| blob | 941ed065532b916ceb8a251d0706681b6bd3e939 |
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) 1996-2022, 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"
37 #include "common/logging.h"
38 #include "fe_utils/psqlscan.h"
40 #include "libpq-fe.h"
41 }
43 %{
45 /* LCOV_EXCL_START */
47 #include "fe_utils/psqlscan_int.h"
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;
55 /*
56 * Set the type of yyextra; we use it as a pointer back to the containing
57 * PsqlScanState.
58 */
59 #define YY_EXTRA_TYPE PsqlScanState
62 /* Return values from yylex() */
63 #define LEXRES_EOL 0 /* end of input */
64 #define LEXRES_SEMI 1 /* command-terminating semicolon found */
65 #define LEXRES_BACKSLASH 2 /* backslash command start */
68 #define ECHO psqlscan_emit(cur_state, yytext, yyleng)
70 /*
71 * Work around a bug in flex 2.5.35: it emits a couple of functions that
72 * it forgets to emit declarations for. Since we use -Wmissing-prototypes,
73 * this would cause warnings. Providing our own declarations should be
74 * harmless even when the bug gets fixed.
75 */
76 extern int psql_yyget_column(yyscan_t yyscanner);
77 extern void psql_yyset_column(int column_no, yyscan_t yyscanner);
79 %}
81 %option reentrant
82 %option bison-bridge
83 %option 8bit
84 %option never-interactive
85 %option nodefault
86 %option noinput
87 %option nounput
88 %option noyywrap
89 %option warn
90 %option prefix="psql_yy"
92 /*
93 * All of the following definitions and rules should exactly match
94 * src/backend/parser/scan.l so far as the flex patterns are concerned.
95 * The rule bodies are just ECHO as opposed to what the backend does,
96 * however. (But be sure to duplicate code that affects the lexing process,
97 * such as BEGIN() and yyless().) Also, psqlscan uses a single <<EOF>> rule
98 * whereas scan.l has a separate one for each exclusive state.
99 */
101 /*
102 * OK, here is a short description of lex/flex rules behavior.
103 * The longest pattern which matches an input string is always chosen.
104 * For equal-length patterns, the first occurring in the rules list is chosen.
105 * INITIAL is the starting state, to which all non-conditional rules apply.
106 * Exclusive states change parsing rules while the state is active. When in
107 * an exclusive state, only those rules defined for that state apply.
108 *
109 * We use exclusive states for quoted strings, extended comments,
110 * and to eliminate parsing troubles for numeric strings.
111 * Exclusive states:
112 * <xb> bit string literal
113 * <xc> extended C-style comments
114 * <xd> delimited identifiers (double-quoted identifiers)
115 * <xh> hexadecimal byte string
116 * <xq> standard quoted strings
117 * <xqs> quote stop (detect continued strings)
118 * <xe> extended quoted strings (support backslash escape sequences)
119 * <xdolq> $foo$ quoted strings
120 * <xui> quoted identifier with Unicode escapes
121 * <xus> quoted string with Unicode escapes
122 *
123 * Note: we intentionally don't mimic the backend's <xeu> state; we have
124 * no need to distinguish it from <xe> state, and no good way to get out
125 * of it in error cases. The backend just throws yyerror() in those
126 * cases, but that's not an option here.
127 */
129 %x xb
130 %x xc
131 %x xd
132 %x xh
133 %x xq
134 %x xqs
135 %x xe
136 %x xdolq
137 %x xui
138 %x xus
140 /*
141 * In order to make the world safe for Windows and Mac clients as well as
142 * Unix ones, we accept either \n or \r as a newline. A DOS-style \r\n
143 * sequence will be seen as two successive newlines, but that doesn't cause
144 * any problems. Comments that start with -- and extend to the next
145 * newline are treated as equivalent to a single whitespace character.
146 *
147 * NOTE a fine point: if there is no newline following --, we will absorb
148 * everything to the end of the input as a comment. This is correct. Older
149 * versions of Postgres failed to recognize -- as a comment if the input
150 * did not end with a newline.
151 *
152 * XXX perhaps \f (formfeed) should be treated as a newline as well?
153 *
154 * XXX if you change the set of whitespace characters, fix scanner_isspace()
155 * to agree.
