Line data Source code
1 : %top{
2 : /*-------------------------------------------------------------------------
3 : *
4 : * psqlscan.l
5 : * lexical scanner for SQL commands
6 : *
7 : * This lexer used to be part of psql, and that heritage is reflected in
8 : * the file name as well as function and typedef names, though it can now
9 : * be used by other frontend programs as well. It's also possible to extend
10 : * this lexer with a compatible add-on lexer to handle program-specific
11 : * backslash commands.
12 : *
13 : * This code is mainly concerned with determining where the end of a SQL
14 : * statement is: we are looking for semicolons that are not within quotes,
15 : * comments, or parentheses. The most reliable way to handle this is to
16 : * borrow the backend's flex lexer rules, lock, stock, and barrel. The rules
17 : * below are (except for a few) the same as the backend's, but their actions
18 : * are just ECHO whereas the backend's actions generally do other things.
19 : *
20 : * XXX The rules in this file must be kept in sync with the backend lexer!!!
21 : *
22 : * XXX Avoid creating backtracking cases --- see the backend lexer for info.
23 : *
24 : * See psqlscan_int.h for additional commentary.
25 : *
26 : *
27 : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
28 : * Portions Copyright (c) 1994, Regents of the University of California
29 : *
30 : * IDENTIFICATION
31 : * src/fe_utils/psqlscan.l
32 : *
33 : *-------------------------------------------------------------------------
34 : */
35 : #include "postgres_fe.h"
36 :
37 : #include "common/logging.h"
38 : #include "fe_utils/psqlscan.h"
39 :
40 : #include "libpq-fe.h"
41 : }
42 :
43 : %{
44 :
45 : /* LCOV_EXCL_START */
46 :
47 : #include "fe_utils/psqlscan_int.h"
48 :
49 : /*
50 : * We must have a typedef YYSTYPE for yylex's first argument, but this lexer
51 : * doesn't presently make use of that argument, so just declare it as int.
52 : */
53 : typedef int YYSTYPE;
54 :
55 :
56 : /* Return values from yylex() */
57 : #define LEXRES_EOL 0 /* end of input */
58 : #define LEXRES_SEMI 1 /* command-terminating semicolon found */
59 : #define LEXRES_BACKSLASH 2 /* backslash command start */
60 :
61 :
62 : #define ECHO psqlscan_emit(cur_state, yytext, yyleng)
63 :
64 : %}
65 :
66 : %option reentrant
67 : %option bison-bridge
68 : %option 8bit
69 : %option never-interactive
70 : %option nodefault
71 : %option noinput
72 : %option nounput
73 : %option noyywrap
74 : %option warn
75 : %option prefix="psql_yy"
76 :
77 : /*
78 : * Set the type of yyextra; we use it as a pointer back to the containing
79 : * PsqlScanState.
80 : */
81 : %option extra-type="PsqlScanState"
82 :
83 : /*
84 : * All of the following definitions and rules should exactly match
85 : * src/backend/parser/scan.l so far as the flex patterns are concerned.
86 : * The rule bodies are just ECHO as opposed to what the backend does,
87 : * however. (But be sure to duplicate code that affects the lexing process,
88 : * such as BEGIN() and yyless().) Also, psqlscan uses a single <<EOF>> rule
89 : * whereas scan.l has a separate one for each exclusive state.
90 : */
91 :
92 : /*
93 : * OK, here is a short description of lex/flex rules behavior.
94 : * The longest pattern which matches an input string is always chosen.
95 : * For equal-length patterns, the first occurring in the rules list is chosen.
96 : * INITIAL is the starting state, to which all non-conditional rules apply.
97 : * Exclusive states change parsing rules while the state is active. When in
98 : * an exclusive state, only those rules defined for that state apply.
99 : *
100 : * We use exclusive states for quoted strings, extended comments,
101 : * and to eliminate parsing troubles for numeric strings.
102 : * Exclusive states:
103 : * <xb> bit string literal
104 : * <xc> extended C-style comments
105 : * <xd> delimited identifiers (double-quoted identifiers)
106 : * <xh> hexadecimal byte string
107 : * <xq> standard quoted strings
108 : * <xqs> quote stop (detect continued strings)
109 : * <xe> extended quoted strings (support backslash escape sequences)
110 : * <xdolq> $foo$ quoted strings
111 : * <xui> quoted identifier with Unicode escapes
112 : * <xus> quoted string with Unicode escapes
113 : *
114 : * Note: we intentionally don't mimic the backend's <xeu> state; we have
115 : * no need to distinguish it from <xe> state, and no good way to get out
116 : * of it in error cases. The backend just throws yyerror() in those
117 : * cases, but that's not an option here.
118 : */
119 :
120 : %x xb
121 : %x xc
122 : %x xd
123 : %x xh
124 : %x xq
125 : %x xqs
126 : %x xe
127 : %x xdolq
128 : %x xui
129 : %x xus
130 :
131 : /*
132 : * In order to make the world safe for Windows and Mac clients as well as
133 : * Unix ones, we accept either \n or \r as a newline. A DOS-style \r\n
134 : * sequence will be seen as two successive newlines, but that doesn't cause
135 : * any problems. Comments that start with -- and extend to the next
136 : * newline are treated as equivalent to a single whitespace character.
137 : *
138 : * NOTE a fine point: if there is no newline following --, we will absorb
139 : * everything to the end of the input as a comment. This is correct. Older
140 : * versions of Postgres failed to recognize -- as a comment if the input
141 : * did not end with a newline.
142 : *
143 : * non_newline_space tracks all space characters except newlines.
144 : *
145 : * XXX if you change the set of whitespace characters, fix scanner_isspace()
146 : * to agree.
147 : */
148 :
149 : space [ \t\n\r\f\v]
150 : non_newline_space [ \t\f\v]
151 : newline [\n\r]
152 : non_newline [^\n\r]
153 :
154 : comment ("--"{non_newline}*)
155 :
156 : whitespace ({space}+|{comment})
157 :
158 : /*
159 : * SQL requires at least one newline in the whitespace separating
160 : * string literals that are to be concatenated. Silly, but who are we
161 : * to argue? Note that {whitespace_with_newline} should not have * after
162 : * it, whereas {whitespace} should generally have a * after it...
163 : */
164 :
165 : special_whitespace ({space}+|{comment}{newline})
166 : non_newline_whitespace ({non_newline_space}|{comment})
167 : whitespace_with_newline ({non_newline_whitespace}*{newline}{special_whitespace}*)
168 :
169 : quote '
170 : /* If we see {quote} then {quotecontinue}, the quoted string continues */
171 : quotecontinue {whitespace_with_newline}{quote}
172 :
173 : /*
174 : * {quotecontinuefail} is needed to avoid lexer backup when we fail to match
175 : * {quotecontinue}. It might seem that this could just be {whitespace}*,
176 : * but if there's a dash after {whitespace_with_newline}, it must be consumed
177 : * to see if there's another dash --- which would start a {comment} and thus
178 : * allow continuation of the {quotecontinue} token.
179 : */
180 : quotecontinuefail {whitespace}*"-"?
181 :
182 : /* Bit string
183 : * It is tempting to scan the string for only those characters
184 : * which are allowed. However, this leads to silently swallowed
185 : * characters if illegal characters are included in the string.
186 : * For example, if xbinside is [01] then B'ABCD' is interpreted
187 : * as a zero-length string, and the ABCD' is lost!
188 : * Better to pass the string forward and let the input routines
189 : * validate the contents.
