doc/pod/libstorage.pod

=head1 NAME

libstorage - routines for managing INN storage

=head1 SYNOPSIS

    #include <inn/storage.h>

    typedef enum {
        RETR_ALL,
        RETR_HEAD,
        RETR_BODY,
        RETR_STAT
    } RETRTYPE;

    typedef enum {
        SM_RDWR,
        SM_PREOPEN
    } SMSETUP;

    typedef unsigned char STORAGECLASS;
    typedef unsigned char STORAGETYPE;

    typedef struct token {
        STORAGETYPE  type;
        STORAGECLASS class;
        char         token[STORAGE_TOKEN_LENGTH];
    } TOKEN;

    typedef struct {
        unsigned char type;
        const char    *data;
        struct iovec  *iov;
        int           iovcnt;
        size_t        len;
        unsigned char nextmethod;
        void          *private;
        time_t        arrived;
        time_t        expires;
        char          *groups;
        int           groupslen;
        TOKEN         *token;
    } ARTHANDLE;

    typedef enum {
        SELFEXPIRE,
        SMARTNGNUM,
        EXPENSIVESTAT
    } PROBETYPE;

    typedef enum {
        SM_ALL,
        SM_HEAD,
        SM_CANCELLEDART
    } FLUSHTYPE;

    struct artngnum {
        char   *groupname;
        ARTNUM artnum;
    };

    bool IsToken(const char *text);

    char *TokenToText(const TOKEN token);

    TOKEN TextToToken(const char *text);

    bool SMsetup(SMSETUP type, void *value);

    bool SMinit(void);

    TOKEN SMstore(const ARTHANDLE article);

    ARTHANDLE *SMretrieve(const TOKEN token, const RETRTYPE amount);

    ARTHANDLE *SMnext(const ARTHANDLE *article, const RETRTYPE amount);

    void SMfreearticle(ARTHANDLE *article);

    bool SMcancel(TOKEN token);

    bool SMprobe(PROBETYPE type, TOKEN *token, void *value);

    void SMprintfiles(FILE *file, TOKEN token, char **xref, int ngroups);

    bool SMflushcacheddata(FLUSHTYPE type);

    char *SMexplaintoken(const TOKEN token);

    void SMshutdown(void);

    int SMerrno;

    char *SMerrorstr;

    #include <inn/ov.h>

    #define OV_NOSPACE ...

    typedef enum {
        OVSPACE,
        OVSORT,
        OVCUTOFFLOW,
        OVGROUPBASEDEXPIRE,
        OVSTATICSEARCH,
        OVSTATALL,
        OVCACHEKEEP,
        OVCACHEFREE
    } OVCTLTYPE;

    typedef enum {
        OVNEWSGROUP,
        OVARRIVED,
        OVNOSORT
    } OVSORTTYPE;

    typedef enum {
        OVADDCOMPLETED,
        OVADDFAILED,
        OVADDGROUPNOMATCH
    } OVADDRESULT;

    bool OVopen(int mode);

    bool OVctl(OVCTLTYPE type, void *val);

    bool OVgroupstats(char *group, int *lo, int *hi, int *count, int *flag);

    bool OVgroupadd(char *group, ARTNUM lo, ARTNUM hi, char *flag);

    bool OVgroupdel(char *group);

    OVADDRESULT OVadd(TOKEN token, char *data, int len, time_t arrived, time_t expires);

    bool OVcancel(TOKEN token);

    void *OVopensearch(char *group, int low, int high);

    bool OVsearch(void *handle, ARTNUM *artnum, char **data, int *len, TOKEN *token, time_t *arrived);

    void OVclosesearch(void *handle);

    bool OVgetartinfo(char *group, ARTNUM artnum, TOKEN *token);

    bool OVexpiregroup(char *group, int *lo, struct history *h);

    typedef struct _OVGE {
        bool   delayrm;
        bool   usepost;
        bool   quiet;
        bool   keep;
        bool   earliest;
        bool   ignoreselfexpire;
        char   *filename;
        time_t now;
        float  timewarp;
    } OVGE;

    void OVclose(void);

=head1 DESCRIPTION

B<libstorage> is a library of common utility (the storage manager) routines
for accessing Usenet articles and related data independent of particular
storage method; this is known as the storage API.

