. It is
only if they are declared readonly or unreal, have magic attached, or
are otherwise configured unusually that they will stop being compatible.

Note that some interpreter functions may attach magic to an AV as part
of normal operations. It is therefore safest, unless you are sure of the
lifecycle of an AV, to only use these new functions close to the point
of AV creation.

=head2 Working with HVs

To create an HV, you use the following routine:

    HV*  newHV();

Once the HV has been created, the following operations are possible on it:

    SV**  hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
    SV**  hv_fetch(HV*, const char* key, U32 klen, I32 lval);

The C<klen> parameter is the length of the key being passed in (Note that
you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
length of the key).  The C<val> argument contains the SV pointer to the
scalar being stored, and C<hash> is the precomputed hash value (zero if
you want C<hv_store> to calculate it for you).  The C<lval> parameter
indicates whether this fetch is actually a part of a store operation, in
which case a new undefined value will be added to the HV with the supplied
key and C<hv_fetch> will return as if the value had already existed.

Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
C<SV*>.  To access the scalar value, you must first dereference the return
value.  However, you should check to make sure that the return value is
not NULL before dereferencing it.

The first of these two functions checks if a hash table entry exists, and the
second deletes it.

    bool  hv_exists(HV*, const char* key, U32 klen);
    SV*   hv_delete(HV*, const char* key, U32 klen, I32 flags);

If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
create and return a mortal copy of the deleted value.

And more miscellaneous functions:

    void   hv_clear(HV*);
    void   hv_undef(HV*);

Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
table but does not actually delete the hash table.  The C<hv_undef> deletes
both the entries and the hash table itself.

Perl keeps the actual data in a linked list of structures with a typedef of HE.
These contain the actual key and value pointers (plus extra administrative
overhead).  The key is a string pointer; the value is an C<SV*>.  However,
once you have an C<HE*>, to get the actual key and value, use the routines
specified below.

=for apidoc_section $HV
=for apidoc Ayh||HE

    I32    hv_iterinit(HV*);
            /* Prepares starting point to traverse hash table */
    HE*    hv_iternext(HV*);
            /* Get the next entry, and return a pointer to a
               structure that has both the key and value */
    char*  hv_iterkey(HE* entry, I32* retlen);
            /* Get the key from an HE structure and also return
               the length of the key string */
    SV*    hv_iterval(HV*, HE* entry);
            /* Return an SV pointer to the value of the HE
               structure */
    SV*    hv_iternextsv(HV*, char** key, I32* retlen);
            /* This convenience routine combines hv_iternext,
	       hv_iterkey, and hv_iterval.  The key and retlen
	       arguments are return values for the key and its
	       length.  The value is returned in the SV* argument */

If you know the name of a hash variable, you can get a pointer to its HV
by using the following:

    HV*  get_hv("package::varname", 0);

This returns NULL if the variable does not exist.

The hash algorithm is defined in the C<PERL_HASH> macro:

    PERL_HASH(hash, key, klen)

The exact implementation of this macro varies by architecture and version
of perl, and the return value may change per invocation, so the value
is only valid for the duration of a single perl process.

See L</Understanding the Magic of Tied Hashes and Arrays> for more
information on how to use the hash access functions on tied hashes.

=for apidoc_section $HV
=for apidoc Amh|void|PERL_HASH|U32 hash|char *key|STRLEN klen

=head2 Hash API Extensions

Beginning with version 5.004, the following functions are also supported:

    HE*     hv_fetch_ent  (HV* tb, SV* key, I32 lval, U32 hash);
    HE*     hv_store_ent  (HV* tb, SV* key, SV* val, U32 hash);

    bool    hv_exists_ent (HV* tb, SV* key, U32 hash);
    SV*     hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);

    SV*     hv_iterkeysv  (HE* entry);

Note that these functions take C<SV*> keys, which simplifies writing
of extension code that deals with hash structures.  These functions
also allow passing of C<SV*> keys to C<tie> functions without forcing
you to stringify the keys (unlike the previous set of functions).

They also return and accept whole hash entries (C<HE*>), making their
use more efficient (since the hash number for a particular string
doesn't have to be recomputed every time).  See L<perlapi> for detailed
descriptions.

The following macros must always be used to access the contents of hash
entries.  Note that the arguments to these macros must be simple
variables, since they may get evaluated more than once.  See
L<perlapi> for detailed descriptions of these macros.

    HePV(HE* he, STRLEN len)
    HeVAL(HE* he)
    HeHASH(HE* he)
    HeSVKEY(HE* he)
    HeSVKEY_force(HE* he)
    HeSVKEY_set(HE* he, SV* sv)

These two lower level macros are defined, but must only be used when
dealing with keys that are not C<SV*>s:

    HeKEY(HE* he)
    HeKLEN(HE* he)

Note that both C<hv_store> and C<hv_store_ent> do not increment the
reference count of the stored C<val>, which is the caller's responsibility.
If these functions return a NULL value, the caller will usually have to
decrement the reference count of C<val> to avoid a memory leak.

=head2 AVs, HVs and undefined values

Sometimes you have to store undefined values in AVs or HVs.  Although
this may be a rare case, it can be tricky.  That's because you're
used to using C<&PL_sv_undef> if you need an undefined SV.

For example, intuition tells you that this XS code:

    AV *av = newAV();
    av_store( av, 0, &PL_sv_undef );

is equivalent to this Perl code:

    my @av;
    $av[0] = undef;

Unfortunately, this isn't true.  In perl 5.18 and earlier, AVs use C<&PL_sv_undef> as a marker
for indicating that an array element has not yet been initialized.
Thus, C<exists $av[0]> would be true for the above Perl code, but
false for the array generated by the XS code.  In perl 5.20, storing
&PL_sv_undef will create a read-only element, because the scalar
&PL_sv_undef itself is stored, not a copy.

Similar problems can occur when storing C<&PL_sv_undef> in HVs:

    hv_store( hv, "key", 3, &PL_sv_undef, 0 );

This will indeed make the value C<undef>, but if you try to modify
the value of C<key>, you'll get the following error:

    Modification of non-creatable hash value attempted

In perl 5.8.0, C<&PL_sv_undef> was also used to mark placeholders
in restricted hashes.  This caused such hash entries not to appear
when iterating over the hash or when checking for the keys
with the C<hv_exists> function.

You can run into similar problems when you store C<&PL_sv_yes> or
C<&PL_sv_no> into AVs or HVs.  Trying to modify such elements
will give you the following error:

    Modification of a read-only value attempted

To make a long story short, you can use the special variables
C<&PL_sv_undef>, C<&PL_sv_yes> and C<&PL_sv_no> with AVs and
HVs, but you have to make sure you know what you're doing.

Generally, if you want to store an undefined value in an AV
or HV, you should not use C<&PL_sv_undef>, but rather create a
new undefined value using the C<newSV> function, for example:

    av_store( av, 42, newSV(0) );
    hv_store( hv, "foo", 3, newSV(0), 0 );

=head2 References

References are a special type of scalar that point to other data types
(including other references).

To create a reference, use either of the following functions:

    SV* newRV_inc((SV*) thing);
    SV* newRV_noinc((SV*) thing);

The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>.  The
functions are identical except that C<newRV_inc> increments the reference
count of the C<thing>, while C<newRV_noinc> does not.  For historical
reasons, C<newRV> is a synonym for C<newRV_inc>.

Once you have a reference, you can use the following macro to dereference
the reference:

    SvRV(SV*)

then call the appropriate routines, casting the returned C<SV*> to either an
C<AV*> or C<HV*>, if required.

To determine if an SV is a reference, you can use the following macro:

    SvROK(SV*)

To discover what type of value the reference refers to, use the following
macro and then check the return value.

    SvTYPE(SvRV(SV*))

The most useful types that will be returned are:

    SVt_PVAV    Array
    SVt_PVHV    Hash
    SVt_PVCV    Code
    SVt_PVGV    Glob (possibly a file handle)

Any numerical value returned which is less than SVt_PVAV will be a scalar
of some form.

See L<perlapi/svtype> for more details.

=head2 Blessed References and Class Objects

References are also used to support object-oriented programming.  In perl's
OO lexicon, an object is simply a reference that has been blessed into a
package (or class).  Once blessed, the programmer may now use the reference
to access the various methods in the class.

A reference can be blessed into a package with the following function:

    SV* sv_bless(SV* sv, HV* stash);

The C<sv> argument must be a reference value.  The C<stash> argument
specifies which class the reference will belong to.  See
L</Stashes and Globs> for information on converting class names into stashes.

/* Still under construction */

The following function upgrades rv to reference if not already one.
Creates a new SV for rv to point to.  If C<classname> is non-null, the SV
is blessed into the specified class.  SV is returned.

	SV* newSVrv(SV* rv, const char* classname);

The following three functions copy integer, unsigned integer or double
into an SV whose reference is C<rv>.  SV is blessed if C<classname> is
non-null.

	SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
	SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
	SV* sv_setref_nv(SV* rv, const char* classname, NV iv);

The following function copies the pointer value (I<the address, not the
string!>) into an SV whose reference is rv.  SV is blessed if C<classname>
is non-null.

	SV* sv_setref_pv(SV* rv, const char* classname, void* pv);

The following function copies a string into an SV whose reference is C<rv>.
Set length to 0 to let Perl calculate the string length.  SV is blessed if
C<classname> is non-null.

    SV* sv_setref_pvn(SV* rv, const char* classname, char* pv,
                                                         STRLEN length);

The following function tests whether the SV is blessed into the specified
class.  It does not check inheritance relationships.

	int  sv_isa(SV* sv, const char* name);

The following function tests whether the SV is a reference to a blessed object.

	int  sv_isobject(SV* sv);

The following function tests whether the SV is derived from the specified
class.  SV can be either a reference to a blessed object or a string
containing a class name.  This is the function implementing the
C<UNIVERSAL::isa> functionality.

	bool sv_derived_from(SV* sv, const char* name);

To check if you've got an object derived from a specific class you have
to write:

	if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }

=head2 Creating New Variables

To create a new Perl variable with an undef value which can be accessed from
your Perl script, use the following routines, depending on the variable type.

    SV*  get_sv("package::varname", GV_ADD);
    AV*  get_av("package::varname", GV_ADD);
    HV*  get_hv("package::varname", GV_ADD);

Notice the use of GV_ADD as the second parameter.  The new variable can now
be set, using the routines appropriate to the data type.

There are additional macros whose values may be bitwise OR'ed with the
C<GV_ADD> argument to enable certain extra features.  Those bits are:

=over

=item GV_ADDMULTI

Marks the variable as multiply defined, thus preventing the:

  Name <varname> used only once: possible typo

warning.

=item GV_ADDWARN

Issues the warning:

  Had to create <varname> unexpectedly

if the variable did not exist before the function was called.

=back

If you do not specify a package name, the variable is created in the current
package.

=head2 Reference Counts and Mortality

Perl uses a reference count-driven garbage collection mechanism.  SVs,
AVs, or HVs (xV for short in the following) start their life with a
reference count of 1.  If the reference count of an xV ever drops to 0,
then it will be destroyed and its memory made available for reuse.
At the most basic internal level, reference counts can be manipulated
with the following macros:

    int SvREFCNT(SV* sv);
    SV* SvREFCNT_inc(SV* sv);
    void SvREFCNT_dec(SV* sv);

(There are also suffixed versions of the increment and decrement macros,
for situations where the full generality of these basic macros can be
exchanged for some performance.)

However, the way a programmer should think about references is not so
much in terms of the bare reference count, but in terms of I<ownership>
of references.  A reference to an xV can be owned by any of a variety
of entities: another xV, the Perl interpreter, an XS data structure,
a piece of running code, or a dynamic scope.  An xV generally does not
know what entities own the references to it; it only knows how many
references there are, which is the reference count.

To correctly maintain reference counts, it is essential to keep track
of what references the XS code is manipulating.  The programmer should
always know where a reference has come from and who owns it, and be
aware of any creation or destruction of references, and any transfers
of ownership.  Because ownership isn't represented explicitly in the xV
data structures, only the reference count need be actually maintained
by the code, and that means that this understanding of ownership is not
actually evident in the code.  For example, transferring ownership of a
reference from one owner to another doesn't change the reference count
at all, so may be achieved with no actual code.  (The transferring code
doesn't touch the referenced object, but does need to ensure that the
former owner knows that it no longer owns the reference, and that the
new owner knows that it now does.)

An xV that is visible at the Perl level should not become unreferenced
and thus be destroyed.  Normally, an object will only become unreferenced
when it is no longer visible, often by the same means that makes it
invisible.  For example, a Perl reference value (RV) owns a reference to
its referent, so if the RV is overwritten that reference gets destroyed,
and the no-longer-reachable referent may be destroyed as a result.

Many functions have some kind of reference manipulation as
part of their purpose.  Sometimes this is documented in terms
of ownership of references, and sometimes it is (less helpfully)
documented in terms of changes to reference counts.  For example, the
L<newRV_inc()|perlapi/newRV_inc> function is documented to create a new RV
(with reference count 1) and increment the reference count of the referent
that was supplied by the caller.  This is best understood as creating
a new reference to the referent, which is owned by the created RV,
and returning to the caller ownership of the sole reference to the RV.
The L<newRV_noinc()|perlapi/newRV_noinc> function instead does not
increment the reference count of the referent, but the RV nevertheless
ends up owning a reference to the referent.  It is therefore implied
that the caller of C<newRV_noinc()> is relinquishing a reference to the
referent, making this conceptually a more complicated operation even
though it does less to the data structures.

For example, imagine you want to return a reference from an XSUB
function.  Inside the XSUB routine, you create an SV which initially
has just a single reference, owned by the XSUB routine.  This reference
needs to be disposed of before the routine is complete, otherwise it
will leak, preventing the SV from ever being destroyed.  So to create
an RV referencing the SV, it is most convenient to pass the SV to
C<newRV_noinc()>, which consumes that reference.  Now the XSUB routine
no longer owns a reference to the SV, but does own a reference to the RV,
which in turn owns a reference to the SV.  The ownership of the reference
to the RV is then transferred by the process of returning the RV from
the XSUB.

There are some convenience functions available that can help with the
destruction of xVs.  These functions introduce the concept of "mortality".
Much documentation speaks of an xV itself being mortal, but this is
misleading.  It is really I<a reference to> an xV that is mortal, and it
is possible for there to be more than one mortal reference to a single xV.
For a reference to be mortal means that it is owned by the temps stack,
one of perl's many internal stacks, which will destroy that reference
"a short time later".  Usually the "short time later" is the end of
the current Perl statement.  However, it gets more complicated around
dynamic scopes: there can be multiple sets of mortal references hanging
around at the same time, with different death dates.  Internally, the
actual determinant for when mortal xV references are destroyed depends
on two macros, SAVETMPS and FREETMPS.  See L<perlcall> and L<perlxs>
and L</Temporaries Stack> below for more details on these macros.

Mortal references are mainly used for xVs that are placed on perl's
main stack.  The stack is problematic for reference tracking, because it
contains a lot of xV references, but doesn't own those references: they
are not counted.  Currently, there are many bugs resulting from xVs being
destroyed while referenced by the stack, because the stack's uncounted
references aren't enough to keep the xVs alive.  So when putting an
(uncounted) reference on the stack, it is vitally important to ensure that
there will be a counted reference to the same xV that will last at least
as long as the uncounted reference.  But it's also important that that
counted reference be cleaned up at an appropriate time, and not unduly
prolong the xV's life.  For there to be a mortal reference is often the
best way to satisfy this requirement, especially if the xV was created
especially to be put on the stack and would otherwise be unreferenced.

To create a mortal reference, use the functions:

    SV*  sv_newmortal()
    SV*  sv_mortalcopy(SV*)
    SV*  sv_2mortal(SV*)

C<sv_newmortal()> creates an SV (with the undefined value) whose sole
reference is mortal.  C<sv_mortalcopy()> creates an xV whose value is a
copy of a supplied xV and whose sole reference is mortal.  C<sv_2mortal()>
mortalises an existing xV reference: it transfers ownership of a reference
from the caller to the temps stack.  Because C<sv_newmortal> gives the new
SV no value, it must normally be given one via C<sv_setpv>, C<sv_setiv>,
etc. :

    SV *tmp = sv_newmortal();
    sv_setiv(tmp, an_integer);

As that is multiple C statements it is quite common so see this idiom instead:

    SV *tmp = sv_2mortal(newSViv(an_integer));

The mortal routines are not just for SVs; AVs and HVs can be
made mortal by passing their address (type-casted to C<SV*>) to the
C<sv_2mortal> or C<sv_mortalcopy> routines.

=head2 Stashes and Globs

A B<stash> is a hash that contains all variables that are defined
within a package.  Each key of the stash is a symbol
name (shared by all the different types of objects that have the same
name), and each value in the hash table is a GV (Glob Value).  This GV
in turn contains references to the various objects of that name,
including (but not limited to) the following:

    Scalar Value
    Array Value
    Hash Value
    I/O Handle
    Format
    Subroutine

There is a single stash called C<PL_defstash> that holds the items that exist
in the C<main> package.  To get at the items in other packages, append the
string "::" to the package name.  The items in the C<Foo> package are in
the stash C<Foo::> in PL_defstash.  The items in the C<Bar::Baz> package are
in the stash C<Baz::> in C<Bar::>'s stash.

=for apidoc_section $GV
=for apidoc Amnh||PL_defstash

To get the stash pointer for a particular package, use the function:

    HV*  gv_stashpv(const char* name, I32 flags)
    HV*  gv_stashsv(SV*, I32 flags)

The first function takes a literal string, the second uses the string stored
in the SV.  Remember that a stash is just a hash table, so you get back an
C<HV*>.  The C<flags> flag will create a new package if it is set to GV_ADD.

The name that C<gv_stash*v> wants is the name of the package whose symbol table
you want.  The default package is called C<main>.  If you have multiply nested
packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
language itself.

Alternately, if you have an SV that is a blessed reference, you can find
out the stash pointer by using:

    HV*  SvSTASH(SvRV(SV*));

then use the following to get the package name itself:

    char*  HvNAME(HV* stash);

If you need to bless or re-bless an object you can use the following
function:

    SV*  sv_bless(SV*, HV* stash)

where the first argument, an C<SV*>, must be a reference, and the second
argument is a stash.  The returned C<SV*> can now be used in the same way
as any other SV.

For more information on references and blessings, consult L<perlref>.

=head2 I/O Handles

Like AVs and HVs, IO objects are another type of non-scalar SV which
may contain input and output L<PerlIO|perlapio> objects or a C<DIR *>
from opendir().

You can create a new IO object:

    IO*  newIO();

Unlike other SVs, a new IO object is automatically blessed into the
L<IO::File> class.

The IO object contains an input and output PerlIO handle:

  PerlIO *IoIFP(IO *io);
  PerlIO *IoOFP(IO *io);

=for apidoc_section $io
=for apidoc Amh|PerlIO *|IoIFP|IO *io
=for apidoc Amh|PerlIO *|IoOFP|IO *io

Typically if the IO object has been opened on a file, the input handle
is always present, but the output handle is only present if the file
is open for output.  For a file, if both are present they will be the
same PerlIO object.

Distinct input and output PerlIO objects are created for sockets and
character devices.

The IO object also contains other data associated with Perl I/O
handles:

  IV IoLINES(io);                /* $. */
  IV IoPAGE(io);                 /* $% */
  IV IoPAGE_LEN(io);             /* $= */
  IV IoLINES_LEFT(io);           /* $- */
  char *IoTOP_NAME(io);          /* $^ */
  GV *IoTOP_GV(io);              /* $^ */
  char *IoFMT_NAME(io);          /* $~ */
  GV *IoFMT_GV(io);              /* $~ */
  char *IoBOTTOM_NAME(io);
  GV *IoBOTTOM_GV(io);
  char IoTYPE(io);
  U8 IoFLAGS(io);

 =for apidoc_sections $io_scn, $formats_section
=for apidoc_section $reports
=for apidoc Amh|IV|IoLINES|IO *io
=for apidoc Amh|IV|IoPAGE|IO *io
=for apidoc Amh|IV|IoPAGE_LEN|IO *io
=for apidoc Amh|IV|IoLINES_LEFT|IO *io
=for apidoc Amh|char *|IoTOP_NAME|IO *io
=for apidoc Amh|GV *|IoTOP_GV|IO *io
=for apidoc Amh|char *|IoFMT_NAME|IO *io
=for apidoc Amh|GV *|IoFMT_GV|IO *io
=for apidoc Amh|char *|IoBOTTOM_NAME|IO *io
=for apidoc Amh|GV *|IoBOTTOM_GV|IO *io
=for apidoc_section $io
=for apidoc Amh|char|IoTYPE|IO *io
=for apidoc Amh|U8|IoFLAGS|IO *io

Most of these are involved with L<formats|perlform>.

IoFLAGs() may contain a combination of flags, the most interesting of
which are C<IOf_FLUSH> (C<$|>) for autoflush and C<IOf_UNTAINT>,
settable with L<< IO::Handle's untaint() method|IO::Handle/"$io->untaint" >>.