156 */
158 space [ \t\n\r\f]
159 horiz_space [ \t\f]
160 newline [\n\r]
161 non_newline [^\n\r]
163 comment ("--"{non_newline}*)
165 whitespace ({space}+|{comment})
167 /*
168 * SQL requires at least one newline in the whitespace separating
169 * string literals that are to be concatenated. Silly, but who are we
170 * to argue? Note that {whitespace_with_newline} should not have * after
171 * it, whereas {whitespace} should generally have a * after it...
172 */
174 special_whitespace ({space}+|{comment}{newline})
175 horiz_whitespace ({horiz_space}|{comment})
176 whitespace_with_newline ({horiz_whitespace}*{newline}{special_whitespace}*)
178 quote '
179 /* If we see {quote} then {quotecontinue}, the quoted string continues */
180 quotecontinue {whitespace_with_newline}{quote}
182 /*
183 * {quotecontinuefail} is needed to avoid lexer backup when we fail to match
184 * {quotecontinue}. It might seem that this could just be {whitespace}*,
185 * but if there's a dash after {whitespace_with_newline}, it must be consumed
186 * to see if there's another dash --- which would start a {comment} and thus
187 * allow continuation of the {quotecontinue} token.
188 */
189 quotecontinuefail {whitespace}*"-"?
191 /* Bit string
192 * It is tempting to scan the string for only those characters
193 * which are allowed. However, this leads to silently swallowed
194 * characters if illegal characters are included in the string.
195 * For example, if xbinside is [01] then B'ABCD' is interpreted
196 * as a zero-length string, and the ABCD' is lost!
197 * Better to pass the string forward and let the input routines
198 * validate the contents.
199 */
200 xbstart [bB]{quote}
201 xbinside [^']*
203 /* Hexadecimal byte string */
204 xhstart [xX]{quote}
205 xhinside [^']*
207 /* National character */
208 xnstart [nN]{quote}
210 /* Quoted string that allows backslash escapes */
211 xestart [eE]{quote}
212 xeinside [^\\']+
213 xeescape [\\][^0-7]
214 xeoctesc [\\][0-7]{1,3}
215 xehexesc [\\]x[0-9A-Fa-f]{1,2}
216 xeunicode [\\](u[0-9A-Fa-f]{4}|U[0-9A-Fa-f]{8})
217 xeunicodefail [\\](u[0-9A-Fa-f]{0,3}|U[0-9A-Fa-f]{0,7})
219 /* Extended quote
220 * xqdouble implements embedded quote, ''''
221 */
222 xqstart {quote}
223 xqdouble {quote}{quote}
224 xqinside [^']+
226 /* $foo$ style quotes ("dollar quoting")
227 * The quoted string starts with $foo$ where "foo" is an optional string
228 * in the form of an identifier, except that it may not contain "$",
229 * and extends to the first occurrence of an identical string.
230 * There is *no* processing of the quoted text.
231 *
232 * {dolqfailed} is an error rule to avoid scanner backup when {dolqdelim}
233 * fails to match its trailing "$".
234 */
235 dolq_start [A-Za-z\200-\377_]
236 dolq_cont [A-Za-z\200-\377_0-9]
237 dolqdelim \$({dolq_start}{dolq_cont}*)?\$
238 dolqfailed \${dolq_start}{dolq_cont}*
239 dolqinside [^$]+
241 /* Double quote
242 * Allows embedded spaces and other special characters into identifiers.
243 */
244 dquote \"
245 xdstart {dquote}
246 xdstop {dquote}
247 xddouble {dquote}{dquote}
248 xdinside [^"]+
250 /* Quoted identifier with Unicode escapes */
251 xuistart [uU]&{dquote}
253 /* Quoted string with Unicode escapes */
254 xusstart [uU]&{quote}
256 /* error rule to avoid backup */
257 xufailed [uU]&
260 /* C-style comments
261 *
262 * The "extended comment" syntax closely resembles allowable operator syntax.
263 * The tricky part here is to get lex to recognize a string starting with
264 * slash-star as a comment, when interpreting it as an operator would produce
265 * a longer match --- remember lex will prefer a longer match! Also, if we
266 * have something like plus-slash-star, lex will think this is a 3-character
267 * operator whereas we want to see it as a + operator and a comment start.
268 * The solution is two-fold:
269 * 1. append {op_chars}* to xcstart so that it matches as much text as
270 * {operator} would. Then the tie-breaker (first matching rule of same
271 * length) ensures xcstart wins. We put back the extra stuff with yyless()
272 * in case it contains a star-slash that should terminate the comment.