190 : */
191 : xbstart [bB]{quote}
192 : xbinside [^']*
193 :
194 : /* Hexadecimal byte string */
195 : xhstart [xX]{quote}
196 : xhinside [^']*
197 :
198 : /* National character */
199 : xnstart [nN]{quote}
200 :
201 : /* Quoted string that allows backslash escapes */
202 : xestart [eE]{quote}
203 : xeinside [^\\']+
204 : xeescape [\\][^0-7]
205 : xeoctesc [\\][0-7]{1,3}
206 : xehexesc [\\]x[0-9A-Fa-f]{1,2}
207 : xeunicode [\\](u[0-9A-Fa-f]{4}|U[0-9A-Fa-f]{8})
208 : xeunicodefail [\\](u[0-9A-Fa-f]{0,3}|U[0-9A-Fa-f]{0,7})
209 :
210 : /* Extended quote
211 : * xqdouble implements embedded quote, ''''
212 : */
213 : xqstart {quote}
214 : xqdouble {quote}{quote}
215 : xqinside [^']+
216 :
217 : /* $foo$ style quotes ("dollar quoting")
218 : * The quoted string starts with $foo$ where "foo" is an optional string
219 : * in the form of an identifier, except that it may not contain "$",
220 : * and extends to the first occurrence of an identical string.
221 : * There is *no* processing of the quoted text.
222 : *
223 : * {dolqfailed} is an error rule to avoid scanner backup when {dolqdelim}
224 : * fails to match its trailing "$".
225 : */
226 : dolq_start [A-Za-z\200-\377_]
227 : dolq_cont [A-Za-z\200-\377_0-9]
228 : dolqdelim \$({dolq_start}{dolq_cont}*)?\$
229 : dolqfailed \${dolq_start}{dolq_cont}*
230 : dolqinside [^$]+
231 :
232 : /* Double quote
233 : * Allows embedded spaces and other special characters into identifiers.
234 : */
235 : dquote \"
236 : xdstart {dquote}
237 : xdstop {dquote}
238 : xddouble {dquote}{dquote}
239 : xdinside [^"]+
240 :
241 : /* Quoted identifier with Unicode escapes */
242 : xuistart [uU]&{dquote}
243 :
244 : /* Quoted string with Unicode escapes */
245 : xusstart [uU]&{quote}
246 :
247 : /* error rule to avoid backup */
248 : xufailed [uU]&
249 :
250 :
251 : /* C-style comments
252 : *
253 : * The "extended comment" syntax closely resembles allowable operator syntax.
254 : * The tricky part here is to get lex to recognize a string starting with
255 : * slash-star as a comment, when interpreting it as an operator would produce
256 : * a longer match --- remember lex will prefer a longer match! Also, if we
257 : * have something like plus-slash-star, lex will think this is a 3-character
258 : * operator whereas we want to see it as a + operator and a comment start.
259 : * The solution is two-fold:
260 : * 1. append {op_chars}* to xcstart so that it matches as much text as
261 : * {operator} would. Then the tie-breaker (first matching rule of same
262 : * length) ensures xcstart wins. We put back the extra stuff with yyless()
263 : * in case it contains a star-slash that should terminate the comment.
264 : * 2. In the operator rule, check for slash-star within the operator, and
265 : * if found throw it back with yyless(). This handles the plus-slash-star
266 : * problem.
267 : * Dash-dash comments have similar interactions with the operator rule.
268 : */
269 : xcstart \/\*{op_chars}*
270 : xcstop \*+\/
271 : xcinside [^*/]+
272 :
273 : ident_start [A-Za-z\200-\377_]
274 : ident_cont [A-Za-z\200-\377_0-9\$]
275 :
276 : identifier {ident_start}{ident_cont}*
277 :
278 : /* Assorted special-case operators and operator-like tokens */
279 : typecast "::"
280 : dot_dot \.\.
281 : colon_equals ":="
282 :
283 : /*
284 : * These operator-like tokens (unlike the above ones) also match the {operator}
285 : * rule, which means that they might be overridden by a longer match if they
286 : * are followed by a comment start or a + or - character. Accordingly, if you
287 : * add to this list, you must also add corresponding code to the {operator}
288 : * block to return the correct token in such cases. (This is not needed in
289 : * psqlscan.l since the token value is ignored there.)
290 : */
291 : equals_greater "=>"
292 : less_equals "<="
293 : greater_equals ">="
294 : less_greater "<>"
295 : not_equals "!="
296 :
297 : /*
298 : * "self" is the set of chars that should be returned as single-character
299 : * tokens. "op_chars" is the set of chars that can make up "Op" tokens,
300 : * which can be one or more characters long (but if a single-char token
301 : * appears in the "self" set, it is not to be returned as an Op). Note
302 : * that the sets overlap, but each has some chars that are not in the other.
303 : *
304 : * If you change either set, adjust the character lists appearing in the
305 : * rule for "operator"!
306 : */
307 : self [,()\[\].;\:\+\-\*\/\%\^\<\>\=]
308 : op_chars [\~\!\@\#\^\&\|\`\?\+\-\*\/\%\<\>\=]
309 : operator {op_chars}+
310 :
311 : /*
312 : * Numbers
313 : *
314 : * Unary minus is not part of a number here. Instead we pass it separately to
315 : * the parser, and there it gets coerced via doNegate().
316 : *
317 : * {numericfail} is used because we would like "1..10" to lex as 1, dot_dot, 10.
318 : *
319 : * {realfail} is added to prevent the need for scanner
320 : * backup when the {real} rule fails to match completely.
321 : */
322 : decdigit [0-9]
323 : hexdigit [0-9A-Fa-f]
324 : octdigit [0-7]
325 : bindigit [0-1]
326 :
327 : decinteger {decdigit}(_?{decdigit})*
328 : hexinteger 0[xX](_?{hexdigit})+
329 : octinteger 0[oO](_?{octdigit})+
330 : bininteger 0[bB](_?{bindigit})+
331 :
332 : hexfail 0[xX]_?
333 : octfail 0[oO]_?
334 : binfail 0[bB]_?
335 :
336 : numeric (({decinteger}\.{decinteger}?)|(\.{decinteger}))
337 : numericfail {decinteger}\.\.
338 :
339 : real ({decinteger}|{numeric})[Ee][-+]?{decinteger}
340 : realfail ({decinteger}|{numeric})[Ee][-+]
341 :
342 : /* Positional parameters don't accept underscores. */
343 : param \${decdigit}+
344 :
345 : /*
346 : * An identifier immediately following an integer literal is disallowed because
347 : * in some cases it's ambiguous what is meant: for example, 0x1234 could be
348 : * either a hexinteger or a decinteger "0" and an identifier "x1234". We can
349 : * detect such problems by seeing if integer_junk matches a longer substring
350 : * than any of the XXXinteger patterns (decinteger, hexinteger, octinteger,
351 : * bininteger). One "junk" pattern is sufficient because
352 : * {decinteger}{identifier} will match all the same strings we'd match with
353 : * {hexinteger}{identifier} etc.
354 : *
355 : * Note that the rule for integer_junk must appear after the ones for
356 : * XXXinteger to make this work correctly: 0x1234 will match both hexinteger
357 : * and integer_junk, and we need hexinteger to be chosen in that case.
358 : *
359 : * Also disallow strings matched by numeric_junk, real_junk and param_junk
360 : * for consistency.
361 : */
362 : integer_junk {decinteger}{identifier}
363 : numeric_junk {numeric}{identifier}
364 : real_junk {real}{identifier}
365 : param_junk \${decdigit}+{identifier}
366 :
367 : /* psql-specific: characters allowed in variable names */
368 : variable_char [A-Za-z\200-\377_0-9]
369 :
370 : other .
371 :
372 : /*
373 : * Dollar quoted strings are totally opaque, and no escaping is done on them.