=for apidoc Amnh||IOf_FLUSH
=for apidoc Amnh||IOf_UNTAINT

The IO object may also contains a directory handle:

  DIR *IoDIRP(io);

=for apidoc Amh|DIR *|IoDIRP|IO *io

suitable for use with PerlDir_read() etc.

All of these accessors macros are lvalues, there are no distinct
C<_set()> macros to modify the members of the IO object.

=head2 Double-Typed SVs

Scalar variables normally contain only one type of value, an integer,
double, pointer, or reference.  Perl will automatically convert the
actual scalar data from the stored type into the requested type.

Some scalar variables contain more than one type of scalar data.  For
example, the variable C<$!> contains either the numeric value of C<errno>
or its string equivalent from either C<strerror> or C<sys_errlist[]>.

To force multiple data values into an SV, you must do two things: use the
C<sv_set*v> routines to add the additional scalar type, then set a flag
so that Perl will believe it contains more than one type of data.  The
four macros to set the flags are:

	SvIOK_on
	SvNOK_on
	SvPOK_on
	SvROK_on

The particular macro you must use depends on which C<sv_set*v> routine
you called first.  This is because every C<sv_set*v> routine turns on
only the bit for the particular type of data being set, and turns off
all the rest.

For example, to create a new Perl variable called "dberror" that contains
both the numeric and descriptive string error values, you could use the
following code:

    extern int  dberror;
    extern char *dberror_list;

    SV* sv = get_sv("dberror", GV_ADD);
    sv_setiv(sv, (IV) dberror);
    sv_setpv(sv, dberror_list[dberror]);
    SvIOK_on(sv);

If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.

=head2 Read-Only Values

In Perl 5.16 and earlier, copy-on-write (see the next section) shared a
flag bit with read-only scalars.  So the only way to test whether
C<sv_setsv>, etc., will raise a "Modification of a read-only value" error
in those versions is:

    SvREADONLY(sv) && !SvIsCOW(sv)

Under Perl 5.18 and later, SvREADONLY only applies to read-only variables,
and, under 5.20, copy-on-write scalars can also be read-only, so the above
check is incorrect.  You just want:

    SvREADONLY(sv)

If you need to do this check often, define your own macro like this:

    #if PERL_VERSION >= 18
    # define SvTRULYREADONLY(sv) SvREADONLY(sv)
    #else
    # define SvTRULYREADONLY(sv) (SvREADONLY(sv) && !SvIsCOW(sv))
    #endif

=head2 Copy on Write

Perl implements a copy-on-write (COW) mechanism for scalars, in which
string copies are not immediately made when requested, but are deferred
until made necessary by one or the other scalar changing.  This is mostly
transparent, but one must take care not to modify string buffers that are
shared by multiple SVs.

You can test whether an SV is using copy-on-write with C<SvIsCOW(sv)>.

You can force an SV to make its own copy of its string buffer by calling C<sv_force_normal(sv)> or SvPV_force_nolen(sv).

If you want to make the SV drop its string buffer, use
C<sv_force_normal_flags(sv, SV_COW_DROP_PV)> or simply
C<sv_setsv(sv, NULL)>.

All of these functions will croak on read-only scalars (see the previous
section for more on those).

To test that your code is behaving correctly and not modifying COW buffers,
on systems that support L<mmap(2)> (i.e., Unix) you can configure perl with
C<-Accflags=-DPERL_DEBUG_READONLY_COW> and it will turn buffer violations
into crashes.  You will find it to be marvellously slow, so you may want to
skip perl's own tests.

=head2 Magic Variables

[This section still under construction.  Ignore everything here.  Post no
bills.  Everything not permitted is forbidden.]

Any SV may be magical, that is, it has special features that a normal
SV does not have.  These features are stored in the SV structure in a
linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.

    struct magic {
        MAGIC*      mg_moremagic;
        MGVTBL*     mg_virtual;
        U16         mg_private;
        char        mg_type;
        U8          mg_flags;
        I32         mg_len;
        SV*         mg_obj;
        char*       mg_ptr;
    };

Note this is current as of patchlevel 0, and could change at any time.

=head2 Assigning Magic

Perl adds magic to an SV using the sv_magic function:

  void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);

The C<sv> argument is a pointer to the SV that is to acquire a new magical
feature.

If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
convert C<sv> to type C<SVt_PVMG>.
Perl then continues by adding new magic
to the beginning of the linked list of magical features.  Any prior entry
of the same type of magic is deleted.  Note that this can be overridden,
and multiple instances of the same type of magic can be associated with an
SV.

The C<name> and C<namlen> arguments are used to associate a string with
the magic, typically the name of a variable.  C<namlen> is stored in the
C<mg_len> field and if C<name> is non-null then either a C<savepvn> copy of
C<name> or C<name> itself is stored in the C<mg_ptr> field, depending on
whether C<namlen> is greater than zero or equal to zero respectively.  As a
special case, if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
to contain an C<SV*> and is stored as-is with its REFCNT incremented.

The sv_magic function uses C<how> to determine which, if any, predefined
"Magic Virtual Table" should be assigned to the C<mg_virtual> field.
See the L</Magic Virtual Tables> section below.  The C<how> argument is also
stored in the C<mg_type> field.  The value of
C<how> should be chosen from the set of macros
C<PERL_MAGIC_foo> found in F<perl.h>.  Note that before
these macros were added, Perl internals used to directly use character
literals, so you may occasionally come across old code or documentation
referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.

The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
structure.  If it is not the same as the C<sv> argument, the reference
count of the C<obj> object is incremented.  If it is the same, or if
the C<how> argument is C<PERL_MAGIC_arylen>, C<PERL_MAGIC_regdatum>,
C<PERL_MAGIC_regdata>, or if it is a NULL pointer, then C<obj> is merely
stored, without the reference count being incremented.

See also C<sv_magicext> in L<perlapi> for a more flexible way to add magic
to an SV.

There is also a function to add magic to an C<HV>:

    void hv_magic(HV *hv, GV *gv, int how);

This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.

To remove the magic from an SV, call the function sv_unmagic:

    int sv_unmagic(SV *sv, int type);

The C<type> argument should be equal to the C<how> value when the C<SV>
was initially made magical.

However, note that C<sv_unmagic> removes all magic of a certain C<type> from the
C<SV>.  If you want to remove only certain
magic of a C<type> based on the magic
virtual table, use C<sv_unmagicext> instead:

    int sv_unmagicext(SV *sv, int type, MGVTBL *vtbl);

=head2 Magic Virtual Tables

The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
C<MGVTBL>, which is a structure of function pointers and stands for
"Magic Virtual Table" to handle the various operations that might be
applied to that variable.

=for apidoc_section $magic
=for apidoc Ayh||MGVTBL

The C<MGVTBL> has five (or sometimes eight) pointers to the following
routine types:

    int  (*svt_get)  (pTHX_ SV* sv, MAGIC* mg);
    int  (*svt_set)  (pTHX_ SV* sv, MAGIC* mg);
    U32  (*svt_len)  (pTHX_ SV* sv, MAGIC* mg);
    int  (*svt_clear)(pTHX_ SV* sv, MAGIC* mg);
    int  (*svt_free) (pTHX_ SV* sv, MAGIC* mg);

    int  (*svt_copy) (pTHX_ SV *sv, MAGIC* mg, SV *nsv,
                                          const char *name, I32 namlen);
    int  (*svt_dup)  (pTHX_ MAGIC *mg, CLONE_PARAMS *param);
    int  (*svt_local)(pTHX_ SV *nsv, MAGIC *mg);


This MGVTBL structure is set at compile-time in F<perl.h> and there are
currently 32 types.  These different structures contain pointers to various
routines that perform additional actions depending on which function is
being called.

   Function pointer    Action taken
   ----------------    ------------
   svt_get             Do something before the value of the SV is
                       retrieved.
   svt_set             Do something after the SV is assigned a value.
   svt_len             Report on the SV's length.
   svt_clear           Clear something the SV represents.
   svt_free            Free any extra storage associated with the SV.

   svt_copy            copy tied variable magic to a tied element
   svt_dup             duplicate a magic structure during thread cloning
   svt_local           copy magic to local value during 'local'

For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
to an C<mg_type> of C<PERL_MAGIC_sv>) contains:

    { magic_get, magic_set, magic_len, 0, 0 }

Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
if a get operation is being performed, the routine C<magic_get> is
called.  All the various routines for the various magical types begin
with C<magic_>.  NOTE: the magic routines are not considered part of
the Perl API, and may not be exported by the Perl library.

The last three slots are a recent addition, and for source code
compatibility they are only checked for if one of the three flags
C<MGf_COPY>, C<MGf_DUP>, or C<MGf_LOCAL> is set in mg_flags.
This means that most code can continue declaring
a vtable as a 5-element value.  These three are
currently used exclusively by the threading code, and are highly subject
to change.

=for apidoc_section $magic
=for apidoc  Amnh||MGf_COPY
=for apidoc_item ||MGf_DUP
=for apidoc_item ||MGf_LOCAL

The current kinds of Magic Virtual Tables are:

=for comment
This table is generated by regen/mg_vtable.pl.  Any changes made here
will be lost.

=for mg_vtable.pl begin

 mg_type
 (old-style char and macro)   MGVTBL         Type of magic
 --------------------------   ------         -------------
 \0 PERL_MAGIC_sv             vtbl_sv        Special scalar variable
 #  PERL_MAGIC_arylen         vtbl_arylen    Array length ($#ary)
 %  PERL_MAGIC_rhash          (none)         Extra data for restricted
                                             hashes
 *  PERL_MAGIC_debugvar       vtbl_debugvar  $DB::single, signal, trace
                                             vars
 .  PERL_MAGIC_pos            vtbl_pos       pos() lvalue
 :  PERL_MAGIC_symtab         (none)         Extra data for symbol
                                             tables
 <  PERL_MAGIC_backref        vtbl_backref   For weak ref data
 @  PERL_MAGIC_arylen_p       (none)         To move arylen out of XPVAV
 B  PERL_MAGIC_bm             vtbl_regexp    Boyer-Moore
                                             (fast string search)
 c  PERL_MAGIC_overload_table vtbl_ovrld     Holds overload table
                                             (AMT) on stash
 D  PERL_MAGIC_regdata        vtbl_regdata   Regex match position data
                                             (@+ and @- vars)
 d  PERL_MAGIC_regdatum       vtbl_regdatum  Regex match position data
                                             element
 E  PERL_MAGIC_env            vtbl_env       %ENV hash
 e  PERL_MAGIC_envelem        vtbl_envelem   %ENV hash element
 f  PERL_MAGIC_fm             vtbl_regexp    Formline
                                             ('compiled' format)
 g  PERL_MAGIC_regex_global   vtbl_mglob     m//g target
 H  PERL_MAGIC_hints          vtbl_hints     %^H hash
 h  PERL_MAGIC_hintselem      vtbl_hintselem %^H hash element
 I  PERL_MAGIC_isa            vtbl_isa       @ISA array
 i  PERL_MAGIC_isaelem        vtbl_isaelem   @ISA array element
 k  PERL_MAGIC_nkeys          vtbl_nkeys     scalar(keys()) lvalue
 L  PERL_MAGIC_dbfile         (none)         Debugger %_<filename
 l  PERL_MAGIC_dbline         vtbl_dbline    Debugger %_<filename
                                             element
 N  PERL_MAGIC_shared         (none)         Shared between threads
 n  PERL_MAGIC_shared_scalar  (none)         Shared between threads
 o  PERL_MAGIC_collxfrm       vtbl_collxfrm  Locale transformation
 P  PERL_MAGIC_tied           vtbl_pack      Tied array or hash
 p  PERL_MAGIC_tiedelem       vtbl_packelem  Tied array or hash element
 q  PERL_MAGIC_tiedscalar     vtbl_packelem  Tied scalar or handle
 r  PERL_MAGIC_qr             vtbl_regexp    Precompiled qr// regex
 S  PERL_MAGIC_sig            vtbl_sig       %SIG hash
 s  PERL_MAGIC_sigelem        vtbl_sigelem   %SIG hash element
 t  PERL_MAGIC_taint          vtbl_taint     Taintedness
 U  PERL_MAGIC_uvar           vtbl_uvar      Available for use by
                                             extensions
 u  PERL_MAGIC_uvar_elem      (none)         Reserved for use by
                                             extensions
 V  PERL_MAGIC_vstring        (none)         SV was vstring literal
 v  PERL_MAGIC_vec            vtbl_vec       vec() lvalue
 w  PERL_MAGIC_utf8           vtbl_utf8      Cached UTF-8 information
 X  PERL_MAGIC_destruct       vtbl_destruct  destruct callback
 x  PERL_MAGIC_substr         vtbl_substr    substr() lvalue
 Y  PERL_MAGIC_nonelem        vtbl_nonelem   Array element that does not
                                             exist
 y  PERL_MAGIC_defelem        vtbl_defelem   Shadow "foreach" iterator
                                             variable / smart parameter
                                             vivification
 Z  PERL_MAGIC_hook           vtbl_hook      %{^HOOK} hash
 z  PERL_MAGIC_hookelem       vtbl_hookelem  %{^HOOK} hash element
 \  PERL_MAGIC_lvref          vtbl_lvref     Lvalue reference
                                             constructor
 ]  PERL_MAGIC_checkcall      vtbl_checkcall Inlining/mutation of call
                                             to this CV
 ^  PERL_MAGIC_extvalue       (none)         Value magic available for
                                             use by extensions
 ~  PERL_MAGIC_ext            (none)         Variable magic available
                                             for use by extensions


=for apidoc_section $magic
=for apidoc AmnhU||PERL_MAGIC_arylen
=for apidoc_item ||PERL_MAGIC_arylen_p
=for apidoc_item ||PERL_MAGIC_backref
=for apidoc_item ||PERL_MAGIC_bm
=for apidoc_item ||PERL_MAGIC_checkcall
=for apidoc_item ||PERL_MAGIC_collxfrm
=for apidoc_item ||PERL_MAGIC_dbfile
=for apidoc_item ||PERL_MAGIC_dbline
=for apidoc_item ||PERL_MAGIC_debugvar
=for apidoc_item ||PERL_MAGIC_defelem
=for apidoc_item ||PERL_MAGIC_destruct
=for apidoc_item ||PERL_MAGIC_env
=for apidoc_item ||PERL_MAGIC_envelem
=for apidoc_item ||PERL_MAGIC_ext
=for apidoc_item ||PERL_MAGIC_extvalue
=for apidoc_item ||PERL_MAGIC_fm
=for apidoc_item ||PERL_MAGIC_hints
=for apidoc_item ||PERL_MAGIC_hintselem
=for apidoc_item ||PERL_MAGIC_hook
=for apidoc_item ||PERL_MAGIC_hookelem
=for apidoc_item ||PERL_MAGIC_isa
=for apidoc_item ||PERL_MAGIC_isaelem
=for apidoc_item ||PERL_MAGIC_lvref
=for apidoc_item ||PERL_MAGIC_nkeys
=for apidoc_item ||PERL_MAGIC_nonelem
=for apidoc_item ||PERL_MAGIC_overload_table
=for apidoc_item ||PERL_MAGIC_pos
=for apidoc_item ||PERL_MAGIC_qr
=for apidoc_item ||PERL_MAGIC_regdata
=for apidoc_item ||PERL_MAGIC_regdatum
=for apidoc_item ||PERL_MAGIC_regex_global
=for apidoc_item ||PERL_MAGIC_rhash
=for apidoc_item ||PERL_MAGIC_shared
=for apidoc_item ||PERL_MAGIC_shared_scalar
=for apidoc_item ||PERL_MAGIC_sig
=for apidoc_item ||PERL_MAGIC_sigelem
=for apidoc_item ||PERL_MAGIC_substr
=for apidoc_item ||PERL_MAGIC_sv
=for apidoc_item ||PERL_MAGIC_symtab
=for apidoc_item ||PERL_MAGIC_taint
=for apidoc_item ||PERL_MAGIC_tied
=for apidoc_item ||PERL_MAGIC_tiedelem
=for apidoc_item ||PERL_MAGIC_tiedscalar
=for apidoc_item ||PERL_MAGIC_utf8
=for apidoc_item ||PERL_MAGIC_uvar
=for apidoc_item ||PERL_MAGIC_uvar_elem
=for apidoc_item ||PERL_MAGIC_vec
=for apidoc_item ||PERL_MAGIC_vstring

=for mg_vtable.pl end

When an uppercase and lowercase letter both exist in the table, then the
uppercase letter is typically used to represent some kind of composite type
(a list or a hash), and the lowercase letter is used to represent an element
of that composite type.  Some internals code makes use of this case
relationship.  However, 'v' and 'V' (vec and v-string) are in no way related.

The C<PERL_MAGIC_ext>, C<PERL_MAGIC_extvalue> and C<PERL_MAGIC_uvar> magic types
are defined specifically for use by extensions and will not be used by perl
itself.  Extensions can use C<PERL_MAGIC_ext> or C<PERL_MAGIC_extvalue> magic to
'attach' private information to variables (typically objects).  This is
especially useful because there is no way for normal perl code to corrupt this
private information (unlike using extra elements of a hash object).
C<PERL_MAGIC_extvalue> is value magic (unlike C<PERL_MAGIC_ext> and
C<PERL_MAGIC_uvar>) meaning that on localization the new value will not be
magical.

Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
C function any time a scalar's value is used or changed.  The C<MAGIC>'s
C<mg_ptr> field points to a C<ufuncs> structure:

    struct ufuncs {
        I32 (*uf_val)(pTHX_ IV, SV*);
        I32 (*uf_set)(pTHX_ IV, SV*);
        IV uf_index;
    };

When the SV is read from or written to, the C<uf_val> or C<uf_set>
function will be called with C<uf_index> as the first arg and a pointer to
the SV as the second.  A simple example of how to add C<PERL_MAGIC_uvar>
magic is shown below.  Note that the ufuncs structure is copied by
sv_magic, so you can safely allocate it on the stack.

    void
    Umagic(sv)
        SV *sv;
    PREINIT:
        struct ufuncs uf;
    CODE:
        uf.uf_val   = &my_get_fn;
        uf.uf_set   = &my_set_fn;
        uf.uf_index = 0;
        sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));

Attaching C<PERL_MAGIC_uvar> to arrays is permissible but has no effect.

For hashes there is a specialized hook that gives control over hash
keys (but not values).  This hook calls C<PERL_MAGIC_uvar> 'get' magic
if the "set" function in the C<ufuncs> structure is NULL.  The hook
is activated whenever the hash is accessed with a key specified as
an C<SV> through the functions C<hv_store_ent>, C<hv_fetch_ent>,
C<hv_delete_ent>, and C<hv_exists_ent>.  Accessing the key as a string
through the functions without the C<..._ent> suffix circumvents the
hook.  See L<Hash::Util::FieldHash/GUTS> for a detailed description.

Note that because multiple extensions may be using C<PERL_MAGIC_ext>
or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
extra care to avoid conflict.  Typically only using the magic on
objects blessed into the same class as the extension is sufficient.
For C<PERL_MAGIC_ext> magic, it is usually a good idea to define an
C<MGVTBL>, even if all its fields will be C<0>, so that individual
C<MAGIC> pointers can be identified as a particular kind of magic
using their magic virtual table.  C<mg_findext> provides an easy way
to do that:

    STATIC MGVTBL my_vtbl = { 0, 0, 0, 0, 0, 0, 0, 0 };

    MAGIC *mg;
    if ((mg = mg_findext(sv, PERL_MAGIC_ext, &my_vtbl))) {
        /* this is really ours, not another module's PERL_MAGIC_ext */
        my_priv_data_t *priv = (my_priv_data_t *)mg->mg_ptr;
        ...
    }

Also note that the C<sv_set*()> and C<sv_cat*()> functions described
earlier do B<not> invoke 'set' magic on their targets.  This must
be done by the user either by calling the C<SvSETMAGIC()> macro after
calling these functions, or by using one of the C<sv_set*_mg()> or
C<sv_cat*_mg()> functions.  Similarly, generic C code must call the
C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
obtained from external sources in functions that don't handle magic.
See L<perlapi> for a description of these functions.
For example, calls to the C<sv_cat*()> functions typically need to be
followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
since their implementation handles 'get' magic.

=head2 Finding Magic

    MAGIC *mg_find(SV *sv, int type); /* Finds the magic pointer of that
                                       * type */

This routine returns a pointer to a C<MAGIC> structure stored in the SV.
If the SV does not have that magical
feature, C<NULL> is returned.  If the
SV has multiple instances of that magical feature, the first one will be
returned.  C<mg_findext> can be used
to find a C<MAGIC> structure of an SV
based on both its magic type and its magic virtual table:

    MAGIC *mg_findext(SV *sv, int type, MGVTBL *vtbl);

Also, if the SV passed to C<mg_find> or C<mg_findext> is not of type
SVt_PVMG, Perl may core dump.

    int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);

This routine checks to see what types of magic C<sv> has.  If the mg_type
field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
the mg_type field is changed to be the lowercase letter.

=head2 Understanding the Magic of Tied Hashes and Arrays

Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
magic type.