273 * 2. In the operator rule, check for slash-star within the operator, and
274 * if found throw it back with yyless(). This handles the plus-slash-star
275 * problem.
276 * Dash-dash comments have similar interactions with the operator rule.
277 */
278 xcstart \/\*{op_chars}*
279 xcstop \*+\/
280 xcinside [^*/]+
282 ident_start [A-Za-z\200-\377_]
283 ident_cont [A-Za-z\200-\377_0-9\$]
285 identifier {ident_start}{ident_cont}*
287 /* Assorted special-case operators and operator-like tokens */
288 typecast "::"
289 dot_dot \.\.
290 colon_equals ":="
292 /*
293 * These operator-like tokens (unlike the above ones) also match the {operator}
294 * rule, which means that they might be overridden by a longer match if they
295 * are followed by a comment start or a + or - character. Accordingly, if you
296 * add to this list, you must also add corresponding code to the {operator}
297 * block to return the correct token in such cases. (This is not needed in
298 * psqlscan.l since the token value is ignored there.)
299 */
300 equals_greater "=>"
301 less_equals "<="
302 greater_equals ">="
303 less_greater "<>"
304 not_equals "!="
306 /*
307 * "self" is the set of chars that should be returned as single-character
308 * tokens. "op_chars" is the set of chars that can make up "Op" tokens,
309 * which can be one or more characters long (but if a single-char token
310 * appears in the "self" set, it is not to be returned as an Op). Note
311 * that the sets overlap, but each has some chars that are not in the other.
312 *
313 * If you change either set, adjust the character lists appearing in the
314 * rule for "operator"!
315 */
316 self [,()\[\].;\:\+\-\*\/\%\^\<\>\=]
317 op_chars [\~\!\@\#\^\&\|\`\?\+\-\*\/\%\<\>\=]
318 operator {op_chars}+
320 /*
321 * Numbers
322 *
323 * Unary minus is not part of a number here. Instead we pass it separately to
324 * the parser, and there it gets coerced via doNegate().
325 *
326 * {decimalfail} is used because we would like "1..10" to lex as 1, dot_dot, 10.
327 *
328 * {realfail1} and {realfail2} are added to prevent the need for scanner
329 * backup when the {real} rule fails to match completely.
330 */
331 digit [0-9]
333 integer {digit}+
334 decimal (({digit}*\.{digit}+)|({digit}+\.{digit}*))
335 decimalfail {digit}+\.\.
336 real ({integer}|{decimal})[Ee][-+]?{digit}+
337 realfail1 ({integer}|{decimal})[Ee]
338 realfail2 ({integer}|{decimal})[Ee][-+]
340 param \${integer}
342 /* psql-specific: characters allowed in variable names */
343 variable_char [A-Za-z\200-\377_0-9]
345 other .
347 /*
348 * Dollar quoted strings are totally opaque, and no escaping is done on them.
349 * Other quoted strings must allow some special characters such as single-quote
350 * and newline.
351 * Embedded single-quotes are implemented both in the SQL standard
352 * style of two adjacent single quotes "''" and in the Postgres/Java style
353 * of escaped-quote "\'".
354 * Other embedded escaped characters are matched explicitly and the leading
355 * backslash is dropped from the string.
356 * Note that xcstart must appear before operator, as explained above!
357 * Also whitespace (comment) must appear before operator.
358 */
360 %%
362 %{
363 /* Declare some local variables inside yylex(), for convenience */
364 PsqlScanState cur_state = yyextra;
365 PQExpBuffer output_buf = cur_state->output_buf;
367 /*
368 * Force flex into the state indicated by start_state. This has a
369 * couple of purposes: it lets some of the functions below set a new
370 * starting state without ugly direct access to flex variables, and it
371 * allows us to transition from one flex lexer to another so that we
372 * can lex different parts of the source string using separate lexers.
373 */
374 BEGIN(cur_state->start_state);
375 %}
377 {whitespace} {
378 /*
379 * Note that the whitespace rule includes both true
380 * whitespace and single-line ("--" style) comments.
381 * We suppress whitespace until we have collected some
382 * non-whitespace data. (This interacts with some
383 * decisions in MainLoop(); see there for details.)