374 : * Other quoted strings must allow some special characters such as single-quote
375 : * and newline.
376 : * Embedded single-quotes are implemented both in the SQL standard
377 : * style of two adjacent single quotes "''" and in the Postgres/Java style
378 : * of escaped-quote "\'".
379 : * Other embedded escaped characters are matched explicitly and the leading
380 : * backslash is dropped from the string.
381 : * Note that xcstart must appear before operator, as explained above!
382 : * Also whitespace (comment) must appear before operator.
383 : */
384 :
385 : %%
386 :
387 : %{
388 : /* Declare some local variables inside yylex(), for convenience */
389 : PsqlScanState cur_state = yyextra;
390 610442 : PQExpBuffer output_buf = cur_state->output_buf;
391 610442 :
392 : /*
393 : * Force flex into the state indicated by start_state. This has a
394 : * couple of purposes: it lets some of the functions below set a new
395 : * starting state without ugly direct access to flex variables, and it
396 : * allows us to transition from one flex lexer to another so that we
397 : * can lex different parts of the source string using separate lexers.
398 : */
399 : BEGIN(cur_state->start_state);
400 610442 : %}
401 :
402 : {whitespace} {
403 : /*
404 : * Note that the whitespace rule includes both true
405 : * whitespace and single-line ("--" style) comments.
406 : * We suppress whitespace until we have collected some
407 : * non-whitespace data. (This interacts with some
408 : * decisions in MainLoop(); see there for details.)
409 : */
410 : if (output_buf->len > 0)
411 1412344 : ECHO;
412 1328122 : }
413 :
414 1412344 : {xcstart} {
415 345 : cur_state->xcdepth = 0;
416 345 : BEGIN(xc);
417 345 : /* Put back any characters past slash-star; see above */
418 : yyless(2);
419 345 : ECHO;
420 345 : }
421 :
422 345 : <xc>{
423 : {xcstart} {
424 9 : cur_state->xcdepth++;
425 9 : /* Put back any characters past slash-star; see above */
426 : yyless(2);
427 9 : ECHO;
428 9 : }
429 :
430 9 : {xcstop} {
431 354 : if (cur_state->xcdepth <= 0)
432 354 : BEGIN(INITIAL);
433 345 : else
434 : cur_state->xcdepth--;
435 9 : ECHO;
436 354 : }
437 :
438 354 : {xcinside} {
439 802 : ECHO;
440 802 : }
441 :
442 802 : {op_chars} {
443 217 : ECHO;
444 217 : }
445 :
446 217 : \*+ {
447 0 : ECHO;
448 0 : }
449 : } /* <xc> */
450 0 :
451 : {xbstart} {
452 375 : BEGIN(xb);
453 375 : ECHO;
454 375 : }
455 : <xh>{xhinside} |
456 375 : <xb>{xbinside} {
457 2025 : ECHO;
458 2025 : }
459 :
460 2025 : {xhstart} {
461 1665 : /* Hexadecimal bit type.
462 : * At some point we should simply pass the string
463 : * forward to the parser and label it there.
464 : * In the meantime, place a leading "x" on the string
465 : * to mark it for the input routine as a hex string.
466 : */
467 : BEGIN(xh);
468 1665 : ECHO;
469 1665 : }
470 :
471 1665 : {xnstart} {
472 0 : yyless(1); /* eat only 'n' this time */
473 0 : ECHO;
474 0 : }
475 :
476 0 : {xqstart} {
477 119710 : if (cur_state->std_strings)
478 119710 : BEGIN(xq);
479 119710 : else
480 : BEGIN(xe);
481 0 : ECHO;
482 119710 : }
483 : {xestart} {
484 119710 : BEGIN(xe);
485 669 : ECHO;
486 669 : }
487 : {xusstart} {
488 669 : BEGIN(xus);
489 277 : ECHO;
490 277 : }
491 :
492 277 : <xb,xh,xq,xe,xus>{quote} {
493 122696 : /*
494 : * When we are scanning a quoted string and see an end
495 : * quote, we must look ahead for a possible continuation.
496 : * If we don't see one, we know the end quote was in fact
497 : * the end of the string. To reduce the lexer table size,
498 : * we use a single "xqs" state to do the lookahead for all
499 : * types of strings.
500 : */
501 : cur_state->state_before_str_stop = YYSTATE;
502 122696 : BEGIN(xqs);
503 122696 : ECHO;
504 122696 : }
505 : <xqs>{quotecontinue} {
506 122696 : /*
507 0 : * Found a quote continuation, so return to the in-quote
508 : * state and continue scanning the literal. Nothing is
509 : * added to the literal's contents.
510 : */
511 : BEGIN(cur_state->state_before_str_stop);
512 0 : ECHO;
513 0 : }
514 : <xqs>{quotecontinuefail} |
515 0 : <xqs>{other} {
516 121908 : /*
517 : * Failed to see a quote continuation. Throw back
518 : * everything after the end quote, and handle the string
519 : * according to the state we were in previously.
520 : */
521 : yyless(0);
522 121908 : BEGIN(INITIAL);
523 121908 : /* There's nothing to echo ... */
524 : }
525 :
526 121908 : <xq,xe,xus>{xqdouble} {
527 3158 : ECHO;
528 3158 : }
529 : <xq,xus>{xqinside} {
530 3158 : ECHO;
531 125617 : }
532 : <xe>{xeinside} {
533 125617 : ECHO;
534 1343 : }
535 : <xe>{xeunicode} {
536 1343 : ECHO;
537 108 : }
538 : <xe>{xeunicodefail} {
539 108 : ECHO;
540 6 : }
541 : <xe>{xeescape} {
542 6 : ECHO;
543 746 : }
544 : <xe>{xeoctesc} {
545 746 : ECHO;
546 12 : }
547 : <xe>{xehexesc} {
548 12 : ECHO;
549 6 : }
550 : <xe>. {
551 6 : /* This is only needed for \ just before EOF */
552 0 : ECHO;
553 0 : }
554 :
555 0 : {dolqdelim} {
556 3634 : cur_state->dolqstart = pg_strdup(yytext);
557 3634 : BEGIN(xdolq);
558 3634 : ECHO;
559 3634 : }
560 : {dolqfailed} {
561 3634 : /* throw back all but the initial "$" */
562 0 : yyless(1);
563 0 : ECHO;
564 0 : }
565 : <xdolq>{dolqdelim} {
566 0 : if (strcmp(yytext, cur_state->dolqstart) == 0)
567 3806 : {
568 : free(cur_state->dolqstart);
569 3634 : cur_state->dolqstart = NULL;
570 3634 : BEGIN(INITIAL);
571 3634 : }
572 : else
573 : {
574 : /*
575 : * When we fail to match $...$ to dolqstart, transfer
576 : * the $... part to the output, but put back the final
577 : * $ for rescanning. Consider $delim$...$junk$delim$
578 : */
579 : yyless(yyleng - 1);
580 172 : }
581 : ECHO;
582 3806 : }
583 : <xdolq>{dolqinside} {
584 3806 : ECHO;
585 19016 : }
586 : <xdolq>{dolqfailed} {
587 19016 : ECHO;
588 559 : }
589 : <xdolq>. {
590 559 : /* This is only needed for $ inside the quoted text */
591 1222 : ECHO;
592 1222 : }
593 :
594 1222 : {xdstart} {
595 5015 : BEGIN(xd);
596 5015 : ECHO;
597 5015 : }
598 : {xuistart} {
599 5015 : BEGIN(xui);
600 12 : ECHO;
601 12 : }
602 : <xd>{xdstop} {
603 12 : BEGIN(INITIAL);
604 5015 : ECHO;
605 5015 : }
606 : <xui>{dquote} {
607 5015 : BEGIN(INITIAL);
608 12 : ECHO;
609 12 : }
610 : <xd,xui>{xddouble} {
611 12 : ECHO;
612 61 : }
613 : <xd,xui>{xdinside} {
614 61 : ECHO;
615 5081 : }
616 :
617 5081 : {xufailed} {
618 0 : /* throw back all but the initial u/U */
619 : yyless(1);
620 0 : ECHO;
621 0 : }
622 :
623 0 : {typecast} {
624 27564 : ECHO;
625 27564 : }
626 :
627 27564 : {dot_dot} {
628 0 : ECHO;
629 0 : }
630 :
631 0 : {colon_equals} {
632 1574 : ECHO;
633 1574 : }
634 :
635 1574 : {equals_greater} {
636 979 : ECHO;
637 979 : }
638 :
639 979 : {less_equals} {
640 1484 : ECHO;
641 1484 : }
642 :
643 1484 : {greater_equals} {
644 3897 : ECHO;
645 3897 : }
646 :
647 3897 : {less_greater} {
648 640 : ECHO;
649 640 : }
650 :
651 640 : {not_equals} {
652 1162 : ECHO;
653 1162 : }
654 :
655 1162 : /*
656 : * These rules are specific to psql --- they implement parenthesis
657 : * counting and detection of command-ending semicolon. These must
658 : * appear before the {self} rule so that they take precedence over it.