WARNING: As of the 5.004 release, proper usage of the array and hash
access functions requires understanding a few caveats.  Some
of these caveats are actually considered bugs in the API, to be fixed
in later releases, and are bracketed with [MAYCHANGE] below.  If
you find yourself actually applying such information in this section, be
aware that the behavior may change in the future, umm, without warning.

The perl tie function associates a variable with an object that implements
the various GET, SET, etc methods.  To perform the equivalent of the perl
tie function from an XSUB, you must mimic this behaviour.  The code below
carries out the necessary steps -- firstly it creates a new hash, and then
creates a second hash which it blesses into the class which will implement
the tie methods.  Lastly it ties the two hashes together, and returns a
reference to the new tied hash.  Note that the code below does NOT call the
TIEHASH method in the MyTie class -
see L</Calling Perl Routines from within C Programs> for details on how
to do this.

    SV*
    mytie()
    PREINIT:
        HV *hash;
        HV *stash;
        SV *tie;
    CODE:
        hash = newHV();
        tie = newRV_noinc((SV*)newHV());
        stash = gv_stashpv("MyTie", GV_ADD);
        sv_bless(tie, stash);
        hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
        SvREFCNT_dec(tie); /* hv_magic() increases tie ref count */
        RETVAL = newRV_noinc(hash);
    OUTPUT:
        RETVAL

The C<av_store> function, when given a tied array argument, merely
copies the magic of the array onto the value to be "stored", using
C<mg_copy>.  It may also return NULL, indicating that the value did not
actually need to be stored in the array.  [MAYCHANGE] After a call to
C<av_store> on a tied array, the caller will usually need to call
C<mg_set(val)> to actually invoke the perl level "STORE" method on the
TIEARRAY object.  If C<av_store> did return NULL, a call to
C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
leak. [/MAYCHANGE]

The previous paragraph is applicable verbatim to tied hash access using the
C<hv_store> and C<hv_store_ent> functions as well.

C<av_fetch> and the corresponding hash functions C<hv_fetch> and
C<hv_fetch_ent> actually return an undefined mortal value whose magic
has been initialized using C<mg_copy>.  Note the value so returned does not
need to be deallocated, as it is already mortal.  [MAYCHANGE] But you will
need to call C<mg_get()> on the returned value in order to actually invoke
the perl level "FETCH" method on the underlying TIE object.  Similarly,
you may also call C<mg_set()> on the return value after possibly assigning
a suitable value to it using C<sv_setsv>,  which will invoke the "STORE"
method on the TIE object. [/MAYCHANGE]

[MAYCHANGE]
In other words, the array or hash fetch/store functions don't really
fetch and store actual values in the case of tied arrays and hashes.  They
merely call C<mg_copy> to attach magic to the values that were meant to be
"stored" or "fetched".  Later calls to C<mg_get> and C<mg_set> actually
do the job of invoking the TIE methods on the underlying objects.  Thus
the magic mechanism currently implements a kind of lazy access to arrays
and hashes.

Currently (as of perl version 5.004), use of the hash and array access
functions requires the user to be aware of whether they are operating on
"normal" hashes and arrays, or on their tied variants.  The API may be
changed to provide more transparent access to both tied and normal data
types in future versions.
[/MAYCHANGE]

You would do well to understand that the TIEARRAY and TIEHASH interfaces
are mere sugar to invoke some perl method calls while using the uniform hash
and array syntax.  The use of this sugar imposes some overhead (typically
about two to four extra opcodes per FETCH/STORE operation, in addition to
the creation of all the mortal variables required to invoke the methods).
This overhead will be comparatively small if the TIE methods are themselves
substantial, but if they are only a few statements long, the overhead
will not be insignificant.

=head2 Localizing changes

Perl has a very handy construction

  {
    local $var = 2;
    ...
  }

This construction is I<approximately> equivalent to

  {
    my $oldvar = $var;
    $var = 2;
    ...
    $var = $oldvar;
  }

The biggest difference is that the first construction would
reinstate the initial value of $var, irrespective of how control exits
the block: C<goto>, C<return>, C<die>/C<eval>, etc.  It is a little bit
more efficient as well.

There is a way to achieve a similar task from C via Perl API: create a
I<pseudo-block>, and arrange for some changes to be automatically
undone at the end of it, either explicit, or via a non-local exit (via
die()).  A I<block>-like construct is created by a pair of
C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
Such a construct may be created specially for some important localized
task, or an existing one (like boundaries of enclosing Perl
subroutine/block, or an existing pair for freeing TMPs) may be
used.  (In the second case the overhead of additional localization must
be almost negligible.)  Note that any XSUB is automatically enclosed in
an C<ENTER>/C<LEAVE> pair.

Inside such a I<pseudo-block> the following service is available:

=over 4

=item C<SAVEINT(int i)>

=item C<SAVEIV(IV i)>

=item C<SAVEI32(I32 i)>

=item C<SAVELONG(long i)>

=item C<SAVEI8(I8 i)>

=item C<SAVEI16(I16 i)>

=item C<SAVEBOOL(int i)>

=item C<SAVESTRLEN(STRLEN i)>

These macros arrange things to restore the value of integer variable
C<i> at the end of the enclosing I<pseudo-block>.

=for apidoc_section $callback
=for apidoc Amh||SAVEINT|int i
=for apidoc Amh||SAVEIV|IV i
=for apidoc Amh||SAVEI32|I32 i
=for apidoc Amh||SAVELONG|long i
=for apidoc Amh||SAVEI8|I8 i
=for apidoc Amh||SAVEI16|I16 i
=for apidoc Amh||SAVEBOOL|bool i
=for apidoc Amh||SAVESTRLEN|STRLEN i

=item C<SAVESPTR(s)>

=item C<SAVEPPTR(p)>

These macros arrange things to restore the value of pointers C<s> and
C<p>.  C<s> must be a pointer of a type which survives conversion to
C<SV*> and back, C<p> should be able to survive conversion to C<char*>
and back.

=for apidoc Amh||SAVESPTR|SV * s
=for apidoc Amh||SAVEPPTR|char * p

=item C<SAVERCPV(char **ppv)>

This macro arranges to restore the value of a C<char *> variable which
was allocated with a call to C<rcpv_new()> to its previous state when
the current pseudo block is completed. The pointer stored in C<*ppv> at
the time of the call will be refcount incremented and stored on the save
stack. Later when the current I<pseudo-block> is completed the value
stored in C<*ppv> will be refcount decremented, and the previous value
restored from the savestack which will also be refcount decremented.

This is the C<RCPV> equivalent of C<SAVEGENERICSV()>.

=for apidoc Amh||SAVERCPV|char *pv

=item C<SAVEGENERICSV(SV **psv)>

This macro arranges to restore the value of a C<SV *> variable to its
previous state when the current pseudo block is completed. The pointer
stored in C<*psv> at the time of the call will be refcount incremented
and stored on the save stack. Later when the current I<pseudo-block> is
completed the value stored in C<*ppv> will be refcount decremented, and
the previous value restored from the savestack which will also be refcount
decremented. This the C equivalent of C<local $sv>.

=for apidoc Amh||SAVEGENERICSV|char **psv

=item C<SAVEFREESV(SV *sv)>

The refcount of C<sv> will be decremented at the end of
I<pseudo-block>.  This is similar to C<sv_2mortal> in that it is also a
mechanism for doing a delayed C<SvREFCNT_dec>.  However, while C<sv_2mortal>
extends the lifetime of C<sv> until the beginning of the next statement,
C<SAVEFREESV> extends it until the end of the enclosing scope.  These
lifetimes can be wildly different.

Also compare C<SAVEMORTALIZESV>.

=for apidoc Amh||SAVEFREESV|SV* sv

=item C<SAVEMORTALIZESV(SV *sv)>

Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
scope instead of decrementing its reference count.  This usually has the
effect of keeping C<sv> alive until the statement that called the currently
live scope has finished executing.

=for apidoc Amh||SAVEMORTALIZESV|SV* sv

=item C<SAVEFREEOP(OP *op)>

The C<OP *> is C<op_free()>ed at the end of I<pseudo-block>.

=for apidoc Amh||SAVEFREEOP|OP *op

=item C<SAVEFREEPV(p)>

The chunk of memory which is pointed to by C<p> is C<Safefree()>ed at the
end of the current I<pseudo-block>.

=for apidoc Amh||SAVEFREEPV|char *pv

=item C<SAVEFREERCPV(char *pv)>

Ensures that a C<char *> which was created by a call to C<rcpv_new()> is
C<rcpv_free()>ed at the end of the current I<pseudo-block>.

This is the RCPV equivalent of C<SAVEFREESV()>.

=for apidoc Amh||SAVEFREERCPV|char *pv

=item C<SAVECLEARSV(SV *sv)>

Clears a slot in the current scratchpad which corresponds to C<sv> at
the end of I<pseudo-block>.

=item C<SAVEDELETE(HV *hv, char *key, I32 length)>

The key C<key> of C<hv> is deleted at the end of I<pseudo-block>.  The
string pointed to by C<key> is Safefree()ed.  If one has a I<key> in
short-lived storage, the corresponding string may be reallocated like
this:

  SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));

=for apidoc Amh||SAVEDELETE|HV * hv|char * key|I32 length

=item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>

At the end of I<pseudo-block> the function C<f> is called with the
only argument C<p> which may be NULL.

=for apidoc Ayh||DESTRUCTORFUNC_NOCONTEXT_t
=for apidoc Amh||SAVEDESTRUCTOR|DESTRUCTORFUNC_NOCONTEXT_t f|void *p

=item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>

At the end of I<pseudo-block> the function C<f> is called with the
implicit context argument (if any), and C<p> which may be NULL.

Note the I<end of the current pseudo-block> may occur much later than
the I<end of the current statement>. You may wish to look at the
C<MORTALSVFUNC_X()> macro instead.

=for apidoc Ayh||DESTRUCTORFUNC_t
=for apidoc Amh||SAVEDESTRUCTOR_X|DESTRUCTORFUNC_t f|void *p

=item C<MORTALSVFUNC_X(SVFUNC_t f, SV *sv)>

At the end of I<the current statement> the function C<f> is called with
the implicit context argument (if any), and C<sv> which may be NULL.

Be aware that the parameter argument to the destructor function differs
from the related C<SAVEDESTRUCTOR_X()> in that it MUST be either NULL or
an C<SV*>.

Note the I<end of the current statement> may occur much before the
the I<end of the current pseudo-block>.  You may wish to look at the
C<SAVEDESTRUCTOR_X()> macro instead.

=for apidoc Amh||MORTALSVFUNC_X|SVFUNC_t f|SV *sv

=item C<MORTALDESTRUCTOR_SV(SV *coderef, SV *args)>

At the end of I<the current statement> the Perl function contained in
C<coderef> is called with the arguments provided (if any) in C<args>.
See the documentation for C<mortal_destructor_sv()> for details on
the C<args> parameter is handled.

Note the I<end of the current statement> may occur much before the
the I<end of the current pseudo-block>.  If you wish to call a perl
function at the end of the current pseudo block you should use the
C<SAVEDESTRUCTOR_X()> API instead, which will require you create a
C wrapper to call the Perl function.

=for apidoc Amh||MORTALDESTRUCTOR_SV|SV *coderef|SV *args

=item C<SAVESTACK_POS()>

The current offset on the Perl internal stack (cf. C<SP>) is restored
at the end of I<pseudo-block>.

=for apidoc Amh||SAVESTACK_POS

=back

The following API list contains functions, thus one needs to
provide pointers to the modifiable data explicitly (either C pointers,
or Perlish C<GV *>s).  Where the above macros take C<int>, a similar
function takes C<int *>.

Other macros above have functions implementing them, but its probably
best to just use the macro, and not those or the ones below.

=over 4

=item C<SV* save_scalar(GV *gv)>

=for apidoc save_scalar

Equivalent to Perl code C<local $gv>.

=item C<AV* save_ary(GV *gv)>

=for apidoc save_ary

=item C<HV* save_hash(GV *gv)>

=for apidoc save_hash

Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.

=item C<void save_item(SV *item)>

=for apidoc save_item

Duplicates the current value of C<SV>. On the exit from the current
C<ENTER>/C<LEAVE> I<pseudo-block> the value of C<SV> will be restored
using the stored value.  It doesn't handle magic.  Use C<save_scalar> if
magic is affected.

=item C<SV* save_svref(SV **sptr)>

=for apidoc save_svref

Similar to C<save_scalar>, but will reinstate an C<SV *>.

=item C<void save_aptr(AV **aptr)>

=item C<void save_hptr(HV **hptr)>

=for apidoc save_aptr
=for apidoc save_hptr

Similar to C<save_svref>, but localize C<AV *> and C<HV *>.

=back

The C<Alias> module implements localization of the basic types within the
I<caller's scope>.  People who are interested in how to localize things in
the containing scope should take a look there too.

=head1 Subroutines

=head2 XSUBs and the Argument Stack

The XSUB mechanism is a simple way for Perl programs to access C subroutines.
An XSUB routine will have a stack that contains the arguments from the Perl
program, and a way to map from the Perl data structures to a C equivalent.

The stack arguments are accessible through the C<ST(n)> macro, which returns
the C<n>'th stack argument.  Argument 0 is the first argument passed in the
Perl subroutine call.  These arguments are C<SV*>, and can be used anywhere
an C<SV*> is used.

Most of the time, output from the C routine can be handled through use of
the RETVAL and OUTPUT directives.  However, there are some cases where the
argument stack is not already long enough to handle all the return values.
An example is the POSIX tzname() call, which takes no arguments, but returns
two, the local time zone's standard and summer time abbreviations.

To handle this situation, the PPCODE directive is used and the stack is
extended using the macro:

    EXTEND(SP, num);

where C<SP> is the macro that represents the local copy of the stack pointer,
and C<num> is the number of elements the stack should be extended by.

Now that there is room on the stack, values can be pushed on it using C<PUSHs>
macro.  The pushed values will often need to be "mortal" (See
L</Reference Counts and Mortality>):

    PUSHs(sv_2mortal(newSViv(an_integer)))
    PUSHs(sv_2mortal(newSVuv(an_unsigned_integer)))
    PUSHs(sv_2mortal(newSVnv(a_double)))
    PUSHs(sv_2mortal(newSVpv("Some String",0)))
    /* Although the last example is better written as the more
     * efficient: */
    PUSHs(newSVpvs_flags("Some String", SVs_TEMP))

And now the Perl program calling C<tzname>, the two values will be assigned
as in:

    ($standard_abbrev, $summer_abbrev) = POSIX::tzname;

An alternate (and possibly simpler) method to pushing values on the stack is
to use the macro:

    XPUSHs(SV*)

This macro automatically adjusts the stack for you, if needed.  Thus, you
do not need to call C<EXTEND> to extend the stack.

Despite their suggestions in earlier versions of this document the macros
C<(X)PUSH[iunp]> are I<not> suited to XSUBs which return multiple results.
For that, either stick to the C<(X)PUSHs> macros shown above, or use the new
C<m(X)PUSH[iunp]> macros instead; see L</Putting a C value on Perl stack>.

For more information, consult L<perlxs> and L<perlxstut>.

=head2 Autoloading with XSUBs

If an AUTOLOAD routine is an XSUB, as with Perl subroutines, Perl puts the
fully-qualified name of the autoloaded subroutine in the $AUTOLOAD variable
of the XSUB's package.

But it also puts the same information in certain fields of the XSUB itself:

    HV *stash           = CvSTASH(cv);
    const char *subname = SvPVX(cv);
    STRLEN name_length  = SvCUR(cv); /* in bytes */
    U32 is_utf8         = SvUTF8(cv);

C<SvPVX(cv)> contains just the sub name itself, not including the package.
For an AUTOLOAD routine in UNIVERSAL or one of its superclasses,
C<CvSTASH(cv)> returns NULL during a method call on a nonexistent package.

B<Note>: Setting $AUTOLOAD stopped working in 5.6.1, which did not support
XS AUTOLOAD subs at all.  Perl 5.8.0 introduced the use of fields in the
XSUB itself.  Perl 5.16.0 restored the setting of $AUTOLOAD.  If you need
to support 5.8-5.14, use the XSUB's fields.

=head2 Calling Perl Routines from within C Programs

There are four routines that can be used to call a Perl subroutine from
within a C program.  These four are:

    I32  call_sv(SV*, I32);
    I32  call_pv(const char*, I32);
    I32  call_method(const char*, I32);
    I32  call_argv(const char*, I32, char**);

The routine most often used is C<call_sv>.  The C<SV*> argument
contains either the name of the Perl subroutine to be called, or a
reference to the subroutine.  The second argument consists of flags
that control the context in which the subroutine is called, whether
or not the subroutine is being passed arguments, how errors should be
trapped, and how to treat return values.

All four routines return the number of arguments that the subroutine returned
on the Perl stack.

These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
but those names are now deprecated; macros of the same name are provided for
compatibility.

When using any of these routines (except C<call_argv>), the programmer
must manipulate the Perl stack.  These include the following macros and
functions:

    dSP
    SP
    PUSHMARK()
    PUTBACK
    SPAGAIN
    ENTER
    SAVETMPS
    FREETMPS
    LEAVE
    XPUSH*()
    POP*()

For a detailed description of calling conventions from C to Perl,
consult L<perlcall>.

=head2 Putting a C value on Perl stack

A lot of opcodes (this is an elementary operation in the internal perl
stack machine) put an SV* on the stack.  However, as an optimization
the corresponding SV is (usually) not recreated each time.  The opcodes
reuse specially assigned SVs (I<target>s) which are (as a corollary)
not constantly freed/created.

Each of the targets is created only once (but see
L</Scratchpads and recursion> below), and when an opcode needs to put
an integer, a double, or a string on the stack, it just sets the
corresponding parts of its I<target> and puts the I<target> on stack.

The macro to put this target on stack is C<PUSHTARG>, and it is
directly used in some opcodes, as well as indirectly in zillions of
others, which use it via C<(X)PUSH[iunp]>.

Because the target is reused, you must be careful when pushing multiple
values on the stack.  The following code will not do what you think:

    XPUSHi(10);
    XPUSHi(20);

This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
At the end of the operation, the stack does not contain the values 10
and 20, but actually contains two pointers to C<TARG>, which we have set
to 20.

If you need to push multiple different values then you should either use
the C<(X)PUSHs> macros, or else use the new C<m(X)PUSH[iunp]> macros,
none of which make use of C<TARG>.  The C<(X)PUSHs> macros simply push an
SV* on the stack, which, as noted under L</XSUBs and the Argument Stack>,
will often need to be "mortal".  The new C<m(X)PUSH[iunp]> macros make
this a little easier to achieve by creating a new mortal for you (via
C<(X)PUSHmortal>), pushing that onto the stack (extending it if necessary
in the case of the C<mXPUSH[iunp]> macros), and then setting its value.
Thus, instead of writing this to "fix" the example above:

    XPUSHs(sv_2mortal(newSViv(10)))
    XPUSHs(sv_2mortal(newSViv(20)))

you can simply write:

    mXPUSHi(10)
    mXPUSHi(20)

On a related note, if you do use C<(X)PUSH[iunp]>, then you're going to
need a C<dTARG> in your variable declarations so that the C<*PUSH*>
macros can make use of the local variable C<TARG>.  See also
C<dTARGET> and C<dXSTARG>.

=head2 Scratchpads

The question remains on when the SVs which are I<target>s for opcodes
are created.  The answer is that they are created when the current
unit--a subroutine or a file (for opcodes for statements outside of
subroutines)--is compiled.  During this time a special anonymous Perl
array is created, which is called a scratchpad for the current unit.

A scratchpad keeps SVs which are lexicals for the current unit and are
targets for opcodes.  A previous version of this document
stated that one can deduce that an SV lives on a scratchpad
by looking on its flags: lexicals have C<SVs_PADMY> set, and
I<target>s have C<SVs_PADTMP> set.  But this has never been fully true.
C<SVs_PADMY> could be set on a variable that no longer resides in any pad.
While I<target>s do have C<SVs_PADTMP> set, it can also be set on variables
that have never resided in a pad, but nonetheless act like I<target>s.  As
of perl 5.21.5, the C<SVs_PADMY> flag is no longer used and is defined as
0.  C<SvPADMY()> now returns true for anything without C<SVs_PADTMP>.

=for apidoc_section $pad
=for apidoc Amnh||SVs_PADTMP
=for apidoc AmnhD||SVs_PADMY

The correspondence between OPs and I<target>s is not 1-to-1.  Different
OPs in the compile tree of the unit can use the same target, if this
would not conflict with the expected life of the temporary.

=head2 Scratchpads and recursion

In fact it is not 100% true that a compiled unit contains a pointer to
the scratchpad AV.  In fact it contains a pointer to an AV of
(initially) one element, and this element is the scratchpad AV.  Why do
we need an extra level of indirection?

The answer is B<recursion>, and maybe B<threads>.  Both
these can create several execution pointers going into the same
subroutine.  For the subroutine-child not write over the temporaries
for the subroutine-parent (lifespan of which covers the call to the
child), the parent and the child should have different
scratchpads.  (I<And> the lexicals should be separate anyway!)