384 */
385 if (output_buf->len > 0)
386 ECHO;
387 }
389 {xcstart} {
390 cur_state->xcdepth = 0;
391 BEGIN(xc);
392 /* Put back any characters past slash-star; see above */
393 yyless(2);
394 ECHO;
395 }
397 <xc>{
398 {xcstart} {
399 cur_state->xcdepth++;
400 /* Put back any characters past slash-star; see above */
401 yyless(2);
402 ECHO;
403 }
405 {xcstop} {
406 if (cur_state->xcdepth <= 0)
407 BEGIN(INITIAL);
408 else
409 cur_state->xcdepth--;
410 ECHO;
411 }
413 {xcinside} {
414 ECHO;
415 }
417 {op_chars} {
418 ECHO;
419 }
421 \*+ {
422 ECHO;
423 }
424 } /* <xc> */
426 {xbstart} {
427 BEGIN(xb);
428 ECHO;
429 }
430 <xh>{xhinside} |
431 <xb>{xbinside} {
432 ECHO;
433 }
435 {xhstart} {
436 /* Hexadecimal bit type.
437 * At some point we should simply pass the string
438 * forward to the parser and label it there.
439 * In the meantime, place a leading "x" on the string
440 * to mark it for the input routine as a hex string.
441 */
442 BEGIN(xh);
443 ECHO;
444 }
446 {xnstart} {
447 yyless(1); /* eat only 'n' this time */
448 ECHO;
449 }
451 {xqstart} {
452 if (cur_state->std_strings)
453 BEGIN(xq);
454 else
455 BEGIN(xe);
456 ECHO;
457 }
458 {xestart} {
459 BEGIN(xe);
460 ECHO;
461 }
462 {xusstart} {
463 BEGIN(xus);
464 ECHO;
465 }
467 <xb,xh,xq,xe,xus>{quote} {
468 /*
469 * When we are scanning a quoted string and see an end
470 * quote, we must look ahead for a possible continuation.
471 * If we don't see one, we know the end quote was in fact
472 * the end of the string. To reduce the lexer table size,
473 * we use a single "xqs" state to do the lookahead for all
474 * types of strings.
475 */
476 cur_state->state_before_str_stop = YYSTATE;
477 BEGIN(xqs);
478 ECHO;
479 }
480 <xqs>{quotecontinue} {
481 /*
482 * Found a quote continuation, so return to the in-quote
483 * state and continue scanning the literal. Nothing is
484 * added to the literal's contents.
485 */
486 BEGIN(cur_state->state_before_str_stop);
487 ECHO;
488 }
489 <xqs>{quotecontinuefail} |
490 <xqs>{other} {
491 /*
492 * Failed to see a quote continuation. Throw back
493 * everything after the end quote, and handle the string
494 * according to the state we were in previously.
495 */
496 yyless(0);
497 BEGIN(INITIAL);
498 /* There's nothing to echo ... */
499 }
501 <xq,xe,xus>{xqdouble} {
502 ECHO;
503 }
504 <xq,xus>{xqinside} {
505 ECHO;
506 }
507 <xe>{xeinside} {
508 ECHO;
509 }
510 <xe>{xeunicode} {
511 ECHO;
512 }
513 <xe>{xeunicodefail} {
514 ECHO;
515 }
516 <xe>{xeescape} {
517 ECHO;
518 }
519 <xe>{xeoctesc} {
520 ECHO;
521 }
522 <xe>{xehexesc} {
523 ECHO;
524 }
525 <xe>. {
526 /* This is only needed for \ just before EOF */
527 ECHO;
528 }
530 {dolqdelim} {
531 cur_state->dolqstart = pg_strdup(yytext);
532 BEGIN(xdolq);
533 ECHO;
534 }
535 {dolqfailed} {
536 /* throw back all but the initial "$" */
537 yyless(1);
538 ECHO;
539 }
540 <xdolq>{dolqdelim} {
541 if (strcmp(yytext, cur_state->dolqstart) == 0)
542 {
543 free(cur_state->dolqstart);
544 cur_state->dolqstart = NULL;
545 BEGIN(INITIAL);
546 }
547 else
548 {
549 /*
550 * When we fail to match $...$ to dolqstart, transfer
551 * the $... part to the output, but put back the final
552 * $ for rescanning. Consider $delim$...$junk$delim$
553 */
554 yyless(yyleng - 1);
555 }
556 ECHO;
557 }
558 <xdolq>{dolqinside} {
559 ECHO;
560 }