659 : */
660 :
661 191237 : "(" {
662 : cur_state->paren_depth++;
663 191237 : ECHO;
664 191237 : }
665 :
666 191237 : ")" {
667 191230 : if (cur_state->paren_depth > 0)
668 191230 : cur_state->paren_depth--;
669 191230 : ECHO;
670 191230 : }
671 :
672 191230 : ";" {
673 184214 : ECHO;
674 184214 : if (cur_state->paren_depth == 0 && cur_state->begin_depth == 0)
675 184214 : {
676 : /* Terminate lexing temporarily */
677 : cur_state->start_state = YY_START;
678 184110 : cur_state->identifier_count = 0;
679 184110 : return LEXRES_SEMI;
680 184110 : }
681 : }
682 :
683 104 : /*
684 : * psql-specific rules to handle backslash commands and variable
685 : * substitution. We want these before {self}, also.
686 : */
687 :
688 397 : "\\"[;:] {
689 : /* Force a semi-colon or colon into the query buffer */
690 : psqlscan_emit(cur_state, yytext + 1, 1);
691 397 : if (yytext[1] == ';')
692 397 : cur_state->identifier_count = 0;
693 397 : }
694 :
695 397 : "\\" {
696 28941 : /* Terminate lexing temporarily */
697 : cur_state->start_state = YY_START;
698 28941 : return LEXRES_BACKSLASH;
699 28941 : }
700 :
701 : :{variable_char}+ {
702 1496 : /* Possible psql variable substitution */
703 : char *varname;
704 : char *value;
705 :
706 : varname = psqlscan_extract_substring(cur_state,
707 1496 : yytext + 1,
708 1496 : yyleng - 1);
709 1496 : if (cur_state->callbacks->get_variable)
710 1496 : value = cur_state->callbacks->get_variable(varname,
711 900 : PQUOTE_PLAIN,
712 : cur_state->cb_passthrough);
713 : else
714 : value = NULL;
715 596 :
716 : if (value)
717 1496 : {
718 : /* It is a variable, check for recursion */
719 : if (psqlscan_var_is_current_source(cur_state, varname))
720 669 : {
721 : /* Recursive expansion --- don't go there */
722 : pg_log_warning("skipping recursive expansion of variable \"%s\"",
723 0 : varname);
724 : /* Instead copy the string as is */
725 : ECHO;
726 0 : }
727 : else
728 : {
729 : /* OK, perform substitution */
730 : psqlscan_push_new_buffer(cur_state, value, varname);
731 669 : /* yy_scan_string already made buffer active */
732 : }
733 : free(value);
734 669 : }
735 : else
736 : {
737 : /*
738 : * if the variable doesn't exist we'll copy the string
739 : * as is
740 : */
741 : ECHO;
742 827 : }
743 :
744 : free(varname);
745 1496 : }
746 :
747 1496 : :'{variable_char}+' {
748 467 : psqlscan_escape_variable(cur_state, yytext, yyleng,
749 467 : PQUOTE_SQL_LITERAL);
750 : }
751 :
752 467 : :\"{variable_char}+\" {
753 17 : psqlscan_escape_variable(cur_state, yytext, yyleng,
754 17 : PQUOTE_SQL_IDENT);
755 : }
756 :
757 17 : :\{\?{variable_char}+\} {
758 6 : psqlscan_test_variable(cur_state, yytext, yyleng);
759 6 : }
760 :
761 6 : /*
762 : * These rules just avoid the need for scanner backup if one of the
763 : * three rules above fails to match completely.
764 : */
765 :
766 0 : :'{variable_char}* {
767 : /* Throw back everything but the colon */
768 : yyless(1);
769 0 : ECHO;
770 0 : }
771 :
772 0 : :\"{variable_char}* {
773 0 : /* Throw back everything but the colon */
774 : yyless(1);
775 0 : ECHO;
776 0 : }
777 :
778 0 : :\{\?{variable_char}* {
779 0 : /* Throw back everything but the colon */
780 : yyless(1);
781 0 : ECHO;
782 0 : }
783 : :\{ {
784 0 : /* Throw back everything but the colon */
785 0 : yyless(1);
786 0 : ECHO;
787 0 : }
788 :
789 0 : /*
790 : * Back to backend-compatible rules.
791 : */
792 :
793 333060 : {self} {
794 : ECHO;
795 333060 : }
796 :
797 333060 : {operator} {
798 10124 : /*
799 : * Check for embedded slash-star or dash-dash; those
800 : * are comment starts, so operator must stop there.
801 : * Note that slash-star or dash-dash at the first
802 : * character will match a prior rule, not this one.
803 : */
804 : int nchars = yyleng;
805 10124 : char *slashstar = strstr(yytext, "/*");
806 10124 : char *dashdash = strstr(yytext, "--");
807 10124 :
808 : if (slashstar && dashdash)
809 10124 : {
810 : /* if both appear, take the first one */
811 : if (slashstar > dashdash)
812 0 : slashstar = dashdash;
813 0 : }
814 : else if (!slashstar)
815 10124 : slashstar = dashdash;
816 10094 : if (slashstar)
817 10124 : nchars = slashstar - yytext;
818 36 :
819 : /*
820 : * For SQL compatibility, '+' and '-' cannot be the
821 : * last char of a multi-char operator unless the operator
822 : * contains chars that are not in SQL operators.
823 : * The idea is to lex '=-' as two operators, but not
824 : * to forbid operator names like '?-' that could not be
825 : * sequences of SQL operators.