So each subroutine is born with an array of scratchpads (of length 1).
On each entry to the subroutine it is checked that the current
depth of the recursion is not more than the length of this array, and
if it is, new scratchpad is created and pushed into the array.

The I<target>s on this scratchpad are C<undef>s, but they are already
marked with correct flags.

=head1 Memory Allocation

=head2 Allocation

All memory meant to be used with the Perl API functions should be manipulated
using the macros described in this section.  The macros provide the necessary
transparency between differences in the actual malloc implementation that is
used within perl.

The following three macros are used to initially allocate memory :

    Newx(pointer, number, type);
    Newxc(pointer, number, type, cast);
    Newxz(pointer, number, type);

The first argument C<pointer> should be the name of a variable that will
point to the newly allocated memory.

The second and third arguments C<number> and C<type> specify how many of
the specified type of data structure should be allocated.  The argument
C<type> is passed to C<sizeof>.  The final argument to C<Newxc>, C<cast>,
should be used if the C<pointer> argument is different from the C<type>
argument.

Unlike the C<Newx> and C<Newxc> macros, the C<Newxz> macro calls C<memzero>
to zero out all the newly allocated memory.

=head2 Reallocation

    Renew(pointer, number, type);
    Renewc(pointer, number, type, cast);
    Safefree(pointer)

These three macros are used to change a memory buffer size or to free a
piece of memory no longer needed.  The arguments to C<Renew> and C<Renewc>
match those of C<New> and C<Newc> with the exception of not needing the
"magic cookie" argument.

=head2 Moving

    Move(source, dest, number, type);
    Copy(source, dest, number, type);
    Zero(dest, number, type);

These three macros are used to move, copy, or zero out previously allocated
memory.  The C<source> and C<dest> arguments point to the source and
destination starting points.  Perl will move, copy, or zero out C<number>
instances of the size of the C<type> data structure (using the C<sizeof>
function).

=head1 PerlIO

The most recent development releases of Perl have been experimenting with
removing Perl's dependency on the "normal" standard I/O suite and allowing
other stdio implementations to be used.  This involves creating a new
abstraction layer that then calls whichever implementation of stdio Perl
was compiled with.  All XSUBs should now use the functions in the PerlIO
abstraction layer and not make any assumptions about what kind of stdio
is being used.

For a complete description of the PerlIO abstraction, consult L<perlapio>.

=head1 Compiled code

=head2 Code tree

Here we describe the internal form your code is converted to by
Perl.  Start with a simple example:

  $a = $b + $c;

This is converted to a tree similar to this one:

             assign-to
           /           \
          +             $a
        /   \
      $b     $c

(but slightly more complicated).  This tree reflects the way Perl
parsed your code, but has nothing to do with the execution order.
There is an additional "thread" going through the nodes of the tree
which shows the order of execution of the nodes.  In our simplified
example above it looks like:

     $b ---> $c ---> + ---> $a ---> assign-to

But with the actual compile tree for C<$a = $b + $c> it is different:
some nodes I<optimized away>.  As a corollary, though the actual tree
contains more nodes than our simplified example, the execution order
is the same as in our example.

=head2 Examining the tree

If you have your perl compiled for debugging (usually done with
C<-DDEBUGGING> on the C<Configure> command line), you may examine the
compiled tree by specifying C<-Dx> on the Perl command line.  The
output takes several lines per node, and for C<$b+$c> it looks like
this:

    5           TYPE = add  ===> 6
                TARG = 1
                FLAGS = (SCALAR,KIDS)
                {
                    TYPE = null  ===> (4)
                      (was rv2sv)
                    FLAGS = (SCALAR,KIDS)
                    {
    3                   TYPE = gvsv  ===> 4
                        FLAGS = (SCALAR)
                        GV = main::b
                    }
                }
                {
                    TYPE = null  ===> (5)
                      (was rv2sv)
                    FLAGS = (SCALAR,KIDS)
                    {
    4                   TYPE = gvsv  ===> 5
                        FLAGS = (SCALAR)
                        GV = main::c
                    }
                }

This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
not optimized away (one per number in the left column).  The immediate
children of the given node correspond to C<{}> pairs on the same level
of indentation, thus this listing corresponds to the tree:

                   add
                 /     \
               null    null
                |       |
               gvsv    gvsv

The execution order is indicated by C<===E<gt>> marks, thus it is C<3
4 5 6> (node C<6> is not included into above listing), i.e.,
C<gvsv gvsv add whatever>.

Each of these nodes represents an op, a fundamental operation inside the
Perl core.  The code which implements each operation can be found in the
F<pp*.c> files; the function which implements the op with type C<gvsv>
is C<pp_gvsv>, and so on.  As the tree above shows, different ops have
different numbers of children: C<add> is a binary operator, as one would
expect, and so has two children.  To accommodate the various different
numbers of children, there are various types of op data structure, and
they link together in different ways.

The simplest type of op structure is C<OP>: this has no children.  Unary
operators, C<UNOP>s, have one child, and this is pointed to by the
C<op_first> field.  Binary operators (C<BINOP>s) have not only an
C<op_first> field but also an C<op_last> field.  The most complex type of
op is a C<LISTOP>, which has any number of children.  In this case, the
first child is pointed to by C<op_first> and the last child by
C<op_last>.  The children in between can be found by iteratively
following the C<OpSIBLING> pointer from the first child to the last (but
see below).

=for apidoc_section $optree_construction
=for apidoc Ayh||OP
=for apidoc Ayh||BINOP
=for apidoc Ayh||LISTOP
=for apidoc Ayh||UNOP

There are also some other op types: a C<PMOP> holds a regular expression,
and has no children, and a C<LOOP> may or may not have children.  If the
C<op_children> field is non-zero, it behaves like a C<LISTOP>.  To
complicate matters, if a C<UNOP> is actually a C<null> op after
optimization (see L</Compile pass 2: context propagation>) it will still
have children in accordance with its former type.

=for apidoc Ayh||LOOP
=for apidoc Ayh||PMOP

Finally, there is a C<LOGOP>, or logic op. Like a C<LISTOP>, this has one
or more children, but it doesn't have an C<op_last> field: so you have to
follow C<op_first> and then the C<OpSIBLING> chain itself to find the
last child. Instead it has an C<op_other> field, which is comparable to
the C<op_next> field described below, and represents an alternate
execution path. Operators like C<and>, C<or> and C<?> are C<LOGOP>s. Note
that in general, C<op_other> may not point to any of the direct children
of the C<LOGOP>.

=for apidoc Ayh||LOGOP

Starting in version 5.21.2, perls built with the experimental
define C<-DPERL_OP_PARENT> add an extra boolean flag for each op,
C<op_moresib>.  When not set, this indicates that this is the last op in an
C<OpSIBLING> chain. This frees up the C<op_sibling> field on the last
sibling to point back to the parent op. Under this build, that field is
also renamed C<op_sibparent> to reflect its joint role. The macro
C<OpSIBLING(o)> wraps this special behaviour, and always returns NULL on
the last sibling.  With this build the C<op_parent(o)> function can be
used to find the parent of any op. Thus for forward compatibility, you
should always use the C<OpSIBLING(o)> macro rather than accessing
C<op_sibling> directly.

Another way to examine the tree is to use a compiler back-end module, such
as L<B::Concise>.

=head2 Compile pass 1: check routines

The tree is created by the compiler while I<yacc> code feeds it
the constructions it recognizes.  Since I<yacc> works bottom-up, so does
the first pass of perl compilation.

What makes this pass interesting for perl developers is that some
optimization may be performed on this pass.  This is optimization by
so-called "check routines".  The correspondence between node names
and corresponding check routines is described in F<opcode.pl> (do not
forget to run C<make regen_headers> if you modify this file).

A check routine is called when the node is fully constructed except
for the execution-order thread.  Since at this time there are no
back-links to the currently constructed node, one can do most any
operation to the top-level node, including freeing it and/or creating
new nodes above/below it.

The check routine returns the node which should be inserted into the
tree (if the top-level node was not modified, check routine returns
its argument).

By convention, check routines have names C<ck_*>.  They are usually
called from C<new*OP> subroutines (or C<convert>) (which in turn are
called from F<perly.y>).

=head2 Compile pass 1a: constant folding

Immediately after the check routine is called the returned node is
checked for being compile-time executable.  If it is (the value is
judged to be constant) it is immediately executed, and a I<constant>
node with the "return value" of the corresponding subtree is
substituted instead.  The subtree is deleted.

If constant folding was not performed, the execution-order thread is
created.

=head2 Compile pass 2: context propagation

When a context for a part of compile tree is known, it is propagated
down through the tree.  At this time the context can have 5 values
(instead of 2 for runtime context): void, boolean, scalar, list, and
lvalue.  In contrast with the pass 1 this pass is processed from top
to bottom: a node's context determines the context for its children.

Additional context-dependent optimizations are performed at this time.
Since at this moment the compile tree contains back-references (via
"thread" pointers), nodes cannot be free()d now.  To allow
optimized-away nodes at this stage, such nodes are null()ified instead
of free()ing (i.e. their type is changed to OP_NULL).

=head2 Compile pass 3: peephole optimization

After the compile tree for a subroutine (or for an C<eval> or a file)
is created, an additional pass over the code is performed.  This pass
is neither top-down or bottom-up, but in the execution order (with
additional complications for conditionals).  Optimizations performed
at this stage are subject to the same restrictions as in the pass 2.

Peephole optimizations are done by calling the function pointed to
by the global variable C<PL_peepp>.  By default, C<PL_peepp> just
calls the function pointed to by the global variable C<PL_rpeepp>.
By default, that performs some basic op fixups and optimisations along
the execution-order op chain, and recursively calls C<PL_rpeepp> for
each side chain of ops (resulting from conditionals).  Extensions may
provide additional optimisations or fixups, hooking into either the
per-subroutine or recursive stage, like this:

    static peep_t prev_peepp;
    static void my_peep(pTHX_ OP *o)
    {
        /* custom per-subroutine optimisation goes here */
        prev_peepp(aTHX_ o);
        /* custom per-subroutine optimisation may also go here */
    }
    BOOT:
        prev_peepp = PL_peepp;
        PL_peepp = my_peep;

    static peep_t prev_rpeepp;
    static void my_rpeep(pTHX_ OP *first)
    {
        OP *o = first, *t = first;
        for(; o = o->op_next, t = t->op_next) {
            /* custom per-op optimisation goes here */
            o = o->op_next;
            if (!o || o == t) break;
            /* custom per-op optimisation goes AND here */
        }
        prev_rpeepp(aTHX_ orig_o);
    }
    BOOT:
        prev_rpeepp = PL_rpeepp;
        PL_rpeepp = my_rpeep;

=for apidoc_section $optree_manipulation
=for apidoc Ayh||peep_t

=head2 Pluggable runops

The compile tree is executed in a runops function.  There are two runops
functions, in F<run.c> and in F<dump.c>.  C<Perl_runops_debug> is used
with DEBUGGING and C<Perl_runops_standard> is used otherwise.  For fine
control over the execution of the compile tree it is possible to provide
your own runops function.

It's probably best to copy one of the existing runops functions and
change it to suit your needs.  Then, in the BOOT section of your XS
file, add the line:

  PL_runops = my_runops;

=for apidoc_section $debugging
=for apidoc runops_debug
=for apidoc runops_standard
=for apidoc Amnh|runops_proc_t|PL_runops

This function should be as efficient as possible to keep your programs
running as fast as possible.

=head2 Compile-time scope hooks

As of perl 5.14 it is possible to hook into the compile-time lexical
scope mechanism using C<Perl_blockhook_register>.  This is used like
this:

    STATIC void my_start_hook(pTHX_ int full);
    STATIC BHK my_hooks;

    BOOT:
        BhkENTRY_set(&my_hooks, bhk_start, my_start_hook);
        Perl_blockhook_register(aTHX_ &my_hooks);

This will arrange to have C<my_start_hook> called at the start of
compiling every lexical scope.  The available hooks are:

=for apidoc_section $lexer
=for apidoc Ayh||BHK

=over 4

=item C<void bhk_start(pTHX_ int full)>

This is called just after starting a new lexical scope.  Note that Perl
code like

    if ($x) { ... }

creates two scopes: the first starts at the C<(> and has C<full == 1>,
the second starts at the C<{> and has C<full == 0>.  Both end at the
C<}>, so calls to C<start> and C<pre>/C<post_end> will match.  Anything
pushed onto the save stack by this hook will be popped just before the
scope ends (between the C<pre_> and C<post_end> hooks, in fact).

=item C<void bhk_pre_end(pTHX_ OP **o)>

This is called at the end of a lexical scope, just before unwinding the
stack.  I<o> is the root of the optree representing the scope; it is a
double pointer so you can replace the OP if you need to.

=item C<void bhk_post_end(pTHX_ OP **o)>

This is called at the end of a lexical scope, just after unwinding the
stack.  I<o> is as above.  Note that it is possible for calls to C<pre_>
and C<post_end> to nest, if there is something on the save stack that
calls string eval.

=item C<void bhk_eval(pTHX_ OP *const o)>

This is called just before starting to compile an C<eval STRING>, C<do
FILE>, C<require> or C<use>, after the eval has been set up.  I<o> is the
OP that requested the eval, and will normally be an C<OP_ENTEREVAL>,
C<OP_DOFILE> or C<OP_REQUIRE>.

=back

Once you have your hook functions, you need a C<BHK> structure to put
them in.  It's best to allocate it statically, since there is no way to
free it once it's registered.  The function pointers should be inserted
into this structure using the C<BhkENTRY_set> macro, which will also set
flags indicating which entries are valid.  If you do need to allocate
your C<BHK> dynamically for some reason, be sure to zero it before you
start.

Once registered, there is no mechanism to switch these hooks off, so if
that is necessary you will need to do this yourself.  An entry in C<%^H>
is probably the best way, so the effect is lexically scoped; however it
is also possible to use the C<BhkDISABLE> and C<BhkENABLE> macros to
temporarily switch entries on and off.  You should also be aware that
generally speaking at least one scope will have opened before your
extension is loaded, so you will see some C<pre>/C<post_end> pairs that
didn't have a matching C<start>.

=head1 Examining internal data structures with the C<dump> functions

To aid debugging, the source file F<dump.c> contains a number of
functions which produce formatted output of internal data structures.

The most commonly used of these functions is C<Perl_sv_dump>; it's used
for dumping SVs, AVs, HVs, and CVs.  The C<Devel::Peek> module calls
C<sv_dump> to produce debugging output from Perl-space, so users of that
module should already be familiar with its format.

C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
derivatives, and produces output similar to C<perl -Dx>; in fact,
C<Perl_dump_eval> will dump the main root of the code being evaluated,
exactly like C<-Dx>.

=for apidoc_section $debugging
=for apidoc dump_eval

Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
subroutines in a package like so: (Thankfully, these are all xsubs, so
there is no op tree)

=for apidoc_section $debugging
=for apidoc dump_sub

    (gdb) print Perl_dump_packsubs(PL_defstash)

    SUB attributes::bootstrap = (xsub 0x811fedc 0)

    SUB UNIVERSAL::can = (xsub 0x811f50c 0)

    SUB UNIVERSAL::isa = (xsub 0x811f304 0)

    SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)

    SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)

and C<Perl_dump_all>, which dumps all the subroutines in the stash and
the op tree of the main root.

=head1 How multiple interpreters and concurrency are supported

=head2 Background and MULTIPLICITY

=for apidoc_section $concurrency
=for apidoc Amnh||PERL_IMPLICIT_CONTEXT

The Perl interpreter can be regarded as a closed box: it has an API
for feeding it code or otherwise making it do things, but it also has
functions for its own use.  This smells a lot like an object, and
there is a way for you to build Perl so that you can have multiple
interpreters, with one interpreter represented either as a C structure,
or inside a thread-specific structure.  These structures contain all
the context, the state of that interpreter.

The macro that controls the major Perl build flavor is MULTIPLICITY.  The
MULTIPLICITY build has a C structure that packages all the interpreter
state, which is being passed to various perl functions as a "hidden"
first argument. MULTIPLICITY makes multi-threaded perls possible (with the
ithreads threading model, related to the macro USE_ITHREADS.)

PERL_IMPLICIT_CONTEXT is a legacy synonym for MULTIPLICITY.

=for apidoc_section $concurrency
=for apidoc Amnh||MULTIPLICITY

To see whether you have non-const data you can use a BSD (or GNU)
compatible C<nm>:

  nm libperl.a | grep -v ' [TURtr] '

If this displays any C<D> or C<d> symbols (or possibly C<C> or C<c>),
you have non-const data.  The symbols the C<grep> removed are as follows:
C<Tt> are I<text>, or code, the C<Rr> are I<read-only> (const) data,
and the C<U> is <undefined>, external symbols referred to.

The test F<t/porting/libperl.t> does this kind of symbol sanity
checking on C<libperl.a>.

All this obviously requires a way for the Perl internal functions to be
either subroutines taking some kind of structure as the first
argument, or subroutines taking nothing as the first argument.  To
enable these two very different ways of building the interpreter,
the Perl source (as it does in so many other situations) makes heavy
use of macros and subroutine naming conventions.

First problem: deciding which functions will be public API functions and
which will be private.  All functions whose names begin C<S_> are private
(think "S" for "secret" or "static").  All other functions begin with
"Perl_", but just because a function begins with "Perl_" does not mean it is
part of the API.  (See L</Internal
Functions>.)  The easiest way to be B<sure> a
function is part of the API is to find its entry in L<perlapi>.
If it exists in L<perlapi>, it's part of the API.  If it doesn't, and you
think it should be (i.e., you need it for your extension), submit an issue at
L<https://github.com/Perl/perl5/issues> explaining why you think it should be.

Second problem: there must be a syntax so that the same subroutine
declarations and calls can pass a structure as their first argument,
or pass nothing.  To solve this, the subroutines are named and
declared in a particular way.  Here's a typical start of a static
function used within the Perl guts:

  STATIC void
  S_incline(pTHX_ char *s)

STATIC becomes "static" in C, and may be #define'd to nothing in some
configurations in the future.

=for apidoc_section $directives
=for apidoc Ayh||STATIC

A public function (i.e. part of the internal API, but not necessarily
sanctioned for use in extensions) begins like this:

  void
  Perl_sv_setiv(pTHX_ SV* dsv, IV num)

C<pTHX_> is one of a number of macros (in F<perl.h>) that hide the
details of the interpreter's context.  THX stands for "thread", "this",
or "thingy", as the case may be.  (And no, George Lucas is not involved. :-)
The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
their variants.

=for apidoc_section $concurrency
=for apidoc Amnh||aTHX
=for apidoc Amnh||aTHX_
=for apidoc Amnh||dTHX
=for apidoc Amnh||pTHX
=for apidoc Amnh||pTHX_

When Perl is built without options that set MULTIPLICITY, there is no
first argument containing the interpreter's context.  The trailing underscore
in the pTHX_ macro indicates that the macro expansion needs a comma
after the context argument because other arguments follow it.  If
MULTIPLICITY is not defined, pTHX_ will be ignored, and the
subroutine is not prototyped to take the extra argument.  The form of the
macro without the trailing underscore is used when there are no additional
explicit arguments.

When a core function calls another, it must pass the context.  This
is normally hidden via macros.  Consider C<sv_setiv>.  It expands into
something like this:

    #ifdef MULTIPLICITY
      #define sv_setiv(a,b)      Perl_sv_setiv(aTHX_ a, b)
      /* can't do this for vararg functions, see below */
    #else
      #define sv_setiv           Perl_sv_setiv
    #endif

This works well, and means that XS authors can gleefully write:

    sv_setiv(foo, bar);

and still have it work under all the modes Perl could have been
compiled with.

This doesn't work so cleanly for varargs functions, though, as macros
imply that the number of arguments is known in advance.  Instead we
either need to spell them out fully, passing C<aTHX_> as the first
argument (the Perl core tends to do this with functions like
Perl_warner), or use a context-free version.

The context-free version of Perl_warner is called
Perl_warner_nocontext, and does not take the extra argument.  Instead
it does C<dTHX;> to get the context from thread-local storage.  We
C<#define warner Perl_warner_nocontext> so that extensions get source
compatibility at the expense of performance.  (Passing an arg is
cheaper than grabbing it from thread-local storage.)

You can ignore [pad]THXx when browsing the Perl headers/sources.
Those are strictly for use within the core.  Extensions and embedders
need only be aware of [pad]THX.

=head2 So what happened to dTHR?

=for apidoc_section $concurrency
=for apidoc Amnh||dTHR

C<dTHR> was introduced in perl 5.005 to support the older thread model.
The older thread model now uses the C<THX> mechanism to pass context
pointers around, so C<dTHR> is not useful any more.  Perl 5.6.0 and
later still have it for backward source compatibility, but it is defined
to be a no-op.

=head2 How do I use all this in extensions?

See also L<perlclib/Dealing with embedded perls and threads>.