561 <xdolq>{dolqfailed} {
562 ECHO;
563 }
564 <xdolq>. {
565 /* This is only needed for $ inside the quoted text */
566 ECHO;
567 }
569 {xdstart} {
570 BEGIN(xd);
571 ECHO;
572 }
573 {xuistart} {
574 BEGIN(xui);
575 ECHO;
576 }
577 <xd>{xdstop} {
578 BEGIN(INITIAL);
579 ECHO;
580 }
581 <xui>{dquote} {
582 BEGIN(INITIAL);
583 ECHO;
584 }
585 <xd,xui>{xddouble} {
586 ECHO;
587 }
588 <xd,xui>{xdinside} {
589 ECHO;
590 }
592 {xufailed} {
593 /* throw back all but the initial u/U */
594 yyless(1);
595 ECHO;
596 }
598 {typecast} {
599 ECHO;
600 }
602 {dot_dot} {
603 ECHO;
604 }
606 {colon_equals} {
607 ECHO;
608 }
610 {equals_greater} {
611 ECHO;
612 }
614 {less_equals} {
615 ECHO;
616 }
618 {greater_equals} {
619 ECHO;
620 }
622 {less_greater} {
623 ECHO;
624 }
626 {not_equals} {
627 ECHO;
628 }
630 /*
631 * These rules are specific to psql --- they implement parenthesis
632 * counting and detection of command-ending semicolon. These must
633 * appear before the {self} rule so that they take precedence over it.
634 */
636 "(" {
637 cur_state->paren_depth++;
638 ECHO;
639 }
641 ")" {
642 if (cur_state->paren_depth > 0)
643 cur_state->paren_depth--;
644 ECHO;
645 }
647 ";" {
648 ECHO;
649 if (cur_state->paren_depth == 0 && cur_state->begin_depth == 0)
650 {
651 /* Terminate lexing temporarily */
652 cur_state->start_state = YY_START;
653 cur_state->identifier_count = 0;
654 return LEXRES_SEMI;
655 }
656 }
658 /*
659 * psql-specific rules to handle backslash commands and variable
660 * substitution. We want these before {self}, also.
661 */
663 "\\"[;:] {
664 /* Force a semi-colon or colon into the query buffer */
665 psqlscan_emit(cur_state, yytext + 1, 1);
666 if (yytext[1] == ';')
667 cur_state->identifier_count = 0;
668 }
670 "\\" {
671 /* Terminate lexing temporarily */
672 cur_state->start_state = YY_START;
673 return LEXRES_BACKSLASH;
674 }
676 :{variable_char}+ {
677 /* Possible psql variable substitution */
678 char *varname;
679 char *value;
681 varname = psqlscan_extract_substring(cur_state,
682 yytext + 1,
683 yyleng - 1);
684 if (cur_state->callbacks->get_variable)
685 value = cur_state->callbacks->get_variable(varname,
686 PQUOTE_PLAIN,
687 cur_state->cb_passthrough);
688 else
689 value = NULL;
691 if (value)
692 {
693 /* It is a variable, check for recursion */
694 if (psqlscan_var_is_current_source(cur_state, varname))
695 {
696 /* Recursive expansion --- don't go there */
697 pg_log_warning("skipping recursive expansion of variable \"%s\"",
698 varname);
699 /* Instead copy the string as is */
700 ECHO;
701 }
702 else
703 {
704 /* OK, perform substitution */
705 psqlscan_push_new_buffer(cur_state, value, varname);
706 /* yy_scan_string already made buffer active */
707 }
708 free(value);
709 }
710 else
711 {
712 /*
713 * if the variable doesn't exist we'll copy the string
714 * as is
715 */
716 ECHO;
717 }
719 free(varname);
720 }
722 :'{variable_char}+' {
723 psqlscan_escape_variable(cur_state, yytext, yyleng,
724 PQUOTE_SQL_LITERAL);
725 }
727 :\"{variable_char}+\" {
728 psqlscan_escape_variable(cur_state, yytext, yyleng,
729 PQUOTE_SQL_IDENT);
730 }
732 :\{\?{variable_char}+\} {
733 psqlscan_test_variable(cur_state, yytext, yyleng);
734 }
736 /*
737 * These rules just avoid the need for scanner backup if one of the
738 * three rules above fails to match completely.