826 : */
827 : if (nchars > 1 &&
828 10124 : (yytext[nchars - 1] == '+' ||
829 9242 : yytext[nchars - 1] == '-'))
830 9239 : {
831 : int ic;
832 :
833 : for (ic = nchars - 2; ic >= 0; ic--)
834 279 : {
835 : char c = yytext[ic];
836 240 : if (c == '~' || c == '!' || c == '@' ||
837 240 : c == '#' || c == '^' || c == '&' ||
838 198 : c == '|' || c == '`' || c == '?' ||
839 75 : c == '%')
840 : break;
841 : }
842 : if (ic < 0)
843 213 : {
844 : /*
845 : * didn't find a qualifying character, so remove
846 : * all trailing [+-]
847 : */
848 : do {
849 : nchars--;
850 39 : } while (nchars > 1 &&
851 39 : (yytext[nchars - 1] == '+' ||
852 18 : yytext[nchars - 1] == '-'));
853 18 : }
854 : }
855 :
856 : if (nchars < yyleng)
857 10124 : {
858 : /* Strip the unwanted chars from the token */
859 : yyless(nchars);
860 75 : }
861 : ECHO;
862 10124 : }
863 :
864 10124 : {param} {
865 607 : ECHO;
866 607 : }
867 : {param_junk} {
868 607 : ECHO;
869 6 : }
870 :
871 6 : {decinteger} {
872 106670 : ECHO;
873 106670 : }
874 : {hexinteger} {
875 106670 : ECHO;
876 67 : }
877 : {octinteger} {
878 67 : ECHO;
879 30 : }
880 : {bininteger} {
881 30 : ECHO;
882 30 : }
883 : {hexfail} {
884 30 : ECHO;
885 3 : }
886 : {octfail} {
887 3 : ECHO;
888 3 : }
889 : {binfail} {
890 3 : ECHO;
891 3 : }
892 : {numeric} {
893 3 : ECHO;
894 4000 : }
895 : {numericfail} {
896 4000 : /* throw back the .., and treat as integer */
897 0 : yyless(yyleng - 2);
898 0 : ECHO;
899 0 : }
900 : {real} {
901 0 : ECHO;
902 316 : }
903 : {realfail} {
904 316 : ECHO;
905 3 : }
906 : {integer_junk} {
907 3 : ECHO;
908 33 : }
909 : {numeric_junk} {
910 33 : ECHO;
911 24 : }
912 : {real_junk} {
913 24 : ECHO;
914 0 : }
915 :
916 0 :
917 1370161 : {identifier} {
918 : /*
919 : * We need to track if we are inside a BEGIN .. END block
920 : * in a function definition, so that semicolons contained
921 : * therein don't terminate the whole statement. Short of
922 : * writing a full parser here, the following heuristic
923 : * should work. First, we track whether the beginning of
924 : * the statement matches CREATE [OR REPLACE]
925 : * {FUNCTION|PROCEDURE}
926 : */
927 :
928 : if (cur_state->identifier_count == 0)
929 1370161 : memset(cur_state->identifiers, 0, sizeof(cur_state->identifiers));
930 190646 :
931 : if (pg_strcasecmp(yytext, "create") == 0 ||
932 2708162 : pg_strcasecmp(yytext, "function") == 0 ||
933 2670531 : pg_strcasecmp(yytext, "procedure") == 0 ||
934 2663307 : pg_strcasecmp(yytext, "or") == 0 ||
935 2658212 : pg_strcasecmp(yytext, "replace") == 0)
936 1327435 : {
937 : if (cur_state->identifier_count < sizeof(cur_state->identifiers))
938 43876 : cur_state->identifiers[cur_state->identifier_count] = pg_tolower((unsigned char) yytext[0]);
939 39541 : }
940 :
941 : cur_state->identifier_count++;
942 1370161 :
943 : if (cur_state->identifiers[0] == 'c' &&
944 1370161 : (cur_state->identifiers[1] == 'f' || cur_state->identifiers[1] == 'p' ||
945 321874 : (cur_state->identifiers[1] == 'o' && cur_state->identifiers[2] == 'r' &&
946 292595 : (cur_state->identifiers[3] == 'f' || cur_state->identifiers[3] == 'p'))) &&
947 11436 : cur_state->paren_depth == 0)
948 36477 : {
949 : if (pg_strcasecmp(yytext, "begin") == 0)
950 29201 : cur_state->begin_depth++;
951 68 : else if (pg_strcasecmp(yytext, "case") == 0)
952 29133 : {
953 : /*
954 : * CASE also ends with END. We only need to track
955 : * this if we are already inside a BEGIN.
956 : */
957 : if (cur_state->begin_depth >= 1)
958 3 : cur_state->begin_depth++;
959 3 : }
960 : else if (pg_strcasecmp(yytext, "end") == 0)
961 29130 : {
962 : if (cur_state->begin_depth > 0)
963 71 : cur_state->begin_depth--;
964 71 : }
965 : }
966 :
967 : ECHO;
968 1370161 : }
969 :
970 1370161 : {other} {
971 0 : ECHO;
972 0 : }
973 :
974 0 : <<EOF>> {
975 398060 : if (cur_state->buffer_stack == NULL)
976 398060 : {
977 : cur_state->start_state = YY_START;
978 397391 : return LEXRES_EOL; /* end of input reached */
979 397391 : }
980 :
981 : /*
982 : * We were expanding a variable, so pop the inclusion
983 : * stack and keep lexing
984 : */
985 : psqlscan_pop_buffer_stack(cur_state);
986 669 : psqlscan_select_top_buffer(cur_state);
987 669 : }
988 :
989 669 : %%
990 0 :
991 : /* LCOV_EXCL_STOP */
992 :
993 : /*
994 : * Create a lexer working state struct.
995 : *
996 : * callbacks is a struct of function pointers that encapsulate some
997 : * behavior we need from the surrounding program. This struct must
998 : * remain valid for the lifespan of the PsqlScanState.
999 : */
1000 : PsqlScanState
1001 : psql_scan_create(const PsqlScanCallbacks *callbacks)
1002 10023 : {
1003 : PsqlScanState state;
1004 :
1005 : state = pg_malloc0_object(PsqlScanStateData);
1006 10023 :
1007 : state->callbacks = callbacks;
1008 10023 :
1009 : yylex_init(&state->scanner);
1010 10023 :
1011 : yyset_extra(state, state->scanner);
1012 10023 :
1013 : psql_scan_reset(state);
1014 10023 :
1015 : return state;
1016 10023 : }
1017 :
1018 : /*
1019 : * Destroy a lexer working state struct, releasing all resources.
1020 : */
1021 : void
1022 : psql_scan_destroy(PsqlScanState state)
1023 9967 : {
1024 : psql_scan_finish(state);
1025 9967 :
1026 : psql_scan_reset(state);
1027 9967 :
1028 : yylex_destroy(state->scanner);
1029 9967 :
1030 : free(state);
1031 9967 : }
1032 9967 :
1033 : /*
1034 : * Set the callback passthrough pointer for the lexer.
1035 : *
1036 : * This could have been integrated into psql_scan_create, but keeping it
1037 : * separate allows the application to change the pointer later, which might
1038 : * be useful.
1039 : */
1040 : void
1041 : psql_scan_set_passthrough(PsqlScanState state, void *passthrough)
1042 9420 : {
1043 : state->cb_passthrough = passthrough;
1044 9420 : }
1045 9420 :
1046 : /*
1047 : * Set up to perform lexing of the given input line.
1048 : *
1049 : * The text at *line, extending for line_len bytes, will be scanned by
1050 : * subsequent calls to the psql_scan routines. psql_scan_finish should
1051 : * be called when scanning is complete. Note that the lexer retains
1052 : * a pointer to the storage at *line --- this string must not be altered
1053 : * or freed until after psql_scan_finish is called.
1054 : *
1055 : * encoding is the libpq identifier for the character encoding in use,
1056 : * and std_strings says whether standard_conforming_strings is on.