When Perl is built with MULTIPLICITY, extensions that call
any functions in the Perl API will need to pass the initial context
argument somehow.  The kicker is that you will need to write it in
such a way that the extension still compiles when Perl hasn't been
built with MULTIPLICITY enabled.

There are three ways to do this.  First, the easy but inefficient way,
which is also the default, in order to maintain source compatibility
with extensions: whenever F<XSUB.h> is #included, it redefines the aTHX
and aTHX_ macros to call a function that will return the context.
Thus, something like:

        sv_setiv(sv, num);

in your extension will translate to this when MULTIPLICITY is
in effect:

        Perl_sv_setiv(Perl_get_context(), sv, num);

or to this otherwise:

        Perl_sv_setiv(sv, num);

You don't have to do anything new in your extension to get this; since
the Perl library provides Perl_get_context(), it will all just
work.

The second, more efficient way is to use the following template for
your Foo.xs:

        #define PERL_NO_GET_CONTEXT     /* we want efficiency */
        #include "EXTERN.h"
        #include "perl.h"
        #include "XSUB.h"

        STATIC void my_private_function(int arg1, int arg2);

        STATIC void
        my_private_function(int arg1, int arg2)
        {
            dTHX;       /* fetch context */
            ... call many Perl API functions ...
        }

        [... etc ...]

        MODULE = Foo            PACKAGE = Foo

        /* typical XSUB */

        void
        my_xsub(arg)
                int arg
            CODE:
                my_private_function(arg, 10);

Note that the only two changes from the normal way of writing an
extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
including the Perl headers, followed by a C<dTHX;> declaration at
the start of every function that will call the Perl API.  (You'll
know which functions need this, because the C compiler will complain
that there's an undeclared identifier in those functions.)  No changes
are needed for the XSUBs themselves, because the XS() macro is
correctly defined to pass in the implicit context if needed.

=for apidoc_section $concurrency
=for apidoc AmnhU#||PERL_NO_GET_CONTEXT

The third, even more efficient way is to ape how it is done within
the Perl guts:


        #define PERL_NO_GET_CONTEXT     /* we want efficiency */
        #include "EXTERN.h"
        #include "perl.h"
        #include "XSUB.h"

        /* pTHX_ only needed for functions that call Perl API */
        STATIC void my_private_function(pTHX_ int arg1, int arg2);

        STATIC void
        my_private_function(pTHX_ int arg1, int arg2)
        {
            /* dTHX; not needed here, because THX is an argument */
            ... call Perl API functions ...
        }

        [... etc ...]

        MODULE = Foo            PACKAGE = Foo

        /* typical XSUB */

        void
        my_xsub(arg)
                int arg
            CODE:
                my_private_function(aTHX_ arg, 10);

This implementation never has to fetch the context using a function
call, since it is always passed as an extra argument.  Depending on
your needs for simplicity or efficiency, you may mix the previous
two approaches freely.

Never add a comma after C<pTHX> yourself--always use the form of the
macro with the underscore for functions that take explicit arguments,
or the form without the argument for functions with no explicit arguments.

=head2 Should I do anything special if I call perl from multiple threads?

If you create interpreters in one thread and then proceed to call them in
another, you need to make sure perl's own Thread Local Storage (TLS) slot is
initialized correctly in each of those threads.

The C<perl_alloc> and C<perl_clone> API functions will automatically set
the TLS slot to the interpreter they created, so that there is no need to do
anything special if the interpreter is always accessed in the same thread that
created it, and that thread did not create or call any other interpreters
afterwards.  If that is not the case, you have to set the TLS slot of the
thread before calling any functions in the Perl API on that particular
interpreter.  This is done by calling the C<PERL_SET_CONTEXT> macro in that
thread as the first thing you do:

	/* do this before doing anything else with some_perl */
	PERL_SET_CONTEXT(some_perl);

	... other Perl API calls on some_perl go here ...

=for apidoc_section $embedding
=for apidoc Amh|void|PERL_SET_CONTEXT|PerlInterpreter* i

(You can always get the current context via C<PERL_GET_CONTEXT>.)

=for apidoc Amnh|PerlInterpreter*|PERL_GET_CONTEXT|

=head2 Future Plans and PERL_IMPLICIT_SYS

Just as MULTIPLICITY provides a way to bundle up everything
that the interpreter knows about itself and pass it around, so too are
there plans to allow the interpreter to bundle up everything it knows
about the environment it's running on.  This is enabled with the
PERL_IMPLICIT_SYS macro.  Currently it only works with USE_ITHREADS on
Windows.

This allows the ability to provide an extra pointer (called the "host"
environment) for all the system calls.  This makes it possible for
all the system stuff to maintain their own state, broken down into
seven C structures.  These are thin wrappers around the usual system
calls (see F<win32/perllib.c>) for the default perl executable, but for a
more ambitious host (like the one that would do fork() emulation) all
the extra work needed to pretend that different interpreters are
actually different "processes", would be done here.

The Perl engine/interpreter and the host are orthogonal entities.
There could be one or more interpreters in a process, and one or
more "hosts", with free association between them.

=head1 Internal Functions

All of Perl's internal functions which will be exposed to the outside
world are prefixed by C<Perl_> so that they will not conflict with XS
functions or functions used in a program in which Perl is embedded.
Similarly, all global variables begin with C<PL_>.  (By convention,
static functions start with C<S_>.)

Inside the Perl core (C<PERL_CORE> defined), you can get at the functions
either with or without the C<Perl_> prefix, thanks to a bunch of defines
that live in F<embed.h>.  Note that extension code should I<not> set
C<PERL_CORE>; this exposes the full perl internals, and is likely to cause
breakage of the XS in each new perl release.

The file F<embed.h> is generated automatically from
F<embed.pl> and F<embed.fnc>.  F<embed.pl> also creates the prototyping
header files for the internal functions, generates the documentation
and a lot of other bits and pieces.  It's important that when you add
a new function to the core or change an existing one, you change the
data in the table in F<embed.fnc> as well.  Here's a sample entry from
that table:

    Apd |SV**   |av_fetch   |AV* ar|I32 key|I32 lval

The first column is a set of flags, the second column the return type,
the third column the name.  Columns after that are the arguments.
The flags are documented at the top of F<embed.fnc>.

If you edit F<embed.pl> or F<embed.fnc>, you will need to run
C<make regen_headers> to force a rebuild of F<embed.h> and other
auto-generated files.

=head2 Formatted Printing of IVs, UVs, and NVs

If you are printing IVs, UVs, or NVS instead of the stdio(3) style
formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
following macros for portability

        IVdf            IV in decimal
        UVuf            UV in decimal
        UVof            UV in octal
        UVxf            UV in hexadecimal
        NVef            NV %e-like
        NVff            NV %f-like
        NVgf            NV %g-like

These will take care of 64-bit integers and long doubles.
For example:

        printf("IV is %" IVdf "\n", iv);

The C<IVdf> will expand to whatever is the correct format for the IVs.
Note that the spaces are required around the format in case the code is
compiled with C++, to maintain compliance with its standard.

Note that there are different "long doubles": Perl will use
whatever the compiler has.

If you are printing addresses of pointers, use %p or UVxf combined
with PTR2UV().

=head2 Formatted Printing of SVs

The contents of SVs may be printed using the C<SVf> format, like so:

 Perl_croak(aTHX_ "This croaked because: %" SVf "\n", SVfARG(err_msg))

where C<err_msg> is an SV.

=for apidoc_section $io_formats
=for apidoc Amnh||SVf
=for apidoc Amh||SVfARG|SV *sv

Not all scalar types are printable.  Simple values certainly are: one of
IV, UV, NV, or PV.  Also, if the SV is a reference to some value,
either it will be dereferenced and the value printed, or information
about the type of that value and its address are displayed.  The results
of printing any other type of SV are undefined and likely to lead to an
interpreter crash.  NVs are printed using a C<%g>-ish format.

Note that the spaces are required around the C<SVf> in case the code is
compiled with C++, to maintain compliance with its standard.

Note that any filehandle being printed to under UTF-8 must be expecting
UTF-8 in order to get good results and avoid Wide-character warnings.
One way to do this for typical filehandles is to invoke perl with the
C<-C> parameter.  (See L<perlrun/-C [numberE<sol>list]>.

You can use this to concatenate two scalars:

 SV *var1 = get_sv("var1", GV_ADD);
 SV *var2 = get_sv("var2", GV_ADD);
 SV *var3 = newSVpvf("var1=%" SVf " and var2=%" SVf,
                     SVfARG(var1), SVfARG(var2));

=for apidoc Amnh||SVf_QUOTEDPREFIX

C<SVf_QUOTEDPREFIX> is similar to C<SVf> except that it restricts the
number of the characters printed, showing at most the first
C<PERL_QUOTEDPREFIX_LEN> characters of the argument, and rendering it with
double quotes and with the contents escaped using double quoted string
escaping rules. If the string is longer than this then ellipses "..."
will be appended after the trailing quote. This is intended for error
messages where the string is assumed to be a class name.

=for apidoc Amnh||HvNAMEf
=for apidoc Amnh||HvNAMEf_QUOTEDPREFIX

C<HvNAMEf> and C<HvNAMEf_QUOTEDPREFIX> are similar to C<SVf> except they
extract the string, length and utf8 flags from the argument using the
C<HvNAME()>, C<HvNAMELEN()>, C<HvNAMEUTF8()> macros. This is intended
for stringifying a class name directly from an stash HV.

=head2 Formatted Printing of Strings

If you just want the bytes printed in a 7bit NUL-terminated string, you can
just use C<%s> (assuming they are all really only 7bit).  But if there is a
possibility the value will be encoded as UTF-8 or contains bytes above
C<0x7F> (and therefore 8bit), you should instead use the C<UTF8f> format.
And as its parameter, use the C<UTF8fARG()> macro:

 chr * msg;

 /* U+2018: \xE2\x80\x98 LEFT SINGLE QUOTATION MARK
    U+2019: \xE2\x80\x99 RIGHT SINGLE QUOTATION MARK */
 if (can_utf8)
   msg = "\xE2\x80\x98Uses fancy quotes\xE2\x80\x99";
 else
   msg = "'Uses simple quotes'";

 Perl_croak(aTHX_ "The message is: %" UTF8f "\n",
                  UTF8fARG(can_utf8, strlen(msg), msg));

The first parameter to C<UTF8fARG> is a boolean: 1 if the string is in
UTF-8; 0 if string is in native byte encoding (Latin1).
The second parameter is the number of bytes in the string to print.
And the third and final parameter is a pointer to the first byte in the
string.

Note that any filehandle being printed to under UTF-8 must be expecting
UTF-8 in order to get good results and avoid Wide-character warnings.
One way to do this for typical filehandles is to invoke perl with the
C<-C> parameter.  (See L<perlrun/-C [numberE<sol>list]>.

=for apidoc_section $io_formats
=for apidoc Amnh||UTF8f
Output a possibly UTF8 value. Be sure to use UTF8fARG() to compose
the arguments for this format.
=for apidoc Amnh||UTF8f_QUOTEDPREFIX
Same as C<UTF8f> but the output is quoted, escaped and length limited.
See C<SVf_QUOTEDPREFIX> for more details on escaping.
=for apidoc Amh||UTF8fARG|bool is_utf8|Size_t byte_len|char *str

=cut

=head2 Formatted Printing of C<Size_t> and C<SSize_t>

The most general way to do this is to cast them to a UV or IV, and
print as in the
L<previous section|/Formatted Printing of IVs, UVs, and NVs>.

But if you're using C<PerlIO_printf()>, it's less typing and visual
clutter to use the C<%z> length modifier (for I<siZe>):

        PerlIO_printf("STRLEN is %zu\n", len);

This modifier is not portable, so its use should be restricted to
C<PerlIO_printf()>.

=head2 Formatted Printing of C<Ptrdiff_t>, C<intmax_t>, C<short> and other special sizes

There are modifiers for these special situations if you are using
C<PerlIO_printf()>.  See L<perlfunc/size>.

=head2 Pointer-To-Integer and Integer-To-Pointer

Because pointer size does not necessarily equal integer size,
use the follow macros to do it right.

        PTR2UV(pointer)
        PTR2IV(pointer)
        PTR2NV(pointer)
        INT2PTR(pointertotype, integer)

=for apidoc_section $casting
=for apidoc Amh|type|INT2PTR|type|int value
=for apidoc Amh|UV|PTR2UV|void * ptr
=for apidoc Amh|IV|PTR2IV|void * ptr
=for apidoc Amh|NV|PTR2NV|void * ptr

For example:

        IV  iv = ...;
        SV *sv = INT2PTR(SV*, iv);

and

        AV *av = ...;
        UV  uv = PTR2UV(av);

There are also

 PTR2nat(pointer)   /* pointer to integer of PTRSIZE */
 PTR2ul(pointer)    /* pointer to unsigned long */

=for apidoc Amh|IV|PTR2nat|void *
=for apidoc Amh|unsigned long|PTR2ul|void *

And C<PTRV> which gives the native type for an integer the same size as
pointers, such as C<unsigned> or C<unsigned long>.

=for apidoc Ayh|type|PTRV

=head2 Exception Handling

There are a couple of macros to do very basic exception handling in XS
modules.  You have to define C<NO_XSLOCKS> before including F<XSUB.h> to
be able to use these macros:

        #define NO_XSLOCKS
        #include "XSUB.h"

You can use these macros if you call code that may croak, but you need
to do some cleanup before giving control back to Perl.  For example:

        dXCPT;    /* set up necessary variables */

        XCPT_TRY_START {
          code_that_may_croak();
        } XCPT_TRY_END

        XCPT_CATCH
        {
          /* do cleanup here */
          XCPT_RETHROW;
        }

Note that you always have to rethrow an exception that has been
caught.  Using these macros, it is not possible to just catch the
exception and ignore it.  If you have to ignore the exception, you
have to use the C<call_*> function.

The advantage of using the above macros is that you don't have
to setup an extra function for C<call_*>, and that using these
macros is faster than using C<call_*>.

=head2 Source Documentation

There's an effort going on to document the internal functions and
automatically produce reference manuals from them -- L<perlapi> is one
such manual which details all the functions which are available to XS
writers.  L<perlintern> is the autogenerated manual for the functions
which are not part of the API and are supposedly for internal use only.

Source documentation is created by putting POD comments into the C
source, like this:

 /*
 =for apidoc sv_setiv

 Copies an integer into the given SV.  Does not handle 'set' magic.  See
 L<perlapi/sv_setiv_mg>.

 =cut
 */

Please try and supply some documentation if you add functions to the
Perl core.

=head2 Backwards compatibility

The Perl API changes over time.  New functions are
added or the interfaces of existing functions are
changed.  The C<Devel::PPPort> module tries to
provide compatibility code for some of these changes, so XS writers don't
have to code it themselves when supporting multiple versions of Perl.

C<Devel::PPPort> generates a C header file F<ppport.h> that can also
be run as a Perl script.  To generate F<ppport.h>, run:

    perl -MDevel::PPPort -eDevel::PPPort::WriteFile

Besides checking existing XS code, the script can also be used to retrieve
compatibility information for various API calls using the C<--api-info>
command line switch.  For example:

  % perl ppport.h --api-info=sv_magicext

For details, see S<C<perldoc ppport.h>>.

=head1 Unicode Support

Perl 5.6.0 introduced Unicode support.  It's important for porters and XS
writers to understand this support and make sure that the code they
write does not corrupt Unicode data.

=head2 What B<is> Unicode, anyway?

In the olden, less enlightened times, we all used to use ASCII.  Most of
us did, anyway.  The big problem with ASCII is that it's American.  Well,
no, that's not actually the problem; the problem is that it's not
particularly useful for people who don't use the Roman alphabet.  What
used to happen was that particular languages would stick their own
alphabet in the upper range of the sequence, between 128 and 255.  Of
course, we then ended up with plenty of variants that weren't quite
ASCII, and the whole point of it being a standard was lost.

Worse still, if you've got a language like Chinese or
Japanese that has hundreds or thousands of characters, then you really
can't fit them into a mere 256, so they had to forget about ASCII
altogether, and build their own systems using pairs of numbers to refer
to one character.

To fix this, some people formed Unicode, Inc. and
produced a new character set containing all the characters you can
possibly think of and more.  There are several ways of representing these
characters, and the one Perl uses is called UTF-8.  UTF-8 uses
a variable number of bytes to represent a character.  You can learn more
about Unicode and Perl's Unicode model in L<perlunicode>.

(On EBCDIC platforms, Perl uses instead UTF-EBCDIC, which is a form of
UTF-8 adapted for EBCDIC platforms.  Below, we just talk about UTF-8.
UTF-EBCDIC is like UTF-8, but the details are different.  The macros
hide the differences from you, just remember that the particular numbers
and bit patterns presented below will differ in UTF-EBCDIC.)

=head2 How can I recognise a UTF-8 string?

You can't.  This is because UTF-8 data is stored in bytes just like
non-UTF-8 data.  The Unicode character 200, (C<0xC8> for you hex types)
capital E with a grave accent, is represented by the two bytes
C<v196.172>.  Unfortunately, the non-Unicode string C<chr(196).chr(172)>
has that byte sequence as well.  So you can't tell just by looking -- this
is what makes Unicode input an interesting problem.

In general, you either have to know what you're dealing with, or you
have to guess.  The API function C<is_utf8_string> can help; it'll tell
you if a string contains only valid UTF-8 characters, and the chances
of a non-UTF-8 string looking like valid UTF-8 become very small very
quickly with increasing string length.  On a character-by-character
basis, C<isUTF8_CHAR>
will tell you whether the current character in a string is valid UTF-8.

=head2 How does UTF-8 represent Unicode characters?

As mentioned above, UTF-8 uses a variable number of bytes to store a
character.  Characters with values 0...127 are stored in one
byte, just like good ol' ASCII.  Character 128 is stored as
C<v194.128>; this continues up to character 191, which is
C<v194.191>.  Now we've run out of bits (191 is binary
C<10111111>) so we move on; character 192 is C<v195.128>.  And
so it goes on, moving to three bytes at character 2048.
L<perlunicode/Unicode Encodings> has pictures of how this works.

Assuming you know you're dealing with a UTF-8 string, you can find out
how long the first character in it is with the C<UTF8SKIP> macro:

    char *utf = "\305\233\340\240\201";
    I32 len;

    len = UTF8SKIP(utf); /* len is 2 here */
    utf += len;
    len = UTF8SKIP(utf); /* len is 3 here */

Another way to skip over characters in a UTF-8 string is to use
C<utf8_hop>, which takes a string and a number of characters to skip
over.  You're on your own about bounds checking, though, so don't use it
lightly.

All bytes in a multi-byte UTF-8 character will have the high bit set,
so you can test if you need to do something special with this
character like this (the C<UTF8_IS_INVARIANT()> is a macro that tests
whether the byte is encoded as a single byte even in UTF-8):

    U8 *utf;     /* Initialize this to point to the beginning of the
                    sequence to convert */
    U8 *utf_end; /* Initialize this to 1 beyond the end of the sequence
                    pointed to by 'utf' */
    UV uv;	 /* Returned code point; note: a UV, not a U8, not a
                    char */
    STRLEN len; /* Returned length of character in bytes */

    if (!UTF8_IS_INVARIANT(*utf))
        /* Must treat this as UTF-8 */
        uv = utf8_to_uvchr_buf(utf, utf_end, &len);
    else
        /* OK to treat this character as a byte */
        uv = *utf;

You can also see in that example that we use C<utf8_to_uvchr_buf> to get the
value of the character; the inverse function C<uvchr_to_utf8> is available
for putting a UV into UTF-8:

    if (!UVCHR_IS_INVARIANT(uv))
        /* Must treat this as UTF8 */
        utf8 = uvchr_to_utf8(utf8, uv);
    else
        /* OK to treat this character as a byte */
        *utf8++ = uv;

You B<must> convert characters to UVs using the above functions if
you're ever in a situation where you have to match UTF-8 and non-UTF-8
characters.  You may not skip over UTF-8 characters in this case.  If you
do this, you'll lose the ability to match hi-bit non-UTF-8 characters;
for instance, if your UTF-8 string contains C<v196.172>, and you skip
that character, you can never match a C<chr(200)> in a non-UTF-8 string.
So don't do that!

(Note that we don't have to test for invariant characters in the
examples above.  The functions work on any well-formed UTF-8 input.
It's just that its faster to avoid the function overhead when it's not
needed.)

=head2 How does Perl store UTF-8 strings?

Currently, Perl deals with UTF-8 strings and non-UTF-8 strings
slightly differently.  A flag in the SV, C<SVf_UTF8>, indicates that the
string is internally encoded as UTF-8.  Without it, the byte value is the
codepoint number and vice versa.  This flag is only meaningful if the SV
is C<SvPOK> or immediately after stringification via C<SvPV> or a
similar macro.  You can check and manipulate this flag with the
following macros:

    SvUTF8(sv)
    SvUTF8_on(sv)
    SvUTF8_off(sv)

This flag has an important effect on Perl's treatment of the string: if
UTF-8 data is not properly distinguished, regular expressions,
C<length>, C<substr> and other string handling operations will have
undesirable (wrong) results.

