1 /*-------------------------------------------------------------------------
2  *
3  * psqlscan_int.h
4  *	  lexical scanner internal declarations
5  *
6  * This file declares the PsqlScanStateData structure used by psqlscan.l
7  * and shared by other lexers compatible with it, such as psqlscanslash.l.
8  *
9  * One difficult aspect of this code is that we need to work in multibyte
10  * encodings that are not ASCII-safe.  A "safe" encoding is one in which each
11  * byte of a multibyte character has the high bit set (it's >= 0x80).  Since
12  * all our lexing rules treat all high-bit-set characters alike, we don't
13  * really need to care whether such a byte is part of a sequence or not.
14  * In an "unsafe" encoding, we still expect the first byte of a multibyte
15  * sequence to be >= 0x80, but later bytes might not be.  If we scan such
16  * a sequence as-is, the lexing rules could easily be fooled into matching
17  * such bytes to ordinary ASCII characters.  Our solution for this is to
18  * substitute 0xFF for each non-first byte within the data presented to flex.
19  * The flex rules will then pass the FF's through unmolested.  The
20  * psqlscan_emit() subroutine is responsible for looking back to the original
21  * string and replacing FF's with the corresponding original bytes.
22  *
23  * Another interesting thing we do here is scan different parts of the same
24  * input with physically separate flex lexers (ie, lexers written in separate
25  * .l files).  We can get away with this because the only part of the
26  * persistent state of a flex lexer that depends on its parsing rule tables
27  * is the start state number, which is easy enough to manage --- usually,
28  * in fact, we just need to set it to INITIAL when changing lexers.  But to
29  * make that work at all, we must use re-entrant lexers, so that all the
30  * relevant state is in the yyscan_t attached to the PsqlScanState;
31  * if we were using lexers with separate static state we would soon end up
32  * with dangling buffer pointers in one or the other.  Also note that this
33  * is unlikely to work very nicely if the lexers aren't all built with the
34  * same flex version, or if they don't use the same flex options.
35  *
36  *
37  * Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
38  * Portions Copyright (c) 1994, Regents of the University of California
39  *
40  * src/include/fe_utils/psqlscan_int.h
41  *
42  *-------------------------------------------------------------------------
43  */
44 #ifndef PSQLSCAN_INT_H
45 #define PSQLSCAN_INT_H
46 
47 #include "fe_utils/psqlscan.h"
48 
49 /*
50  * These are just to allow this file to be compilable standalone for header
51  * validity checking; in actual use, this file should always be included
52  * from the body of a flex file, where these symbols are already defined.
53  */
54 #ifndef YY_TYPEDEF_YY_BUFFER_STATE
55 #define YY_TYPEDEF_YY_BUFFER_STATE
56 typedef struct yy_buffer_state *YY_BUFFER_STATE;
57 #endif
58 #ifndef YY_TYPEDEF_YY_SCANNER_T
59 #define YY_TYPEDEF_YY_SCANNER_T
60 typedef void *yyscan_t;
61 #endif
62 
63 /*
64  * We use a stack of flex buffers to handle substitution of psql variables.
65  * Each stacked buffer contains the as-yet-unread text from one psql variable.
66  * When we pop the stack all the way, we resume reading from the outer buffer
67  * identified by scanbufhandle.
68  */
69 typedef struct StackElem
70 {
71 	YY_BUFFER_STATE buf;		/* flex input control structure */
72 	char	   *bufstring;		/* data actually being scanned by flex */
73 	char	   *origstring;		/* copy of original data, if needed */
74 	char	   *varname;		/* name of variable providing data, or NULL */
75 	struct StackElem *next;
76 } StackElem;
77 
78 /*
79  * All working state of the lexer must be stored in PsqlScanStateData
80  * between calls.  This allows us to have multiple open lexer operations,
81  * which is needed for nested include files.  The lexer itself is not
82  * recursive, but it must be re-entrant.
83  */
84 typedef struct PsqlScanStateData
85 {
86 	yyscan_t	scanner;		/* Flex's state for this PsqlScanState */
87 
88 	PQExpBuffer output_buf;		/* current output buffer */
89 
90 	StackElem  *buffer_stack;	/* stack of variable expansion buffers */
91 
92 	/*
93 	 * These variables always refer to the outer buffer, never to any stacked
94 	 * variable-expansion buffer.
95 	 */
96 	YY_BUFFER_STATE scanbufhandle;
97 	char	   *scanbuf;		/* start of outer-level input buffer */
98 	const char *scanline;		/* current input line at outer level */
99 
100 	/* safe_encoding, curline, refline are used by emit() to replace FFs */
101 	int			encoding;		/* encoding being used now */
102 	bool		safe_encoding;	/* is current encoding "safe"? */
103 	bool		std_strings;	/* are string literals standard? */
104 	const char *curline;		/* actual flex input string for cur buf */
105 	const char *refline;		/* original data for cur buffer */
106 
107 	/*
108 	 * All this state lives across successive input lines, until explicitly
109 	 * reset by psql_scan_reset.  start_state is adopted by yylex() on entry,
110 	 * and updated with its finishing state on exit.
111 	 */
112 	int			start_state;	/* yylex's starting/finishing state */
113 	int			state_before_str_stop;	/* start cond. before end quote */
114 	int			paren_depth;	/* depth of nesting in parentheses */
115 	int			xcdepth;		/* depth of nesting in slash-star comments */
116 	char	   *dolqstart;		/* current $foo$ quote start string */
117 
118 	/*
119 	 * State to track boundaries of BEGIN ... END blocks in function
120 	 * definitions, so that semicolons do not send query too early.
121 	 */
122 	int			identifier_count;	/* identifiers since start of statement */
123 	char		identifiers[4]; /* records the first few identifiers */
124 	int			begin_depth;	/* depth of begin/end pairs */
125 
126 	/*
127 	 * Callback functions provided by the program making use of the lexer,
128 	 * plus a void* callback passthrough argument.
129 	 */
130 	const PsqlScanCallbacks *callbacks;
131 	void	   *cb_passthrough;
132 } PsqlScanStateData;
133 
134 
135 /*
136  * Functions exported by psqlscan.l, but only meant for use within
137  * compatible lexers.
138  */
139 extern void psqlscan_push_new_buffer(PsqlScanState state,
140 									 const char *newstr, const char *varname);
141 extern void psqlscan_pop_buffer_stack(PsqlScanState state);
142 extern void psqlscan_select_top_buffer(PsqlScanState state);
143 extern bool psqlscan_var_is_current_source(PsqlScanState state,
144 										   const char *varname);
145 extern YY_BUFFER_STATE psqlscan_prepare_buffer(PsqlScanState state,
146 											   const char *txt, int len,
147 											   char **txtcopy);
148 extern void psqlscan_emit(PsqlScanState state, const char *txt, int len);
149 extern char *psqlscan_extract_substring(PsqlScanState state,
150 										const char *txt, int len);
151 extern void psqlscan_escape_variable(PsqlScanState state,
152 									 const char *txt, int len,
153 									 PsqlScanQuoteType quote);
154 extern void psqlscan_test_variable(PsqlScanState state,
155 								   const char *txt, int len);
156 
157 #endif							/* PSQLSCAN_INT_H */
158