1057 : */
1058 : void
1059 : psql_scan_setup(PsqlScanState state,
1060 397686 : const char *line, int line_len,
1061 : int encoding, bool std_strings)
1062 : {
1063 : /* Mustn't be scanning already */
1064 : Assert(state->scanbufhandle == NULL);
1065 : Assert(state->buffer_stack == NULL);
1066 :
1067 : /* Do we need to hack the character set encoding? */
1068 : state->encoding = encoding;
1069 397686 : state->safe_encoding = pg_valid_server_encoding_id(encoding);
1070 397686 :
1071 : /* Save standard-strings flag as well */
1072 : state->std_strings = std_strings;
1073 397686 :
1074 : /* Set up flex input buffer with appropriate translation and padding */
1075 : state->scanbufhandle = psqlscan_prepare_buffer(state, line, line_len,
1076 397686 : &state->scanbuf);
1077 : state->scanline = line;
1078 397686 :
1079 : /* Set lookaside data in case we have to map unsafe encoding */
1080 : state->curline = state->scanbuf;
1081 397686 : state->refline = state->scanline;
1082 397686 :
1083 : /* Initialize state for psql_scan_get_location() */
1084 : state->cur_line_no = 0; /* yylex not called yet */
1085 397686 : state->cur_line_ptr = state->scanbuf;
1086 397686 : }
1087 397686 :
1088 : /*
1089 : * Do lexical analysis of SQL command text.
1090 : *
1091 : * The text previously passed to psql_scan_setup is scanned, and appended
1092 : * (possibly with transformation) to query_buf.
1093 : *
1094 : * The return value indicates the condition that stopped scanning:
1095 : *
1096 : * PSCAN_SEMICOLON: found a command-ending semicolon. (The semicolon is
1097 : * transferred to query_buf.) The command accumulated in query_buf should
1098 : * be executed, then clear query_buf and call again to scan the remainder
1099 : * of the line.
1100 : *
1101 : * PSCAN_BACKSLASH: found a backslash that starts a special command.
1102 : * Any previous data on the line has been transferred to query_buf.
1103 : * The caller will typically next apply a separate flex lexer to scan
1104 : * the special command.
1105 : *
1106 : * PSCAN_INCOMPLETE: the end of the line was reached, but we have an
1107 : * incomplete SQL command. *prompt is set to the appropriate prompt type.
1108 : *
1109 : * PSCAN_EOL: the end of the line was reached, and there is no lexical
1110 : * reason to consider the command incomplete. The caller may or may not
1111 : * choose to send it. *prompt is set to the appropriate prompt type if
1112 : * the caller chooses to collect more input.
1113 : *
1114 : * In the PSCAN_INCOMPLETE and PSCAN_EOL cases, psql_scan_finish() should
1115 : * be called next, then the cycle may be repeated with a fresh input line.
1116 : *
1117 : * In all cases, *prompt is set to an appropriate prompt type code for the
1118 : * next line-input operation.
1119 : */
1120 : PsqlScanResult
1121 : psql_scan(PsqlScanState state,
1122 610442 : PQExpBuffer query_buf,
1123 : promptStatus_t *prompt)
1124 : {
1125 : PsqlScanResult result;
1126 : int lexresult;
1127 :
1128 : /* Must be scanning already */
1129 : Assert(state->scanbufhandle != NULL);
1130 :
1131 : /* Set current output target */
1132 : state->output_buf = query_buf;
1133 610442 :
1134 : /* Set input source */
1135 : if (state->buffer_stack != NULL)
1136 610442 : yy_switch_to_buffer(state->buffer_stack->buf, state->scanner);
1137 45 : else
1138 : yy_switch_to_buffer(state->scanbufhandle, state->scanner);
1139 610397 :
1140 : /* And lex. */
1141 : lexresult = yylex(NULL, state->scanner);
1142 610442 :
1143 : /* Notify psql_scan_get_location() that a yylex call has been made. */
1144 : if (state->cur_line_no == 0)
1145 610442 : state->cur_line_no = 1;
1146 397684 :
1147 : /*
1148 : * Check termination state and return appropriate result info.
1149 : */
1150 : switch (lexresult)
1151 610442 : {
1152 : case LEXRES_EOL: /* end of input */
1153 397391 : switch (state->start_state)
1154 397391 : {
1155 : case INITIAL:
1156 372428 : case xqs: /* we treat this like INITIAL */
1157 : if (state->paren_depth > 0)
1158 372428 : {
1159 : result = PSCAN_INCOMPLETE;
1160 32051 : *prompt = PROMPT_PAREN;
1161 32051 : }
1162 : else if (state->begin_depth > 0)
1163 340377 : {
1164 : result = PSCAN_INCOMPLETE;
1165 427 : *prompt = PROMPT_CONTINUE;
1166 427 : }
1167 : else if (query_buf->len > 0)
1168 339950 : {
1169 : result = PSCAN_EOL;
1170 73351 : *prompt = PROMPT_CONTINUE;
1171 73351 : }
1172 : else
1173 : {
1174 : /* never bother to send an empty buffer */
1175 : result = PSCAN_INCOMPLETE;
1176 266599 : *prompt = PROMPT_READY;
1177 266599 : }
1178 : break;
1179 372428 : case xb:
1180 0 : result = PSCAN_INCOMPLETE;
1181 0 : *prompt = PROMPT_SINGLEQUOTE;
1182 0 : break;
1183 0 : case xc:
1184 401 : result = PSCAN_INCOMPLETE;
1185 401 : *prompt = PROMPT_COMMENT;
1186 401 : break;
1187 401 : case xd:
1188 19 : result = PSCAN_INCOMPLETE;
1189 19 : *prompt = PROMPT_DOUBLEQUOTE;
1190 19 : break;
1191 19 : case xh:
1192 0 : result = PSCAN_INCOMPLETE;
1193 0 : *prompt = PROMPT_SINGLEQUOTE;
1194 0 : break;
1195 0 : case xe:
1196 301 : result = PSCAN_INCOMPLETE;
1197 301 : *prompt = PROMPT_SINGLEQUOTE;
1198 301 : break;
1199 301 : case xq:
1200 5574 : result = PSCAN_INCOMPLETE;
1201 5574 : *prompt = PROMPT_SINGLEQUOTE;
1202 5574 : break;
1203 5574 : case xdolq:
1204 18668 : result = PSCAN_INCOMPLETE;
1205 18668 : *prompt = PROMPT_DOLLARQUOTE;
1206 18668 : break;
1207 18668 : case xui:
1208 0 : result = PSCAN_INCOMPLETE;
1209 0 : *prompt = PROMPT_DOUBLEQUOTE;
1210 0 : break;
1211 0 : case xus:
1212 0 : result = PSCAN_INCOMPLETE;
1213 0 : *prompt = PROMPT_SINGLEQUOTE;
1214 0 : break;
1215 0 : default:
1216 0 : /* can't get here */
1217 : fprintf(stderr, "invalid YY_START\n");
1218 0 : exit(1);
1219 0 : }
1220 : break;
1221 397391 : case LEXRES_SEMI: /* semicolon */
1222 184110 : result = PSCAN_SEMICOLON;
1223 184110 : *prompt = PROMPT_READY;
1224 184110 : break;
1225 184110 : case LEXRES_BACKSLASH: /* backslash */
1226 28941 : result = PSCAN_BACKSLASH;
1227 28941 : *prompt = PROMPT_READY;
1228 28941 : break;
1229 28941 : default:
1230 0 : /* can't get here */
1231 : fprintf(stderr, "invalid yylex result\n");
1232 0 : exit(1);
1233 0 : }
1234 :
1235 : return result;
1236 610442 : }
1237 :
1238 : /*
1239 : * Clean up after scanning a string. This flushes any unread input and
1240 : * releases resources (but not the PsqlScanState itself). Note however
1241 : * that this does not reset the lexer scan state; that can be done by
1242 : * psql_scan_reset(), which is an orthogonal operation.
1243 : *
1244 : * It is legal to call this when not scanning anything (makes it easier
1245 : * to deal with error recovery).