The problem comes when you have, for instance, a string that isn't
flagged as UTF-8, and contains a byte sequence that could be UTF-8 --
especially when combining non-UTF-8 and UTF-8 strings.

Never forget that the C<SVf_UTF8> flag is separate from the PV value; you
need to be sure you don't accidentally knock it off while you're
manipulating SVs.  More specifically, you cannot expect to do this:

    SV *sv;
    SV *nsv;
    STRLEN len;
    char *p;

    p = SvPV(sv, len);
    frobnicate(p);
    nsv = newSVpvn(p, len);

The C<char*> string does not tell you the whole story, and you can't
copy or reconstruct an SV just by copying the string value.  Check if the
old SV has the UTF8 flag set (I<after> the C<SvPV> call), and act
accordingly:

    p = SvPV(sv, len);
    is_utf8 = SvUTF8(sv);
    frobnicate(p, is_utf8);
    nsv = newSVpvn(p, len);
    if (is_utf8)
        SvUTF8_on(nsv);

In the above, your C<frobnicate> function has been changed to be made
aware of whether or not it's dealing with UTF-8 data, so that it can
handle the string appropriately.

Since just passing an SV to an XS function and copying the data of
the SV is not enough to copy the UTF8 flags, even less right is just
passing a S<C<char *>> to an XS function.

For full generality, use the L<C<DO_UTF8>|perlapi/DO_UTF8> macro to see if the
string in an SV is to be I<treated> as UTF-8.  This takes into account
if the call to the XS function is being made from within the scope of
L<S<C<use bytes>>|bytes>.  If so, the underlying bytes that comprise the
UTF-8 string are to be exposed, rather than the character they
represent.  But this pragma should only really be used for debugging and
perhaps low-level testing at the byte level.  Hence most XS code need
not concern itself with this, but various areas of the perl core do need
to support it.

And this isn't the whole story.  Starting in Perl v5.12, strings that
aren't encoded in UTF-8 may also be treated as Unicode under various
conditions (see L<perlunicode/ASCII Rules versus Unicode Rules>).
This is only really a problem for characters whose ordinals are between
128 and 255, and their behavior varies under ASCII versus Unicode rules
in ways that your code cares about (see L<perlunicode/The "Unicode Bug">).
There is no published API for dealing with this, as it is subject to
change, but you can look at the code for C<pp_lc> in F<pp.c> for an
example as to how it's currently done.

=head2 How do I pass a Perl string to a C library?

A Perl string, conceptually, is an opaque sequence of code points.
Many C libraries expect their inputs to be "classical" C strings, which are
arrays of octets 1-255, terminated with a NUL byte. Your job when writing
an interface between Perl and a C library is to define the mapping between
Perl and that library.

Generally speaking, C<SvPVbyte> and related macros suit this task well.
These assume that your Perl string is a "byte string", i.e., is either
raw, undecoded input into Perl or is pre-encoded to, e.g., UTF-8.

Alternatively, if your C library expects UTF-8 text, you can use
C<SvPVutf8> and related macros. This has the same effect as encoding
to UTF-8 then calling the corresponding C<SvPVbyte>-related macro.

Some C libraries may expect other encodings (e.g., UTF-16LE). To give
Perl strings to such libraries
you must either do that encoding in Perl then use C<SvPVbyte>, or
use an intermediary C library to convert from however Perl stores the
string to the desired encoding.

Take care also that NULs in your Perl string don't confuse the C
library. If possible, give the string's length to the C library; if that's
not possible, consider rejecting strings that contain NUL bytes.

=head3 What about C<SvPV>, C<SvPV_nolen>, etc.?

Consider a 3-character Perl string C<$foo = "\x64\x78\x8c">.
Perl can store these 3 characters either of two ways:

=over

=item * bytes: 0x64 0x78 0x8c

=item * UTF-8: 0x64 0x78 0xc2 0x8c

=back

Now let's say you convert C<$foo> to a C string thus:

    STRLEN strlen;
    char *str = SvPV(foo_sv, strlen);

At this point C<str> could point to a 3-byte C string or a 4-byte one.

Generally speaking, we want C<str> to be the same regardless of how
Perl stores C<$foo>, so the ambiguity here is undesirable. C<SvPVbyte>
and C<SvPVutf8> solve that by giving predictable output: use
C<SvPVbyte> if your C library expects byte strings, or C<SvPVutf8>
if it expects UTF-8.

If your C library happens to support both encodings, then C<SvPV>--always
in tandem with lookups to C<SvUTF8>!--may be safe and (slightly) more
efficient.

B<TESTING> B<TIP:> Use L<utf8>'s C<upgrade> and C<downgrade> functions
in your tests to ensure consistent handling regardless of Perl's
internal encoding.

=head2 How do I convert a string to UTF-8?

If you're mixing UTF-8 and non-UTF-8 strings, it is necessary to upgrade
the non-UTF-8 strings to UTF-8.  If you've got an SV, the easiest way to do
this is:

    sv_utf8_upgrade(sv);

However, you must not do this, for example:

    if (!SvUTF8(left))
        sv_utf8_upgrade(left);

If you do this in a binary operator, you will actually change one of the
strings that came into the operator, and, while it shouldn't be noticeable
by the end user, it can cause problems in deficient code.

Instead, C<bytes_to_utf8> will give you a UTF-8-encoded B<copy> of its
string argument.  This is useful for having the data available for
comparisons and so on, without harming the original SV.  There's also
C<utf8_to_bytes> to go the other way, but naturally, this will fail if
the string contains any characters above 255 that can't be represented
in a single byte.

=head2 How do I compare strings?

L<perlapi/sv_cmp> and L<perlapi/sv_cmp_flags> do a lexigraphic
comparison of two SV's, and handle UTF-8ness properly.  Note, however,
that Unicode specifies a much fancier mechanism for collation, available
via the L<Unicode::Collate> module.

To just compare two strings for equality/non-equality, you can just use
L<C<memEQ()>|perlapi/memEQ> and L<C<memNE()>|perlapi/memEQ> as usual,
except the strings must be both UTF-8 or not UTF-8 encoded.

To compare two strings case-insensitively, use
L<C<foldEQ_utf8()>|perlapi/foldEQ_utf8> (the strings don't have to have
the same UTF-8ness).

=head2 Is there anything else I need to know?

Not really.  Just remember these things:

=over 3

=item *

There's no way to tell if a S<C<char *>> or S<C<U8 *>> string is UTF-8
or not.  But you can tell if an SV is to be treated as UTF-8 by calling
C<DO_UTF8> on it, after stringifying it with C<SvPV> or a similar
macro.  And, you can tell if SV is actually UTF-8 (even if it is not to
be treated as such) by looking at its C<SvUTF8> flag (again after
stringifying it).  Don't forget to set the flag if something should be
UTF-8.
Treat the flag as part of the PV, even though it's not -- if you pass on
the PV to somewhere, pass on the flag too.

=item *

If a string is UTF-8, B<always> use C<utf8_to_uvchr_buf> to get at the value,
unless C<UTF8_IS_INVARIANT(*s)> in which case you can use C<*s>.

=item *

When writing a character UV to a UTF-8 string, B<always> use
C<uvchr_to_utf8>, unless C<UVCHR_IS_INVARIANT(uv))> in which case
you can use C<*s = uv>.

=item *

Mixing UTF-8 and non-UTF-8 strings is
tricky.  Use C<bytes_to_utf8> to get
a new string which is UTF-8 encoded, and then combine them.

=back

=head1 Custom Operators

Custom operator support is an experimental feature that allows you to
define your own ops.  This is primarily to allow the building of
interpreters for other languages in the Perl core, but it also allows
optimizations through the creation of "macro-ops" (ops which perform the
functions of multiple ops which are usually executed together, such as
C<gvsv, gvsv, add>.)

This feature is implemented as a new op type, C<OP_CUSTOM>.  The Perl
core does not "know" anything special about this op type, and so it will
not be involved in any optimizations.  This also means that you can
define your custom ops to be any op structure -- unary, binary, list and
so on -- you like.

It's important to know what custom operators won't do for you.  They
won't let you add new syntax to Perl, directly.  They won't even let you
add new keywords, directly.  In fact, they won't change the way Perl
compiles a program at all.  You have to do those changes yourself, after
Perl has compiled the program.  You do this either by manipulating the op
tree using a C<CHECK> block and the C<B::Generate> module, or by adding
a custom peephole optimizer with the C<optimize> module.

When you do this, you replace ordinary Perl ops with custom ops by
creating ops with the type C<OP_CUSTOM> and the C<op_ppaddr> of your own
PP function.  This should be defined in XS code, and should look like
the PP ops in C<pp_*.c>.  You are responsible for ensuring that your op
takes the appropriate number of values from the stack, and you are
responsible for adding stack marks if necessary.

You should also "register" your op with the Perl interpreter so that it
can produce sensible error and warning messages.  Since it is possible to
have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
Perl uses the value of C<< o->op_ppaddr >> to determine which custom op
it is dealing with.  You should create an C<XOP> structure for each
ppaddr you use, set the properties of the custom op with
C<XopENTRY_set>, and register the structure against the ppaddr using
C<Perl_custom_op_register>.  A trivial example might look like:

=for apidoc_section $optree_manipulation
=for apidoc Ayh||XOP

    static XOP my_xop;
    static OP *my_pp(pTHX);

    BOOT:
        XopENTRY_set(&my_xop, xop_name, "myxop");
        XopENTRY_set(&my_xop, xop_desc, "Useless custom op");
        Perl_custom_op_register(aTHX_ my_pp, &my_xop);

The available fields in the structure are:

=over 4

=item xop_name

A short name for your op.  This will be included in some error messages,
and will also be returned as C<< $op->name >> by the L<B|B> module, so
it will appear in the output of module like L<B::Concise|B::Concise>.

=item xop_desc

A short description of the function of the op.

=item xop_class

Which of the various C<*OP> structures this op uses.  This should be one of
the C<OA_*> constants from F<op.h>, namely

=over 4

=item OA_BASEOP

=item OA_UNOP

=item OA_BINOP

=item OA_LOGOP

=item OA_LISTOP

=item OA_PMOP

=item OA_SVOP

=item OA_PADOP

=item OA_PVOP_OR_SVOP

This should be interpreted as 'C<PVOP>' only.  The C<_OR_SVOP> is because
the only core C<PVOP>, C<OP_TRANS>, can sometimes be a C<SVOP> instead.

=item OA_LOOP

=item OA_COP

=for apidoc_section $optree_manipulation
=for apidoc Amnh||OA_BASEOP
=for apidoc_item OA_BINOP
=for apidoc_item OA_COP
=for apidoc_item OA_LISTOP
=for apidoc_item OA_LOGOP
=for apidoc_item OA_LOOP
=for apidoc_item OA_PADOP
=for apidoc_item OA_PMOP
=for apidoc_item OA_PVOP_OR_SVOP
=for apidoc_item OA_SVOP
=for apidoc_item OA_UNOP

=back

The other C<OA_*> constants should not be used.

=item xop_peep

This member is of type C<Perl_cpeep_t>, which expands to C<void
(*Perl_cpeep_t)(aTHX_ OP *o, OP *oldop)>.  If it is set, this function
will be called from C<Perl_rpeep> when ops of this type are encountered
by the peephole optimizer.  I<o> is the OP that needs optimizing;
I<oldop> is the previous OP optimized, whose C<op_next> points to I<o>.

=for apidoc_section $optree_manipulation
=for apidoc Ayh||Perl_cpeep_t

=back

C<B::Generate> directly supports the creation of custom ops by name.

=head1 Stacks

Descriptions above occasionally refer to "the stack", but there are in fact
many stack-like data structures within the perl interpreter. When otherwise
unqualified, "the stack" usually refers to the value stack.

The various stacks have different purposes, and operate in slightly different
ways. Their differences are noted below.

=head2 Value Stack

This stack stores the values that regular perl code is operating on, usually
intermediate values of expressions within a statement. The stack itself is
formed of an array of SV pointers.

The base of this stack is pointed to by the interpreter variable
C<PL_stack_base>, of type C<SV **>.

=for apidoc_section $stack
=for apidoc Amnh||PL_stack_base

The head of the stack is C<PL_stack_sp>, and points to the most
recently-pushed item.

=for apidoc Amnh||PL_stack_sp

Items are pushed to the stack by using the C<PUSHs()> macro or its variants
described above; C<XPUSHs()>, C<mPUSHs()>, C<mXPUSHs()> and the typed
versions. Note carefully that the non-C<X> versions of these macros do not
check the size of the stack and assume it to be big enough. These must be
paired with a suitable check of the stack's size, such as the C<EXTEND> macro
to ensure it is large enough. For example

    EXTEND(SP, 4);
    mPUSHi(10);
    mPUSHi(20);
    mPUSHi(30);
    mPUSHi(40);

This is slightly more performant than making four separate checks in four
separate C<mXPUSHi()> calls.

As a further performance optimisation, the various C<PUSH> macros all operate
using a local variable C<SP>, rather than the interpreter-global variable
C<PL_stack_sp>. This variable is declared by the C<dSP> macro - though it is
normally implied by XSUBs and similar so it is rare you have to consider it
directly. Once declared, the C<PUSH> macros will operate only on this local
variable, so before invoking any other perl core functions you must use the
C<PUTBACK> macro to return the value from the local C<SP> variable back to
the interpreter variable. Similarly, after calling a perl core function which
may have had reason to move the stack or push/pop values to it, you must use
the C<SPAGAIN> macro which refreshes the local C<SP> value back from the
interpreter one.

Items are popped from the stack by using the C<POPs> macro or its typed
versions, There is also a macro C<TOPs> that inspects the topmost item without
removing it.

=for apidoc_section $stack
=for apidoc Amnh||TOPs

Note specifically that SV pointers on the value stack do not contribute to the
overall reference count of the xVs being referred to. If newly-created xVs are
being pushed to the stack you must arrange for them to be destroyed at a
suitable time; usually by using one of the C<mPUSH*> macros or C<sv_2mortal()>
to mortalise the xV.

=head2 Mark Stack

The value stack stores individual perl scalar values as temporaries between
expressions. Some perl expressions operate on entire lists; for that purpose
we need to know where on the stack each list begins. This is the purpose of the
mark stack.

The mark stack stores integers as I32 values, which are the height of the
value stack at the time before the list began; thus the mark itself actually
points to the value stack entry one before the list. The list itself starts at
C<mark + 1>.

The base of this stack is pointed to by the interpreter variable
C<PL_markstack>, of type C<I32 *>.

=for apidoc_section $stack
=for apidoc Amnh||PL_markstack

The head of the stack is C<PL_markstack_ptr>, and points to the most
recently-pushed item.

=for apidoc Amnh||PL_markstack_ptr

Items are pushed to the stack by using the C<PUSHMARK()> macro. Even though
the stack itself stores (value) stack indices as integers, the C<PUSHMARK>
macro should be given a stack pointer directly; it will calculate the index
offset by comparing to the C<PL_stack_sp> variable. Thus almost always the
code to perform this is

    PUSHMARK(SP);

Items are popped from the stack by the C<POPMARK> macro. There is also a macro
C<TOPMARK> that inspects the topmost item without removing it. These macros
return I32 index values directly. There is also the C<dMARK> macro which
declares a new SV double-pointer variable, called C<mark>, which points at the
marked stack slot; this is the usual macro that C code will use when operating
on lists given on the stack.

As noted above, the C<mark> variable itself will point at the most recently
pushed value on the value stack before the list begins, and so the list itself
starts at C<mark + 1>. The values of the list may be iterated by code such as

    for(SV **svp = mark + 1; svp <= PL_stack_sp; svp++) {
      SV *item = *svp;
      ...
    }

Note specifically in the case that the list is already empty, C<mark> will
equal C<PL_stack_sp>.

Because the C<mark> variable is converted to a pointer on the value stack,
extra care must be taken if C<EXTEND> or any of the C<XPUSH> macros are
invoked within the function, because the stack may need to be moved to
extend it and so the existing pointer will now be invalid. If this may be a
problem, a possible solution is to track the mark offset as an integer and
track the mark itself later on after the stack had been moved.

    I32 markoff = POPMARK;

    ...

    SP **mark = PL_stack_base + markoff;

=head2 Temporaries Stack

As noted above, xV references on the main value stack do not contribute to the
reference count of an xV, and so another mechanism is used to track when
temporary values which live on the stack must be released. This is the job of
the temporaries stack.

The temporaries stack stores pointers to xVs whose reference counts will be
decremented soon.

The base of this stack is pointed to by the interpreter variable
C<PL_tmps_stack>, of type C<SV **>.

=for apidoc_section $stack
=for apidoc Amnh||PL_tmps_stack

The head of the stack is indexed by C<PL_tmps_ix>, an integer which stores the
index in the array of the most recently-pushed item.

=for apidoc Amnh||PL_tmps_ix

There is no public API to directly push items to the temporaries stack. Instead,
the API function C<sv_2mortal()> is used to mortalize an xV, adding its
address to the temporaries stack.

Likewise, there is no public API to read values from the temporaries stack.
Instead, the macros C<SAVETMPS> and C<FREETMPS> are used. The C<SAVETMPS>
macro establishes the base levels of the temporaries stack, by capturing the
current value of C<PL_tmps_ix> into C<PL_tmps_floor> and saving the previous
value to the save stack. Thereafter, whenever C<FREETMPS> is invoked all of
the temporaries that have been pushed since that level are reclaimed.

=for apidoc_section $stack
=for apidoc Amnh||PL_tmps_floor

While it is common to see these two macros in pairs within an C<ENTER>/
C<LEAVE> pair, it is not necessary to match them. It is permitted to invoke
C<FREETMPS> multiple times since the most recent C<SAVETMPS>; for example in a
loop iterating over elements of a list. While you can invoke C<SAVETMPS>
multiple times within a scope pair, it is unlikely to be useful. Subsequent
invocations will move the temporaries floor further up, thus effectively
trapping the existing temporaries to only be released at the end of the scope.

=head2 Save Stack

The save stack is used by perl to implement the C<local> keyword and other
similar behaviours; any cleanup operations that need to be performed when
leaving the current scope. Items pushed to this stack generally capture the
current value of some internal variable or state, which will be restored when
the scope is unwound due to leaving, C<return>, C<die>, C<goto> or other
reasons.

Whereas other perl internal stacks store individual items all of the same type
(usually SV pointers or integers), the items pushed to the save stack are
formed of many different types, having multiple fields to them. For example,
the C<SAVEt_INT> type needs to store both the address of the C<int> variable
to restore, and the value to restore it to. This information could have been
stored using fields of a C<struct>, but would have to be large enough to store
three pointers in the largest case, which would waste a lot of space in most
of the smaller cases.

=for apidoc_section $stack
=for apidoc Amnh||SAVEt_INT

Instead, the stack stores information in a variable-length encoding of C<ANY>
structures. The final value pushed is stored in the C<UV> field which encodes
the kind of item held by the preceding items; the count and types of which
will depend on what kind of item is being stored. The kind field is pushed
last because that will be the first field to be popped when unwinding items
from the stack.

The base of this stack is pointed to by the interpreter variable
C<PL_savestack>, of type C<ANY *>.

=for apidoc_section $stack
=for apidoc Amnh||PL_savestack

The head of the stack is indexed by C<PL_savestack_ix>, an integer which
stores the index in the array at which the next item should be pushed. (Note
that this is different to most other stacks, which reference the most
recently-pushed item).

=for apidoc_section $stack
=for apidoc Amnh||PL_savestack_ix

Items are pushed to the save stack by using the various C<SAVE...()> macros.
Many of these macros take a variable and store both its address and current
value on the save stack, ensuring that value gets restored on scope exit.

    SAVEI8(i8)
    SAVEI16(i16)
    SAVEI32(i32)
    SAVEINT(i)
    ...

There are also a variety of other special-purpose macros which save particular
types or values of interest. C<SAVETMPS> has already been mentioned above.
Others include C<SAVEFREEPV> which arranges for a PV (i.e. a string buffer) to
be freed, or C<SAVEDESTRUCTOR> which arranges for a given function pointer to
be invoked on scope exit. A full list of such macros can be found in
F<scope.h>.

There is no public API for popping individual values or items from the save
stack. Instead, via the scope stack, the C<ENTER> and C<LEAVE> pair form a way
to start and stop nested scopes. Leaving a nested scope via C<LEAVE> will
restore all of the saved values that had been pushed since the most recent
C<ENTER>.

=head2 Scope Stack

As with the mark stack to the value stack, the scope stack forms a pair with
the save stack. The scope stack stores the height of the save stack at which
nested scopes begin, and allows the save stack to be unwound back to that
point when the scope is left.

When perl is built with debugging enabled, there is a second part to this
stack storing human-readable string names describing the type of stack
context. Each push operation saves the name as well as the height of the save
stack, and each pop operation checks the topmost name with what is expected,
causing an assertion failure if the name does not match.

The base of this stack is pointed to by the interpreter variable
C<PL_scopestack>, of type C<I32 *>. If enabled, the scope stack names are
stored in a separate array pointed to by C<PL_scopestack_name>, of type
C<const char **>.