The storage manager's function is to isolate the applications from the
individual methods and make the policy decisions as to which storage method
should be called to store an article.  One of the core concepts in the
storage API is that articles are not manipulated using the message-ID or
article number, but rather a token that uniquely identifies the article
to the method that stored it.  This may seem to be redundant since the
message-ID already is a unique identifier for the article; however, since
the storage method generates the token, it can encode all the information
it needs to locate the article in the token.

B<OV> is a common utility routines for accessing newsgroups and overview
data independent of particular overview method.

The B<OV> function is to isolate the applications from the individual
methods.  All articles passed through the storage API are assumed to
be in wire format.  Wire format means C<\r\n> at the end of lines,
dot-stuffed lines (that is to say C<.> at the beginning of lines which
already start with C<.>), and C<.\r\n> at the end of article on NNTP stream
are not stripped.  This has a performance win when transferring articles.
Note that for the tradspool method, wire format can be disabled.  This is
just for compatibility which is needed by some old tools written for
traditional spool.

The B<IsToken> function checks to see if the text is formatted as a text
token string.  It returns true if formatted correctly or returns false
if not.

The B<TokenToText> function converts a token into a text string for output.

The B<TextToToken> function converts a text string into a token.

The B<SMsetup> function configures some parameters for use by the storage
manager.  I<type> is one of following:

=over 4

=item C<SM_RDWR>

Allow read/write open for storage files (default is false).

=item C<SM_PREOPEN>

Open all storage files at startup time and keep them (default is false).

=back

I<value> is the pointer which tells each type's value.  It returns true
if setup is successful, or false if not.

The B<SMinit> function calls the setup function for all of the configured
methods based on B<SMsetup>.  This function should be called prior to all
other storage API functions which begin with C<SM> except B<SMsetup>.  It
returns true if initialization is successful or returns false if not.
B<SMinit> returns true, unless all storage methods fail initialization.

The B<SMstore> function stores an article specified with I<article>.
The headers and body of the article are supplied to B<SMstore> using the
I<iov> and I<iovcnt> members of B<ARTHANDLE>.  (I<data> and I<private>
are ignored by B<SMstore>.)  If I<arrived> is specified, B<SMstore> uses
its value as article's arrival time; otherwise B<SMstore> uses the current
time for it.  B<SMstore> returns the token type or returns B<TOKEN_EMPTY>
if the article is not stored because some error occurs or simply does not
match any uwildmat(3) expression in F<storage.conf>.  B<SMstore> fails if
B<SM_RDWR> has not been set to true with B<SMsetup>.

The B<SMretrieve> function retrieves an article specified with I<token>.
I<amount> is the one of following which specifies retrieving type:

=over 4

=item C<RETR_ALL>

Retrieve the whole article.

=item C<RETR_HEAD>

Retrieve the headers of the article.

=item C<RETR_BODY>

Retrieve the body of the article.

=item C<RETR_STAT>

Just check to see if the article exists.

=back

B<SMretrieve> provides the article data via the I<data> and I<len> members
of B<ARTHANDLE>.  (I<iov> is not set by B<SMretrieve>.)  The data area
indicated by B<ARTHANDLE> should not be modified.

The B<SMnext> function is similar in function to B<SMretrieve> except that it
is intended for traversing the method's article store sequentially.  To start
a query, B<SMnext> should be called with a NULL pointer B<ARTHANDLE>.
Then B<SMnext> returns B<ARTHANDLE> which should be used for the next
query.  If a NULL pointer B<ARTHANDLE> is returned, no articles are left
to be queried.  If I<data> of B<ARTHANDLE> is NULL pointer or I<len> of
B<ARTHANDLE> is C<0>, it indicates the article may be corrupted and should
be cancelled by I<SMcancel>.  The data area indicated by B<ARTHANDLE>
should not be modified.

The B<SMfreearticle> function frees all allocated memory used by
B<SMretrieve> and B<SMnext>.  If B<SMnext> will be called with previously
returned B<ARTHANDLE>, B<SMfreearticle> should not be called as B<SMnext>
frees allocated memory internally.

The B<SMcancel> function removes the article specified with I<token>.
It returns true if cancellation is successful or returns false if not.
B<SMcancel> fails if B<SM_RDWR> has not been set to true with B<SMsetup>.