1246 : */
1247 : void
1248 : psql_scan_finish(PsqlScanState state)
1249 407278 : {
1250 : /* Drop any incomplete variable expansions. */
1251 : while (state->buffer_stack != NULL)
1252 407278 : psqlscan_pop_buffer_stack(state);
1253 0 :
1254 : /* Done with the outer scan buffer, too */
1255 : if (state->scanbufhandle)
1256 407278 : yy_delete_buffer(state->scanbufhandle, state->scanner);
1257 397631 : state->scanbufhandle = NULL;
1258 407278 : if (state->scanbuf)
1259 407278 : free(state->scanbuf);
1260 397631 : state->scanbuf = NULL;
1261 407278 : }
1262 407278 :
1263 : /*
1264 : * Reset lexer scanning state to start conditions. This is appropriate
1265 : * for executing \r psql commands (or any other time that we discard the
1266 : * prior contents of query_buf). It is not, however, necessary to do this
1267 : * when we execute and clear the buffer after getting a PSCAN_SEMICOLON or
1268 : * PSCAN_EOL scan result, because the scan state must be INITIAL when those
1269 : * conditions are returned.
1270 : *
1271 : * Note that this is unrelated to flushing unread input; that task is
1272 : * done by psql_scan_finish().
1273 : */
1274 : void
1275 : psql_scan_reset(PsqlScanState state)
1276 21592 : {
1277 : state->start_state = INITIAL;
1278 21592 : state->paren_depth = 0;
1279 21592 : state->xcdepth = 0; /* not really necessary */
1280 21592 : if (state->dolqstart)
1281 21592 : free(state->dolqstart);
1282 0 : state->dolqstart = NULL;
1283 21592 : state->identifier_count = 0;
1284 21592 : state->begin_depth = 0;
1285 21592 : }
1286 21592 :
1287 : /*
1288 : * Reselect this lexer (psqlscan.l) after using another one.
1289 : *
1290 : * Currently and for foreseeable uses, it's sufficient to reset to INITIAL
1291 : * state, because we'd never switch to another lexer in a different state.
1292 : * However, we don't want to reset e.g. paren_depth, so this can't be
1293 : * the same as psql_scan_reset().
1294 : *
1295 : * Note: psql setjmp error recovery just calls psql_scan_reset(), so that
1296 : * must be a superset of this.
1297 : *
1298 : * Note: it seems likely that other lexers could just assign INITIAL for
1299 : * themselves, since that probably has the value zero in every flex-generated
1300 : * lexer. But let's not assume that.
1301 : */
1302 : void
1303 : psql_scan_reselect_sql_lexer(PsqlScanState state)
1304 138436 : {
1305 : state->start_state = INITIAL;
1306 138436 : }
1307 138436 :
1308 : /*
1309 : * Return true if lexer is currently in an "inside quotes" state.
1310 : *
1311 : * This is pretty grotty but is needed to preserve the old behavior
1312 : * that mainloop.c drops blank lines not inside quotes without even
1313 : * echoing them.
1314 : */
1315 : bool
1316 : psql_scan_in_quote(PsqlScanState state)
1317 75664 : {
1318 : return state->start_state != INITIAL &&
1319 76156 : state->start_state != xqs;
1320 492 : }
1321 :
1322 : /*
1323 : * Return the current scanning location (end+1 of last scanned token),
1324 : * as a line number counted from 1 and an offset from string start.
1325 : *
1326 : * This considers only the outermost input string, and therefore is of
1327 : * limited use for programs that use psqlscan_push_new_buffer().
1328 : *
1329 : * It would be a bit easier probably to use "%option yylineno" to count
1330 : * lines, but the flex manual says that has a performance cost, and only
1331 : * a minority of programs using psqlscan have need for this functionality.
1332 : * So we implement it ourselves without adding overhead to the lexer itself.
1333 : */
1334 : void
1335 : psql_scan_get_location(PsqlScanState state,
1336 1737 : int *lineno, int *offset)
1337 : {
1338 : const char *line_end;
1339 :
1340 : /*
1341 : * We rely on flex's having stored a NUL after the current token in
1342 : * scanbuf. Therefore we must specially handle the state before yylex()
1343 : * has been called, when obviously that won't have happened yet.
1344 : */
1345 : if (state->cur_line_no == 0)
1346 1737 : {
1347 : *lineno = 1;
1348 0 : *offset = 0;
1349 0 : return;
1350 0 : }
1351 :
1352 : /*
1353 : * Advance cur_line_no/cur_line_ptr past whatever has been lexed so far.
1354 : * Doing this prevents repeated calls from being O(N^2) for long inputs.
1355 : */
1356 : while ((line_end = strchr(state->cur_line_ptr, '\n')) != NULL)
1357 2210 : {
1358 : state->cur_line_no++;
1359 473 : state->cur_line_ptr = line_end + 1;
1360 473 : }
1361 : state->cur_line_ptr += strlen(state->cur_line_ptr);
1362 1737 :
1363 : /* Report current location. */
1364 : *lineno = state->cur_line_no;
1365 1737 : *offset = state->cur_line_ptr - state->scanbuf;
1366 1737 : }
1367 :
1368 : /*
1369 : * Push the given string onto the stack of stuff to scan.
1370 : *
1371 : * NOTE SIDE EFFECT: the new buffer is made the active flex input buffer.
1372 : */
1373 : void
1374 : psqlscan_push_new_buffer(PsqlScanState state, const char *newstr,
1375 669 : const char *varname)
1376 : {
1377 : StackElem *stackelem;
1378 :
1379 : stackelem = pg_malloc_object(StackElem);
1380 669 :
1381 : /*
1382 : * In current usage, the passed varname points at the current flex input
1383 : * buffer; we must copy it before calling psqlscan_prepare_buffer()
1384 : * because that will change the buffer state.
1385 : */
1386 : stackelem->varname = varname ? pg_strdup(varname) : NULL;
1387 669 :
1388 : stackelem->buf = psqlscan_prepare_buffer(state, newstr, strlen(newstr),
1389 669 : &stackelem->bufstring);
1390 : state->curline = stackelem->bufstring;
1391 669 : if (state->safe_encoding)
1392 669 : {
1393 : stackelem->origstring = NULL;
1394 669 : state->refline = stackelem->bufstring;
1395 669 : }
1396 : else
1397 : {
1398 : stackelem->origstring = pg_strdup(newstr);
1399 0 : state->refline = stackelem->origstring;
1400 0 : }
1401 : stackelem->next = state->buffer_stack;
1402 669 : state->buffer_stack = stackelem;
1403 669 : }
1404 669 :
1405 : /*
1406 : * Pop the topmost buffer stack item (there must be one!)
1407 : *
1408 : * NB: after this, the flex input state is unspecified; caller must
1409 : * switch to an appropriate buffer to continue lexing.
1410 : * See psqlscan_select_top_buffer().
1411 : */
1412 : void
1413 : psqlscan_pop_buffer_stack(PsqlScanState state)
1414 669 : {
1415 : StackElem *stackelem = state->buffer_stack;
1416 669 :
1417 : state->buffer_stack = stackelem->next;
1418 669 : yy_delete_buffer(stackelem->buf, state->scanner);
1419 669 : free(stackelem->bufstring);
1420 669 : if (stackelem->origstring)
1421 669 : free(stackelem->origstring);
1422 0 : if (stackelem->varname)
1423 669 : free(stackelem->varname);
1424 669 : free(stackelem);
1425 669 : }
1426 669 :
1427 : /*
1428 : * Select the topmost surviving buffer as the active input.