=for apidoc_section $stack
=for apidoc Amnh||PL_scopestack
=for apidoc Amnh||PL_scopestack_name

The head of the stack is indexed by C<PL_scopestack_ix>, an integer which
stores the index of the array or arrays at which the next item should be
pushed. (Note that this is different to most other stacks, which reference the
most recently-pushed item).

=for apidoc_section $stack
=for apidoc Amnh||PL_scopestack_ix

Values are pushed to the scope stack using the C<ENTER> macro, which begins a
new nested scope. Any items pushed to the save stack are then restored at the
next nested invocation of the C<LEAVE> macro.

=head1 Dynamic Scope and the Context Stack

B<Note:> this section describes a non-public internal API that is subject
to change without notice.

=head2 Introduction to the context stack

In Perl, dynamic scoping refers to the runtime nesting of things like
subroutine calls, evals etc, as well as the entering and exiting of block
scopes. For example, the restoring of a C<local>ised variable is
determined by the dynamic scope.

Perl tracks the dynamic scope by a data structure called the context
stack, which is an array of C<PERL_CONTEXT> structures, and which is
itself a big union for all the types of context. Whenever a new scope is
entered (such as a block, a C<for> loop, or a subroutine call), a new
context entry is pushed onto the stack. Similarly when leaving a block or
returning from a subroutine call etc. a context is popped. Since the
context stack represents the current dynamic scope, it can be searched.
For example, C<next LABEL> searches back through the stack looking for a
loop context that matches the label; C<return> pops contexts until it
finds a sub or eval context or similar; C<caller> examines sub contexts on
the stack.

=for apidoc_section $concurrency
=for apidoc Cyh||PERL_CONTEXT

Each context entry is labelled with a context type, C<cx_type>. Typical
context types are C<CXt_SUB>, C<CXt_EVAL> etc., as well as C<CXt_BLOCK>
and C<CXt_NULL> which represent a basic scope (as pushed by C<pp_enter>)
and a sort block. The type determines which part of the context union are
valid.

=for apidoc  Cyh ||cx_type

=for apidoc  Cmnh||CXt_BLOCK
=for apidoc_item ||CXt_EVAL
=for apidoc_item ||CXt_FORMAT
=for apidoc_item ||CXt_GIVEN
=for apidoc_item ||CXt_LOOP_ARY
=for apidoc_item ||CXt_LOOP_LAZYIV
=for apidoc_item ||CXt_LOOP_LAZYSV
=for apidoc_item ||CXt_LOOP_LIST
=for apidoc_item ||CXt_LOOP_PLAIN
=for apidoc_item ||CXt_NULL
=for apidoc_item ||CXt_SUB
=for apidoc_item ||CXt_SUBST
=for apidoc_item ||CXt_WHEN

The main division in the context struct is between a substitution scope
(C<CXt_SUBST>) and block scopes, which are everything else. The former is
just used while executing C<s///e>, and won't be discussed further
here.

All the block scope types share a common base, which corresponds to
C<CXt_BLOCK>. This stores the old values of various scope-related
variables like C<PL_curpm>, as well as information about the current
scope, such as C<gimme>. On scope exit, the old variables are restored.

Particular block scope types store extra per-type information. For
example, C<CXt_SUB> stores the currently executing CV, while the various
for loop types might hold the original loop variable SV. On scope exit,
the per-type data is processed; for example the CV has its reference count
decremented, and the original loop variable is restored.

The macro C<cxstack> returns the base of the current context stack, while
C<cxstack_ix> is the index of the current frame within that stack.

=for apidoc_section $concurrency
=for apidoc Cmnh|PERL_CONTEXT *|cxstack
=for apidoc Cmnh|I32|cxstack_ix

In fact, the context stack is actually part of a stack-of-stacks system;
whenever something unusual is done such as calling a C<DESTROY> or tie
handler, a new stack is pushed, then popped at the end.

Note that the API described here changed considerably in perl 5.24; prior
to that, big macros like C<PUSHBLOCK> and C<POPSUB> were used; in 5.24
they were replaced by the inline static functions described below. In
addition, the ordering and detail of how these macros/function work
changed in many ways, often subtly. In particular they didn't handle
saving the savestack and temps stack positions, and required additional
C<ENTER>, C<SAVETMPS> and C<LEAVE> compared to the new functions. The
old-style macros will not be described further.


=head2 Pushing contexts

For pushing a new context, the two basic functions are
C<cx = cx_pushblock()>, which pushes a new basic context block and returns
its address, and a family of similar functions with names like
C<cx_pushsub(cx)> which populate the additional type-dependent fields in
the C<cx> struct. Note that C<CXt_NULL> and C<CXt_BLOCK> don't have their
own push functions, as they don't store any data beyond that pushed by
C<cx_pushblock>.

The fields of the context struct and the arguments to the C<cx_*>
functions are subject to change between perl releases, representing
whatever is convenient or efficient for that release.

A typical context stack pushing can be found in C<pp_entersub>; the
following shows a simplified and stripped-down example of a non-XS call,
along with comments showing roughly what each function does.

 dMARK;
 U8 gimme      = GIMME_V;
 bool hasargs  = cBOOL(PL_op->op_flags & OPf_STACKED);
 OP *retop     = PL_op->op_next;
 I32 old_ss_ix = PL_savestack_ix;
 CV *cv        = ....;

 /* ... make mortal copies of stack args which are PADTMPs here ... */

 /* ... do any additional savestack pushes here ... */

 /* Now push a new context entry of type 'CXt_SUB'; initially just
  * doing the actions common to all block types: */

 cx = cx_pushblock(CXt_SUB, gimme, MARK, old_ss_ix);

     /* this does (approximately):
         CXINC;              /* cxstack_ix++ (grow if necessary) */
         cx = CX_CUR();      /* and get the address of new frame */
         cx->cx_type        = CXt_SUB;
         cx->blk_gimme      = gimme;
         cx->blk_oldsp      = MARK - PL_stack_base;
         cx->blk_oldsaveix  = old_ss_ix;
         cx->blk_oldcop     = PL_curcop;
         cx->blk_oldmarksp  = PL_markstack_ptr - PL_markstack;
         cx->blk_oldscopesp = PL_scopestack_ix;
         cx->blk_oldpm      = PL_curpm;
         cx->blk_old_tmpsfloor = PL_tmps_floor;

         PL_tmps_floor        = PL_tmps_ix;
     */


 /* then update the new context frame with subroutine-specific info,
  * such as the CV about to be executed: */

 cx_pushsub(cx, cv, retop, hasargs);

     /* this does (approximately):
         cx->blk_sub.cv          = cv;
         cx->blk_sub.olddepth    = CvDEPTH(cv);
         cx->blk_sub.prevcomppad = PL_comppad;
         cx->cx_type            |= (hasargs) ? CXp_HASARGS : 0;
         cx->blk_sub.retop       = retop;
         SvREFCNT_inc_simple_void_NN(cv);
     */

=for apidoc_section $concurrency
=for apidoc Cmnh||CXINC

Note that C<cx_pushblock()> sets two new floors: for the args stack (to
C<MARK>) and the temps stack (to C<PL_tmps_ix>). While executing at this
scope level, every C<nextstate> (amongst others) will reset the args and
tmps stack levels to these floors. Note that since C<cx_pushblock> uses
the current value of C<PL_tmps_ix> rather than it being passed as an arg,
this dictates at what point C<cx_pushblock> should be called. In
particular, any new mortals which should be freed only on scope exit
(rather than at the next C<nextstate>) should be created first.

Most callers of C<cx_pushblock> simply set the new args stack floor to the
top of the previous stack frame, but for C<CXt_LOOP_LIST> it stores the
items being iterated over on the stack, and so sets C<blk_oldsp> to the
top of these items instead. Note that, contrary to its name, C<blk_oldsp>
doesn't always represent the value to restore C<PL_stack_sp> to on scope
exit.

Note the early capture of C<PL_savestack_ix> to C<old_ss_ix>, which is
later passed as an arg to C<cx_pushblock>. In the case of C<pp_entersub>,
this is because, although most values needing saving are stored in fields
of the context struct, an extra value needs saving only when the debugger
is running, and it doesn't make sense to bloat the struct for this rare
case. So instead it is saved on the savestack. Since this value gets
calculated and saved before the context is pushed, it is necessary to pass
the old value of C<PL_savestack_ix> to C<cx_pushblock>, to ensure that the
saved value gets freed during scope exit.  For most users of
C<cx_pushblock>, where nothing needs pushing on the save stack,
C<PL_savestack_ix> is just passed directly as an arg to C<cx_pushblock>.

Note that where possible, values should be saved in the context struct
rather than on the save stack; it's much faster that way.

Normally C<cx_pushblock> should be immediately followed by the appropriate
C<cx_pushfoo>, with nothing between them; this is because if code
in-between could die (e.g. a warning upgraded to fatal), then the context
stack unwinding code in C<dounwind> would see (in the example above) a
C<CXt_SUB> context frame, but without all the subroutine-specific fields
set, and crashes would soon ensue.

=for apidoc dounwind

Where the two must be separate, initially set the type to C<CXt_NULL> or
C<CXt_BLOCK>, and later change it to C<CXt_foo> when doing the
C<cx_pushfoo>. This is exactly what C<pp_enteriter> does, once it's
determined which type of loop it's pushing.

=head2 Popping contexts

Contexts are popped using C<cx_popsub()> etc. and C<cx_popblock()>. Note
however, that unlike C<cx_pushblock>, neither of these functions actually
decrement the current context stack index; this is done separately using
C<CX_POP()>.

=for apidoc_section $concurrency
=for apidoc Cmh|void|CX_POP|PERL_CONTEXT* cx

There are two main ways that contexts are popped. During normal execution
as scopes are exited, functions like C<pp_leave>, C<pp_leaveloop> and
C<pp_leavesub> process and pop just one context using C<cx_popfoo> and
C<cx_popblock>. On the other hand, things like C<pp_return> and C<next>
may have to pop back several scopes until a sub or loop context is found,
and exceptions (such as C<die>) need to pop back contexts until an eval
context is found. Both of these are accomplished by C<dounwind()>, which
is capable of processing and popping all contexts above the target one.

Here is a typical example of context popping, as found in C<pp_leavesub>
(simplified slightly):

 U8 gimme;
 PERL_CONTEXT *cx;
 SV **oldsp;
 OP *retop;

 cx = CX_CUR();

 gimme = cx->blk_gimme;
 oldsp = PL_stack_base + cx->blk_oldsp; /* last arg of previous frame */

 if (gimme == G_VOID)
     PL_stack_sp = oldsp;
 else
     leave_adjust_stacks(oldsp, oldsp, gimme, 0);

 CX_LEAVE_SCOPE(cx);
 cx_popsub(cx);
 cx_popblock(cx);
 retop = cx->blk_sub.retop;
 CX_POP(cx);

 return retop;

=for apidoc_section $concurrency
=for apidoc Cmh||CX_CUR

The steps above are in a very specific order, designed to be the reverse
order of when the context was pushed. The first thing to do is to copy
and/or protect any return arguments and free any temps in the current
scope. Scope exits like an rvalue sub normally return a mortal copy of
their return args (as opposed to lvalue subs). It is important to make
this copy before the save stack is popped or variables are restored, or
bad things like the following can happen:

    sub f { my $x =...; $x }  # $x freed before we get to copy it
    sub f { /(...)/;    $1 }  # PL_curpm restored before $1 copied

Although we wish to free any temps at the same time, we have to be careful
not to free any temps which are keeping return args alive; nor to free the
temps we have just created while mortal copying return args. Fortunately,
C<leave_adjust_stacks()> is capable of making mortal copies of return args,
shifting args down the stack, and only processing those entries on the
temps stack that are safe to do so.

In void context no args are returned, so it's more efficient to skip
calling C<leave_adjust_stacks()>. Also in void context, a C<nextstate> op
is likely to be imminently called which will do a C<FREETMPS>, so there's
no need to do that either.

The next step is to pop savestack entries: C<CX_LEAVE_SCOPE(cx)> is just
defined as C<< LEAVE_SCOPE(cx->blk_oldsaveix) >>. Note that during the
popping, it's possible for perl to call destructors, call C<STORE> to undo
localisations of tied vars, and so on. Any of these can die or call
C<exit()>. In this case, C<dounwind()> will be called, and the current
context stack frame will be re-processed. Thus it is vital that all steps
in popping a context are done in such a way to support reentrancy.  The
other alternative, of decrementing C<cxstack_ix> I<before> processing the
frame, would lead to leaks and the like if something died halfway through,
or overwriting of the current frame.

=for apidoc_section $concurrency
=for apidoc Cmh|void|CX_LEAVE_SCOPE|PERL_CONTEXT* cx

C<CX_LEAVE_SCOPE> itself is safely re-entrant: if only half the savestack
items have been popped before dying and getting trapped by eval, then the
C<CX_LEAVE_SCOPE>s in C<dounwind> or C<pp_leaveeval> will continue where
the first one left off.

The next step is the type-specific context processing; in this case
C<cx_popsub>. In part, this looks like:

    cv = cx->blk_sub.cv;
    CvDEPTH(cv) = cx->blk_sub.olddepth;
    cx->blk_sub.cv = NULL;
    SvREFCNT_dec(cv);

where its processing the just-executed CV. Note that before it decrements
the CV's reference count, it nulls the C<blk_sub.cv>. This means that if
it re-enters, the CV won't be freed twice. It also means that you can't
rely on such type-specific fields having useful values after the return
from C<cx_popfoo>.

Next, C<cx_popblock> restores all the various interpreter vars to their
previous values or previous high water marks; it expands to:

    PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
    PL_scopestack_ix = cx->blk_oldscopesp;
    PL_curpm         = cx->blk_oldpm;
    PL_curcop        = cx->blk_oldcop;
    PL_tmps_floor    = cx->blk_old_tmpsfloor;

Note that it I<doesn't> restore C<PL_stack_sp>; as mentioned earlier,
which value to restore it to depends on the context type (specifically
C<for (list) {}>), and what args (if any) it returns; and that will
already have been sorted out earlier by C<leave_adjust_stacks()>.

Finally, the context stack pointer is actually decremented by C<CX_POP(cx)>.
After this point, it's possible that that the current context frame could
be overwritten by other contexts being pushed. Although things like ties
and C<DESTROY> are supposed to work within a new context stack, it's best
not to assume this. Indeed on debugging builds, C<CX_POP(cx)> deliberately
sets C<cx> to null to detect code that is still relying on the field
values in that context frame. Note in the C<pp_leavesub()> example above,
we grab C<blk_sub.retop> I<before> calling C<CX_POP>.

=head2 Redoing contexts

Finally, there is C<cx_topblock(cx)>, which acts like a super-C<nextstate>
as regards to resetting various vars to their base values. It is used in
places like C<pp_next>, C<pp_redo> and C<pp_goto> where rather than
exiting a scope, we want to re-initialise the scope. As well as resetting
C<PL_stack_sp> like C<nextstate>, it also resets C<PL_markstack_ptr>,
C<PL_scopestack_ix> and C<PL_curpm>. Note that it doesn't do a
C<FREETMPS>.


=head1 Reference-counted argument stack

=head2 Introduction

As of perl 5.40, there is a build option, C<PERL_RC_STACK>, not enabled by
default, which requires that items pushed onto, or popped off the argument
stack have their reference counts adjusted. It is intended that eventually
this will be the default way (and finally the only way) to configure perl.

The macros which manipulate the stack such as PUSHs() and POPs() don't
adjust the reference count of the SV. Most of the time this is fine, since
something else is keeping the SV alive while on the argument stack,  such
a pointer from the TEMPs stack, or from the pad (e.g. a lexical variable
or a C<PADTMP>). Occasionally this can go horribly wrong. For example,
this code:

    my @a = (1,2,3);
    sub f { @a = (); print "(@_)\n" };
    f(@a, 4);

may print undefined or random freed values, since some of the elements of
@_, which have been aliased to the elements of @a, have been freed.
C<PERL_RC_STACK> is intended to fix this by making each SV pointer on the
argument stack increment the reference count (RC) of the SV  by one.

In this new environment, unmodified existing PP and XS functions, which
have been written assuming a non reference-counted stack (non-RC for
short), are called via special wrapper functions which adjust the stack
before and after. At the moment there is no API to write an RC XS
function, so all XS code will continue to be called via a wrapper (which
makes them slightly slower), but means that in general, CPAN distributions
containing XS code code continue to work without modification.

However, PP functions, either in perl core, or those in XS functions used
to implement custom ops or to override the PP functions for built-in ops,
need dealing with specially. For the latter, they can just be wrapped;
this involves the least work, but has a performance impact. In the longer
term, and for core PP functions, they need unwrapping and rewriting using
a new API. With this, the old macros such as PUSHs() have been replaced
with a new set of (mostly inline) functions with a common prefix, such as
rpp_push_1(). "RPP" stands for "reference-counted push and pop functions".
The new functions modify the reference count on C<PERL_RC_STACK> builds,
while leaving them unadjusted otherwise. Thus in core they generally work
in both cases, while in XS code they are portable to older perl versions
via C<PPPort> (XXX assuming that they get been added to C<PPPort>).

The rest of this section is mainly concerned with how to convert existing
PP functions, and how to write new PP functions to use the new C<rpp_>
API.

A reference-counted perl can be built using the PERL_RC_STACK define.
For development and debugging purposes, it is best to enable leaking
scalar debugging too, as that displays extra information about scalars
that have leaked or been prematurely freed.

    Configure -DDEBUGGING \
      -Accflags='-DPERL_RC_STACK -DDEBUG_LEAKING_SCALARS'

=head2 Reference counted stack states

In the new regime, the current argument stack can be in one of three
states, which can be determined by the shown expression.

=over

=item * not reference-counted

    !AvREAL(PL_curstack)

In this case, perl will assume when emptying the stack (such as during a
croak()) that the items on it don't need freeing. This is the traditional
perl behaviour. On C<PERL_RC_STACK> builds, such stacks will be rarely
encountered.

=item * fully reference-counted

    AvREAL(PL_curstack) && !PL_curstackinfo->si_stack_nonrc_base

All the items on the stack are reference counted, and will be freed by
functions like rpp_popfree_1() or if perl croak()s. This is the normal
state of the stack in C<PERL_RC_STACK> builds.

=item * partially reference-counted (split)

    AvREAL(PL_curstack) && PL_curstackinfo->si_stack_nonrc_base > 0

In this case, items on the stack from the index C<si_stack_nonrc_base>
upwards are non-RC; those below are RC. This state occurs when a PP or XS
function has been wrapped. In this case, the wrapper function pushes a
non-RC copy of the arg pointers above the cut then calls the real
function. When that returns, the wrapper function bumps up the RC of any
returned args. See below for more details.

=back

Note that perl uses a stack-of-stacks, and the AvREAL() and
C<si_stack_nonrc_base> states are per stack. When perl starts up, the main
stack is RC, but by default, new stacks pushed in XS code via PUSHSTACKi()
are non-RC, so it is quite possible to get a mixture. The perl core itself
uses the new push_stackinfo() function which replaces PUSHSTACKi() and
allows you to specify that the new stack should be RC by default.
(XXX core mostly hasn't actually been updated yet to use push_stackinfo())

Most places in the core assume a particular RC environment. In particular,
it is assumed that within a runops loop, all the PP functions are
RC-aware, either because they have been (re)written to be aware, or
because they have been wrapped. Whenever a runops loop is entered via
CALLRUNOPS(), it will check the current state of the stack, and if it's
not fully RC, will temporarily update its contents to be fully RC before
entering the main runops loop. Then if necessary it will restore the stack
to its old state on return. This means that functions like call_sv(),
which can be called from any environment (e.g. RC core or wrapped and
temporarily non-RC XS code) will always do the Right Thing when invoking
the runops loop, no matter what the current stack state is.

Similarly, croaks and the like (which can occur anywhere) have to be able
to handle both stack types. So there are a few places in core - call_sv(),
eval_sv() etc, Perl_die_unwind() and S_my_exit_jump() - which have been
specially crafted to handle both cases; everything else can assume a fixed
environment.

=head2 Wrapping

Normally a core PP function is declared like this:

    PP(pp_foo)
    {
        ...
    }

This expands to something like:

    OP* Perl_pp_foo(pTHX)
    {
        ...
    }

When such a function needs to be wrapped, it is instead declared as:

    PP_wrapped(pp_foo, nargs, nlists)
    {
        ...
    }

which on non-RC builds, expands to the same as PP() (the extra args are
ignored). On RC builds it expands to something like

    OP* Perl_pp_foo(pTHX)
    {
        return Perl_pp_wrap(aTHX_ S_Perl_pp_foo_norc, nargs, nlists);
    }

    STATIC OP* S_Perl_pp_foo_norc(pTHX)
    {
        ...
    }

Here the externally visible PP function calls pp_wrap(), which adjusts
the stack contents, then calls the hidden real body of the PP function,
then on return, adjusts the stack back.

There is an API macro, XSPP_wrapped(), intended for use on PP functions
declared in XS code, It is identical to PP_wrapped(), except that it
doesn't prepend a C<Perl_> prefix to the function name.