The B<SMprobe> function checks the token on B<PROBETYPE>.  I<type> is
one of following:

=over 4

=item C<SELFEXPIRE>

Check to see if the method of the token
has self expire functionality.

=item C<SMARTNGNUM>

Get the newsgroup name and
the article number of the token.

=item C<EXPENSIVESTAT>

Check to see whether
checking the existence of an article is expensive or not.

=back

The B<SMprintfiles> function shows file name or token usable by fastrm(8).

The B<SMflushcacheddata> function flushes cached data on each storage
method.  I<type> is one of following:

=over 4

=item C<SM_HEAD>

Flush cached header.

=item C<SM_CANCELLEDART>

Flush the articles which should be cancelled.

=item C<SM_ALL>

Flush all cached data.

=back

The B<SMexplaintoken> function returns a human-readable text string with
a clear, decoded form of the storage API token.

The B<SMshutdown> function calls the shutdown for each configured storage
method and then frees any resources it has allocated for itself.

B<SMerrno> and B<SMerrorstr> indicate the reason of the last error concerning
storage manager.

B<OVopen> calls the setup function for configured method which is specified
as I<ovmethod> in F<inn.conf>.  I<mode> is constructed from following:

=over 4

=item C<OV_READ>

Allow read open for the overview method.

=item C<OV_WRITE>

Allow write open for the overview method.

=back

This function should be called prior to all other OV functions which begin
with C<OV>.

The B<OVctl> function probes or sets some parameters for the overview method.
I<type> is one of following:

=over 4

=item C<OVGROUPBASEDEXPIRE>

Setup how group-based expiry is done.

=item C<OVCUTOFFLOW>

Do not add overview data if the data is under the lowest article.

=item C<OVSORT>

Probe which key is suitable for the overview method.

=item C<OVSPACE>

Probe the overview space usage.

=item C<OVSTATALL>

Stat all the articles when B<OVexpiregroup> is called.

=item C<OVSTATICSEARCH>

Setup if results of B<OVsearch> are stored in a static buffer and must be
copied before the next call to B<OVsearch>.

=item C<OVCACHEKEEP>

Setup whether the cache should be kept.

=item C<OVCACHEFREE>

Free the cache.

=back

The B<OVgroupstats> function retrieves the specified newsgroup information
from the overview method.

The B<OVgroupadd> function informs the overview method that the specified
newsgroup is being added.

The B<OVgroupdel> function informs the overview method that the specified
newsgroup is being removed.

The B<OVadd> function stores an overview data.

The B<OVcancel> function requests the overview method delete overview data
specified with token.

The B<OVopensearch> function requests the overview method prepare overview
data retrieval.  The request range is determined by I<low> and I<high>.
The setting of B<OVSTATICSEARCH> determines how search result data
must be handled.  (Note that with some storage methods, each call to
B<OVopensearch> may cause internal storage to be remapped.  Therefore,
multiple simultaneous searches may require data to be copied in between
B<OVsearch> calls even if B<OVSTATICSEARCH> is false.)

The B<OVsearch> function retrieves information, article number, overview
data, or arrival time.  It should be called with NULL handle when it
is the first time; subsequent B<OVsearch> calls should use the handle
returned by the previous call to B<OVsearch>.  B<OVsearch> returns
true, unless it reaches I<high>, which is specified by B<OVopensearch>.
Retrieved overview data are sorted by article number, and I<len> is C<0>
if there is no overview data for the article.  Note that the retrieved
data is not necessarily null-terminated; you should only rely on I<len>
octets of overview data being present.

The B<OVclosesearch> function frees all resources which have been allocated
by B<OVopensearch>.

The B<OVgetartinfo> function retrieves the overview data and the token
specified with I<artnum>.

The B<OVexpiregroup> function expires the overview data for the newsgroup.
It checks the existence of the article and purges the overview data if the
article no longer exists.  If I<groupbaseexpiry> in F<inn.conf> is true,
B<OVexpiregroup> also expires articles.

The B<OVclose> function frees all resources which are used by the overview
method.

=head1 HISTORY

Written by Katsuhiro Kondou <kondou@nec.co.jp> for InterNetNews.
Converted to POD by Julien Elie.

=head1 SEE ALSO

expire(8), fastrm(8), inn.conf(5), storage.conf(5).

=cut