1429 : */
1430 : void
1431 : psqlscan_select_top_buffer(PsqlScanState state)
1432 669 : {
1433 : StackElem *stackelem = state->buffer_stack;
1434 669 :
1435 : if (stackelem != NULL)
1436 669 : {
1437 : yy_switch_to_buffer(stackelem->buf, state->scanner);
1438 0 : state->curline = stackelem->bufstring;
1439 0 : state->refline = stackelem->origstring ? stackelem->origstring : stackelem->bufstring;
1440 0 : }
1441 : else
1442 : {
1443 : yy_switch_to_buffer(state->scanbufhandle, state->scanner);
1444 669 : state->curline = state->scanbuf;
1445 669 : state->refline = state->scanline;
1446 669 : }
1447 : }
1448 669 :
1449 : /*
1450 : * Check if specified variable name is the source for any string
1451 : * currently being scanned
1452 : */
1453 : bool
1454 : psqlscan_var_is_current_source(PsqlScanState state, const char *varname)
1455 669 : {
1456 : StackElem *stackelem;
1457 :
1458 : for (stackelem = state->buffer_stack;
1459 669 : stackelem != NULL;
1460 669 : stackelem = stackelem->next)
1461 0 : {
1462 : if (stackelem->varname && strcmp(stackelem->varname, varname) == 0)
1463 0 : return true;
1464 0 : }
1465 : return false;
1466 669 : }
1467 :
1468 : /*
1469 : * Set up a flex input buffer to scan the given data. We always make a
1470 : * copy of the data. If working in an unsafe encoding, the copy has
1471 : * multibyte sequences replaced by FFs to avoid fooling the lexer rules.
1472 : *
1473 : * NOTE SIDE EFFECT: the new buffer is made the active flex input buffer.
1474 : */
1475 : YY_BUFFER_STATE
1476 : psqlscan_prepare_buffer(PsqlScanState state, const char *txt, int len,
1477 398355 : char **txtcopy)
1478 : {
1479 : char *newtxt;
1480 :
1481 : /* Flex wants two \0 characters after the actual data */
1482 : newtxt = pg_malloc_array(char, (len + 2));
1483 398355 : *txtcopy = newtxt;
1484 398355 : newtxt[len] = newtxt[len + 1] = YY_END_OF_BUFFER_CHAR;
1485 398355 :
1486 : if (state->safe_encoding)
1487 398355 : memcpy(newtxt, txt, len);
1488 398215 : else
1489 : {
1490 : /* Gotta do it the hard way */
1491 : int i = 0;
1492 140 :
1493 : while (i < len)
1494 808 : {
1495 : int thislen = PQmblen(txt + i, state->encoding);
1496 668 :
1497 : /* first byte should always be okay... */
1498 : newtxt[i] = txt[i];
1499 668 : i++;
1500 668 : while (--thislen > 0 && i < len)
1501 808 : newtxt[i++] = (char) 0xFF;
1502 140 : }
1503 : }
1504 :
1505 : return yy_scan_buffer(newtxt, len + 2, state->scanner);
1506 398355 : }
1507 :
1508 : /*
1509 : * psqlscan_emit() --- body for ECHO macro
1510 : *
1511 : * NB: this must be used for ALL and ONLY the text copied from the flex
1512 : * input data. If you pass it something that is not part of the yytext
1513 : * string, you are making a mistake. Internally generated text can be
1514 : * appended directly to state->output_buf.
1515 : */
1516 : void
1517 : psqlscan_emit(PsqlScanState state, const char *txt, int len)
1518 5100502 : {
1519 : PQExpBuffer output_buf = state->output_buf;
1520 5100502 :
1521 : if (state->safe_encoding)
1522 5100502 : appendBinaryPQExpBuffer(output_buf, txt, len);
1523 5100026 : else
1524 : {
1525 : /* Gotta do it the hard way */
1526 : const char *reference = state->refline;
1527 476 : int i;
1528 :
1529 : reference += (txt - state->curline);
1530 476 :
1531 : for (i = 0; i < len; i++)
1532 1277 : {
1533 : char ch = txt[i];
1534 801 :
1535 : if (ch == (char) 0xFF)
1536 801 : ch = reference[i];
1537 140 : appendPQExpBufferChar(output_buf, ch);
1538 801 : }
1539 : }
1540 : }
1541 5100502 :
1542 : /*
1543 : * psqlscan_extract_substring --- fetch value of (part of) the current token
1544 : *
1545 : * This is like psqlscan_emit(), except that the data is returned as a
1546 : * malloc'd string rather than being pushed directly to state->output_buf.
1547 : */
1548 : char *
1549 : psqlscan_extract_substring(PsqlScanState state, const char *txt, int len)
1550 2670 : {
1551 : char *result = pg_malloc_array(char, (len + 1));
1552 2670 :
1553 : if (state->safe_encoding)
1554 2670 : memcpy(result, txt, len);
1555 2670 : else
1556 : {
1557 : /* Gotta do it the hard way */
1558 : const char *reference = state->refline;
1559 0 : int i;
1560 :
1561 : reference += (txt - state->curline);
1562 0 :
1563 : for (i = 0; i < len; i++)
1564 0 : {
1565 : char ch = txt[i];
1566 0 :
1567 : if (ch == (char) 0xFF)
1568 0 : ch = reference[i];
1569 0 : result[i] = ch;
1570 0 : }
1571 : }
1572 : result[len] = '\0';
1573 2670 : return result;
1574 2670 : }
1575 :
1576 : /*
1577 : * psqlscan_escape_variable --- process :'VARIABLE' or :"VARIABLE"
1578 : *
1579 : * If the variable name is found, escape its value using the appropriate
1580 : * quoting method and emit the value to output_buf. (Since the result is
1581 : * surely quoted, there is never any reason to rescan it.) If we don't
1582 : * find the variable or escaping fails, emit the token as-is.
1583 : */
1584 : void
1585 : psqlscan_escape_variable(PsqlScanState state, const char *txt, int len,
1586 519 : PsqlScanQuoteType quote)
1587 : {
1588 : char *varname;
1589 : char *value;
1590 :
1591 : /* Variable lookup. */
1592 : varname = psqlscan_extract_substring(state, txt + 2, len - 3);
1593 519 : if (state->callbacks->get_variable)
1594 519 : value = state->callbacks->get_variable(varname, quote,
1595 519 : state->cb_passthrough);
1596 : else
1597 : value = NULL;
1598 0 : free(varname);
1599 519 :
1600 : if (value)
1601 519 : {
1602 : /* Emit the suitably-escaped value */
1603 : appendPQExpBufferStr(state->output_buf, value);
1604 491 : free(value);
1605 491 : }
1606 : else
1607 : {
1608 : /* Emit original token as-is */
1609 : psqlscan_emit(state, txt, len);
1610 28 : }
1611 : }
1612 519 :
1613 : void
1614 : psqlscan_test_variable(PsqlScanState state, const char *txt, int len)
1615 16 : {
1616 : char *varname;
1617 : char *value;
1618 :
1619 : varname = psqlscan_extract_substring(state, txt + 3, len - 4);
1620 16 : if (state->callbacks->get_variable)
1621 16 : value = state->callbacks->get_variable(varname, PQUOTE_PLAIN,
1622 16 : state->cb_passthrough);
1623 : else
1624 : value = NULL;
1625 0 : free(varname);
1626 16 :
1627 : if (value != NULL)
1628 16 : {
1629 : appendPQExpBufferStr(state->output_buf, "TRUE");
1630 7 : free(value);
1631 7 : }
1632 : else
1633 : {
1634 : appendPQExpBufferStr(state->output_buf, "FALSE");
1635 9 : }
1636 : }
1637 16 : /* END: function "psqlscan_test_variable" */
|