The C<nargs> and C<nlists> parameters to the macro are numeric constants
or simple expressions which specify how many arguments the PP function
expects, or how many lists it expects. For example,

    PP_wrapped(pp_add, 2, 0);     /* consumes two args off the stack */

    PP_wrapped(pp_readline,       /* consumes one or two args */
            ((PL_op->op_flags & OPf_STACKED) ? 2 : 1), 0);

    PP_wrapped(pp_push, 0, 1);    /* consumes one list */

    PP_wrapped(pp_aassign, 0, 2); /* consumes two lists */

To understand what pp_wrap() does, consider calling Perl_pp_foo() which
expects three arguments. On entry the stack may look like:

    ... A+ B+ C+

(where the C<+> indicates that the pointers to A, B and C are each
reference counted). The wrapper function pp_wrap() marks a cut at the
current stack position using C<si_stack_nonrc_base>, then, based on the
value of C<nargs>, pushes a copy of those three pointers above the cut:

    ... A+ B+ C+ | A0 B0 C0

(where the C<0> indicates that the pointers aren't RC), then calls the
real PP function, S_Perl_pp_foo_norc(). That function processes A, B and C,
pops them off the stack, and pushes some result SVs. None of this
manipulation adjusts any RCs. On return to pp_wrap(), the stack may look
something like:

    ... A+ B+ C+ | X0 Y0

The wrapper function bumps up the RCs of X and Y, decrements A B C,
shifts the results down and sets C<si_stack_nonrc_base> to zero, leaving
the stack as:

    ... X+ Y+

In places like pp_entersub(), a similar wrapping (via the functions
rpp_invoke_xs() and then xs_wrap()) is done when calling XS subs.

When C<nlists> is positive, a similar action takes place, except that the
mark stack is examined (and adjusted) in order to determine the number of
args that need copying.

A complex calling environment might have multiple nested stacks with
different RC states. Perl starts off with an RC stack. Then for example,
pp_entersub() is called, which (via xs_wrap()) splits the stack and
executes the XS function in a non-RC environment. That function may call
PUSHSTACKi(), which creates a new non-RC stack, then calls call_sv(), which
does CALLRUNOPS(), which causes the new stack to temporarily become RC.
Then a tied method is called, which pushes a new RC stack, and so on. (XXX
currently tied methods actually push a non-RC stack. To be fixed soon).

=head2 (Re)writing a PP function using the rpp_() API

Wrapping a PP function has a performance overhead, and is there mainly as
a temporary crutch. Eventually, PP functions should be updated to use
rpp_() functions, and any new PP functions should be written this way from
scratch and thus not ever need wrapping.

A couple examples of core PP functions being converted can be seen in the
commits C<v5.39.1-304-g205fcd8410> and C<v5.39.1-303-g2fe263a83a>, which
demonstrate a unary and a binary op being converted (pp_not() and
pp_and()).

The traditional PP stack API consisted of a C<dSP> declaration, plus a
number of macros to push, pop and extend the stack. A I<very simplified>
pp_add() function might look something like:

    PP(pp_add)
    {
        dSP;
        dTARGET;
        IV right = SvIV(POPs);
        IV left  = SvIV(POPs);
        TARGi(left + right, 1);
        PUSHs(TARG);
        PUTBACK;
        return NORMAL;
    }

which expands to something like:

    {
        SV **sp = PL_stack_sp;
        SV *targ = PAD_SV(PL_op->op_targ);
        IV right = SvIV(*sp--);
        IV left  = SvIV(*sp--);
        sv_setiv(targ, left + right);
        *++sp = targ;
        PL_stack_sp = sp;
        return PL_op->op_next;
    }

The whole C<dSP> thing harks back to the days before decent optimising
compilers. It was always error-prone, e.g. if you forgot a C<PUTBACK> or
C<SPAGAIN>. The new API always just accesses C<PL_stack_sp> directly. In
fact the first step of upgrading a PP function is always to remove the
C<dSP> declaration. This has the happy side effect that any old-style
macros left in the pp function which implicitly use C<sp> will become
compile errors. The existence of a C<dSP> somewhere in core is a good sign
that that function still needs updating.

An obvious question is: why not just modify the definitions of the PUSHs()
etc macros to modify reference counts on RC builds? The basic problem is
that an SV may now be kept alive only by a single reference count from
the stack (formerly, they tended to be on the TEMPs stack too). So in code
like:

    SV *sv = POPs;
    IV i = SvIV(sv);

including an SvREFCNT_dec() in the C<POPs> macro definition would cause
C<sv> to be freed immediately, before its integer value can be read.

A potential issue with the new regime is that perl can croak at basically
any point in execution (e.g. the SvIV() above might call FETCH() on a tied
variable which then croaks). Thus at all times, the RC of each SV must be
properly accounted for. In the example above, a naive approach to avoiding
a premature free of C<sv> might be:

    SV *sv = *PL_stack_sp--;
    IV i = SvIV(sv);
    SvREFCNT_dec(sv); // got i, so ok to free sv now

but that means that C<sv> leaks if SvIV() triggers a croak.

To avoid that, the new regime has the general outline that arguments are
left on the stack I<until they are finished with>, then removed and their
reference count adjusted at that point. With the new API, the pp_add()
function looks something like:

    {
        dTARGET;
        IV right = SvIV(PL_stack_sp[ 0]); // NB: arguments left on stack
        IV left  = SvIV(PL_stack_sp[-1]);
        TARGi(left + right, 1);
        rpp_replace_2_1(targ);
        return NORMAL;
    }

The rpp_replace_2_1() function pops two values off the stack and pushes
one new value on, while adjusting reference counts as appropriate
(depending on whether built with C<PERL_RC_STACK> or not).

The rpp_() functions in the new API will be described in detail below, but
in summary:

 new function            approximate old equivant
 ------------            -----------------------

 rpp_extend(n)           EXTEND(SP, n)

 rpp_push_1(sv)          PUSHs(sv)
 rpp_push_2(sv1, sv2))   PUSHs(sv1); PUSHs(sv2)
 rpp_xpush_1(sv)         XPUSHs(sv)
 rpp_xpush_2(sv1, sv2))  EXTEND(SP,2); PUSHs(sv1); PUSHs(sv2);

 rpp_push_1_norc(sv)     mPUSHs(sv)     // on RC bulds, skips RC++;
                                        // on non-RC builds, mortalises
 rpp_popfree_1()         (void)POPs;
 rpp_popfree_2()         (void)POPs; (void)POPs;
 rpp_popfree_to(svp)     PL_stack_sp = svp;
 rpp_obliterate_stack_to(ix)           // see description below

 sv = rpp_pop_1_norc()   sv = SvREFCNT_inc(POPs)

 rpp_replace_1_1(sv)     (void)POPs; PUSHs(sv);
 rpp_replace_2_1(sv)     (void)POPs; (void)POPs; PUSHs(sv);
 rpp_replace_at(sp, sv)  *sp = sv;
 rpp_replace_at_norc(sp, sv) *sp = sv_2mortal(sv);

 rpp_context(mark, gimme,
             extra)      SP -= extra;
                         // impose void/scalar/list context on return args
                         SP = (gimme == G_VOID) ? mark : ....

 rpp_try_AMAGIC_1()      tryAMAGICun_MG()
 rpp_try_AMAGIC_2()      tryAMAGICbin_MG()

 rpp_is_lone(sv)         SvTEMP(sv) && SvREFCNT(sv) == 1
 rpp_stack_is_rc()       no equivalent

 rpp_invoke_xs(cv)       CvXSUB(cv)(aTHX_ cv);


 (no replacement)        dATARGET   // just write the macro body in full

There are also some C<_NN> variants which assume that any items being
removed from the stack are non-NULL, and so are slightly more efficient:

 rpp_popfree_1_NN()
 rpp_popfree_2_NN()
 rpp_popfree_to_NN(svp)

 rpp_replace_1_1_NN(sv)
 rpp_replace_2_1_NN(sv)
 rpp_replace_at_NN(sp, sv)
 rpp_replace_at_norc_NN(sp, sv)

There are also a few C<_IMM> variants, which expect the single pushed or
replacement value to be an immortal, such as C<&PL_sv_undef> - this skips
incrementing the ref count of the immortal SV. It doesn't matter if the
ref count of the SV prematurely reaches zero, as sv_free2() will just
resurrect it. Not every variant is provided; if a suitable one
doesn't exist, just using a standard C<_1> version is fine, albeit
slightly slower.

 rpp_push_IMM(&PL_sv_undef)
 rpp_xpush_IMM(&PL_sv_zero)
 rpp_replace_1_IMM_NN(&PL_sv_yes)
 rpp_replace_2_IMM_NN(&PL_sv_no)

Other new C and perl functions related to reference-counted stacks are:

 push_stackinfo(type,rc) PUSHSTACKi(type)
 pop_stackinfo()         POPSTACK()
 switch_argstack(to)     SWITCHSTACK(from,to)

 (Internals::stack_refcounted() & 1) # perl built with PERL_RC_STACK

Some of these new functions are trivial, but should be used in preference
to writing direct code because they will work on both RC and non-RC
builds, and may do extra checks and assertions on C<DEBUGGING> builds.

Note that rpp_popfree_1() etc aren't direct replacements for C<POPs>. The
rpp_() variants don't return a value and are intended to be called when
the SV is finished with. So

    SV *sv = POPs;
    ... do stuff with sv ...

becomes

    SV *sv = *PL_stack_sp;
    ... do stuff with sv ...
    rpp_popfree_1(); /* does SvREFCNT_dec(*PL_stack_sp--) */

The rpp_replace_M_N() functions are shortcuts for popping and freeing C<M>
items then pushing and bumping up the RCs of C<N> items. Note that they
handle edge cases such as an old and new SV being the same.

rpp_replace_at(sp, sv) is similar to rpp_replace_1_1(), except that
it replaces an SV at an address in the stack rather than at the top.

rpp_replace_at_norc(sp, sv) is similar to rpp_replace_at(), except that
it assumes that C<sv> already has a bumped reference count. So, a bit
like rpp_push_1_norc() (see below), it doesn't bother increasing C<sv>'s
reference count, or on non-RC builds it mortalises it instead.

rpp_popfree_to(svp) is designed to replace code like

    PL_stack_sp = PL_stack_base + cx->blk_oldsp;

which typically appears in list ops or scope exits when the arguments are
finished with. Left unaltered, all the SVs above C<oldsp> would leak. The
new approach is

    rpp_popfree_to(PL_stack_base + cx->blk_oldsp);

There is a rarely-used variant of this, rpp_obliterate_stack_to(), which
pops the stack back to the specified index regardless of the current RC
state of the stack. So for example if the stack is split, it will only
adjust the RCs of any SVs which are below the split point, while
rpp_popfree_to() would mindlessly free I<all> SVs (on RC builds anyway).
For normal PP functions you should only ever use rpp_popfree_to(), which
is faster.

There are no new equivalents for all the convenience macros like POPi()
and (shudder) dPOPPOPiirl(). These should be replaced with the rpp_()
functions above and with the conversions and variable declarations being
made explicit, e.g. dPOPPOPiirl() becomes:

    IV right = SvIV(PL_stack_sp[ 0]);
    IV left  = SvIV(PL_stack_sp[-1]);
    rpp_popfree_2();

A couple of the rpp_() functions with C<norc> in their names don't adjust
the reference count on RC builds (but, conversely, do on non-RC builds).

rpp_push_1_norc(sv) does a simple C<*++PL_stack_sp = sv> on RC builds. It
is typically used to "root" a newly-created SV, which already has an RC of
1. On non-RC builds it mortalises the SV instead. So for example, code
which used to look like

    mPUSHs(newSViv(i));

and which expanded to the equivalent of:

    PUSHs(sv_2mortal(newSViv(i));

should be rewritten as:

    rpp_push_1_norc(newSViv(i));

This is because newSViv() and similar create a new SV with a reference
count one too high (1 rather than 0). This count is then "donated" to the
stack by pushing it. Conversely on non-RC builds, the count is donated to
the TEMPs stack.

Similarly, on RC builds, C<sv = rpp_pop_1_norc()> does a simple
C<sv = *PL_stack_sv--> without adjusting the reference count, while on
non-RC builds it actually increments the SV's reference count. It is
intended for cases where you immediately want to increment the reference
count again after popping, e.g. where the SV is to be immediately embedded
somewhere. For example this code:

    SV *sv = PL_stack_sp[0];
    SvREFCNT_inc(sv);
    av_store(av, i, sv); /* in real life should check return value */
    rpp_popfree_1();

can be more efficiently written as

    av_store(av, i, rpp_pop_1_norc());

By using this function, the code works correctly on both RC and non-RC
builds.

A common operation on list ops is to impose void, scalar or list context
on the return arguments, possibly discarding all, or all except one, of
them. rpp_context(mark, gimme, extra) does this. As a first step (for
convenience and efficiency) it notionally pops C<extra> args off the
stack. Then for list context, leaves things as is. For void context, the
stack pointer is reset to mark, and everything above is popped. For
scalar, the top argument (or &PL_sv_undef) is moved from the top to
mark+1 and everything above is discarded.

The macros which appear at the start of many PP functions to check for
unary or binary op overloading (among other things) have been replaced
with rpp_try_AMAGIC_1() and _2() inline functions, which now rely on the
calling PP function to choose whether to return immediately rather than
the return being hidden away in the macro.

The rpp_invoke_xs() function calls the XS function associated with the CV,
but may do so via a wrapper function to adjust the stack as necessary.

In the spirit of hiding away less in macros, C<dATARGET> hasn't been given
a replacement; where its effect is needed, it is now written out in full;
see pp_add() for an example.

Finally, a couple of rpp() functions provide information rather than
manipulate the stack.

rpp_is_lone(sv) indicates whether C<sv>, assumed to be still on the stack,
it kept alive only by a single reference-counted pointer from the argument
and/or temps stacks, and thus is a candidate for some optimisations (like
skipping the copying of return arguments from a subroutine call).

rpp_stack_is_rc() indicates whether the current stack is currently
reference-counted. It's used mainly in a few places like call_sv() which
can be called from anywhere, and thus have to deal with both cases.

So for example, rather than using rpp_xpush_1(), call_sv() has lines like:

        rpp_extend(1);
        *++PL_stack_sp = sv;
    #ifdef PERL_RC_STACK
        if (rpp_stack_is_rc())
            SvREFCNT_inc_simple_void_NN(sv);
    #endif

which works on both standard builds and RC builds, and works whether
call_sv() is called from a standard PP function (rpp_stack_is_rc() is
true) or from a wrapped PP or XS function (rpp_stack_is_rc() is false).
Note that you're unlikely to need to use this function, as in most places,
such as PP or XS functions, it is always RC or non-RC respectively. In
fact on debugging builds under C<PERL_RC_STACK>, PUSHs() and similar
macros include an C<assert(!rpp_stack_is_rc())>, while rpp_push_1() and
similar functions have C<assert(rpp_stack_is_rc())>.

The macros for pushing new stackinfos have been replaced with inline
functions which don't rely on C<dSP> being in scope, and which have less
ambiguous names: they make it clear that a new I<stackinfo> is being
pushed, rather than just some sort of I<stack>. push_stackinfo() also has
a boolean argument indicating whether the new argument stack should be
reference-counted or not. For backwards compatibility, PUSHSTACKi(type) is
defined to be push_stackinfo(type, 0).

Some test scripts check for things like leaks by testing that the
reference count of a particular variable has an expected value. If this
is different on a perl built with C<PERL_RC_STACK>, then the perl
function Internals::stack_refcounted() can be used. This returns an
integer, the lowest bit of which indicates that perl was built with
C<PERL_RC_STACK>. Other bits are reserved for future use and should be
masked out.

=head1 Slab-based operator allocation

B<Note:> this section describes a non-public internal API that is subject
to change without notice.

Perl's internal error-handling mechanisms implement C<die> (and its internal
equivalents) using longjmp. If this occurs during lexing, parsing or
compilation, we must ensure that any ops allocated as part of the compilation
process are freed. (Older Perl versions did not adequately handle this
situation: when failing a parse, they would leak ops that were stored in
C C<auto> variables and not linked anywhere else.)

To handle this situation, Perl uses I<op slabs> that are attached to the
currently-compiling CV. A slab is a chunk of allocated memory. New ops are
allocated as regions of the slab. If the slab fills up, a new one is created
(and linked from the previous one). When an error occurs and the CV is freed,
any ops remaining are freed.

Each op is preceded by two pointers: one points to the next op in the slab, and
the other points to the slab that owns it. The next-op pointer is needed so
that Perl can iterate over a slab and free all its ops. (Op structures are of
different sizes, so the slab's ops can't merely be treated as a dense array.)
The slab pointer is needed for accessing a reference count on the slab: when
the last op on a slab is freed, the slab itself is freed.

The slab allocator puts the ops at the end of the slab first. This will tend to
allocate the leaves of the op tree first, and the layout will therefore
hopefully be cache-friendly. In addition, this means that there's no need to
store the size of the slab (see below on why slabs vary in size), because Perl
can follow pointers to find the last op.

It might seem possible to eliminate slab reference counts altogether, by having
all ops implicitly attached to C<PL_compcv> when allocated and freed when the
CV is freed. That would also allow C<op_free> to skip C<FreeOp> altogether, and
thus free ops faster. But that doesn't work in those cases where ops need to
survive beyond their CVs, such as re-evals.

The CV also has to have a reference count on the slab. Sometimes the first op
created is immediately freed. If the reference count of the slab reaches 0,
then it will be freed with the CV still pointing to it.

CVs use the C<CVf_SLABBED> flag to indicate that the CV has a reference count
on the slab. When this flag is set, the slab is accessible via C<CvSTART> when
C<CvROOT> is not set, or by subtracting two pointers C<(2*sizeof(I32 *))> from
C<CvROOT> when it is set. The alternative to this approach of sneaking the slab
into C<CvSTART> during compilation would be to enlarge the C<xpvcv> struct by
another pointer. But that would make all CVs larger, even though slab-based op
freeing is typically of benefit only for programs that make significant use of
string eval.

=for apidoc_section $concurrency
=for apidoc  Cmnh|    |CVf_SLABBED
=for apidoc_item |OP *|CvROOT|CV * sv
=for apidoc_item |OP *|CvSTART|CV * sv

When the C<CVf_SLABBED> flag is set, the CV takes responsibility for freeing
the slab. If C<CvROOT> is not set when the CV is freed or undeffed, it is
assumed that a compilation error has occurred, so the op slab is traversed and
all the ops are freed.

Under normal circumstances, the CV forgets about its slab (decrementing the
reference count) when the root is attached. So the slab reference counting that
happens when ops are freed takes care of freeing the slab. In some cases, the
CV is told to forget about the slab (C<cv_forget_slab>) precisely so that the
ops can survive after the CV is done away with.

Forgetting the slab when the root is attached is not strictly necessary, but
avoids potential problems with C<CvROOT> being written over. There is code all
over the place, both in core and on CPAN, that does things with C<CvROOT>, so
forgetting the slab makes things more robust and avoids potential problems.

Since the CV takes ownership of its slab when flagged, that flag is never
copied when a CV is cloned, as one CV could free a slab that another CV still
points to, since forced freeing of ops ignores the reference count (but asserts
that it looks right).

To avoid slab fragmentation, freed ops are marked as freed and attached to the
slab's freed chain (an idea stolen from DBM::Deep). Those freed ops are reused
when possible. Not reusing freed ops would be simpler, but it would result in
significantly higher memory usage for programs with large C<if (DEBUG) {...}>
blocks.

C<SAVEFREEOP> is slightly problematic under this scheme. Sometimes it can cause
an op to be freed after its CV. If the CV has forcibly freed the ops on its
slab and the slab itself, then we will be fiddling with a freed slab. Making
C<SAVEFREEOP> a no-op doesn't help, as sometimes an op can be savefreed when
there is no compilation error, so the op would never be freed. It holds
a reference count on the slab, so the whole slab would leak. So C<SAVEFREEOP>
now sets a special flag on the op (C<< ->op_savefree >>). The forced freeing of
ops after a compilation error won't free any ops thus marked.

Since many pieces of code create tiny subroutines consisting of only a few ops,
and since a huge slab would be quite a bit of baggage for those to carry
around, the first slab is always very small. To avoid allocating too many
slabs for a single CV, each subsequent slab is twice the size of the previous.

Smartmatch expects to be able to allocate an op at run time, run it, and then
throw it away. For that to work the op is simply malloced when C<PL_compcv> hasn't
been set up. So all slab-allocated ops are marked as such (C<< ->op_slabbed >>),
to distinguish them from malloced ops.


=head1 AUTHORS

Until May 1997, this document was maintained by Jeff Okamoto
E<lt>okamoto@corp.hp.comE<gt>.  It is now maintained as part of Perl
itself by the Perl 5 Porters E<lt>perl5-porters@perl.orgE<gt>.

With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
Stephen McCamant, and Gurusamy Sarathy.

=head1 SEE ALSO

L<perlapi>, L<perlintern>, L<perlxs>, L<perlembed>