739 */
741 :'{variable_char}* {
742 /* Throw back everything but the colon */
743 yyless(1);
744 ECHO;
745 }
747 :\"{variable_char}* {
748 /* Throw back everything but the colon */
749 yyless(1);
750 ECHO;
751 }
753 :\{\?{variable_char}* {
754 /* Throw back everything but the colon */
755 yyless(1);
756 ECHO;
757 }
758 :\{ {
759 /* Throw back everything but the colon */
760 yyless(1);
761 ECHO;
762 }
764 /*
765 * Back to backend-compatible rules.
766 */
768 {self} {
769 ECHO;
770 }
772 {operator} {
773 /*
774 * Check for embedded slash-star or dash-dash; those
775 * are comment starts, so operator must stop there.
776 * Note that slash-star or dash-dash at the first
777 * character will match a prior rule, not this one.
778 */
779 int nchars = yyleng;
780 char *slashstar = strstr(yytext, "/*");
781 char *dashdash = strstr(yytext, "--");
783 if (slashstar && dashdash)
784 {
785 /* if both appear, take the first one */
786 if (slashstar > dashdash)
787 slashstar = dashdash;
788 }
789 else if (!slashstar)
790 slashstar = dashdash;
791 if (slashstar)
792 nchars = slashstar - yytext;
794 /*
795 * For SQL compatibility, '+' and '-' cannot be the
796 * last char of a multi-char operator unless the operator
797 * contains chars that are not in SQL operators.
798 * The idea is to lex '=-' as two operators, but not
799 * to forbid operator names like '?-' that could not be
800 * sequences of SQL operators.
801 */
802 if (nchars > 1 &&
803 (yytext[nchars - 1] == '+' ||
804 yytext[nchars - 1] == '-'))
805 {
806 int ic;
808 for (ic = nchars - 2; ic >= 0; ic--)
809 {
810 char c = yytext[ic];
811 if (c == '~' || c == '!' || c == '@' ||
812 c == '#' || c == '^' || c == '&' ||
813 c == '|' || c == '`' || c == '?' ||
814 c == '%')
815 break;
816 }
817 if (ic < 0)
818 {
819 /*
820 * didn't find a qualifying character, so remove
821 * all trailing [+-]
822 */
823 do {
824 nchars--;
825 } while (nchars > 1 &&
826 (yytext[nchars - 1] == '+' ||
827 yytext[nchars - 1] == '-'));
828 }
829 }
831 if (nchars < yyleng)
832 {
833 /* Strip the unwanted chars from the token */
834 yyless(nchars);
835 }
836 ECHO;
837 }
839 {param} {
840 ECHO;
841 }
843 {integer} {
844 ECHO;
845 }
846 {decimal} {
847 ECHO;
848 }
849 {decimalfail} {
850 /* throw back the .., and treat as integer */
851 yyless(yyleng - 2);
852 ECHO;
853 }
854 {real} {
855 ECHO;
856 }
857 {realfail1} {
858 /*
859 * throw back the [Ee], and figure out whether what
860 * remains is an {integer} or {decimal}.
861 * (in psql, we don't actually care...)
862 */
863 yyless(yyleng - 1);
864 ECHO;
865 }
866 {realfail2} {
867 /* throw back the [Ee][+-], and proceed as above */
868 yyless(yyleng - 2);
869 ECHO;
870 }
873 {identifier} {
874 /*
875 * We need to track if we are inside a BEGIN .. END block
876 * in a function definition, so that semicolons contained
877 * therein don't terminate the whole statement. Short of
878 * writing a full parser here, the following heuristic
879 * should work. First, we track whether the beginning of
880 * the statement matches CREATE [OR REPLACE]
881 * {FUNCTION|PROCEDURE}
882 */
884 if (cur_state->identifier_count == 0)
885 memset(cur_state->identifiers, 0, sizeof(cur_state->identifiers));
887 if (pg_strcasecmp(yytext, "create") == 0 ||
888 pg_strcasecmp(yytext, "function") == 0 ||
889 pg_strcasecmp(yytext, "procedure") == 0 ||
890 pg_strcasecmp(yytext, "or") == 0 ||
891 pg_strcasecmp(yytext, "replace") == 0)
892 {
893 if (cur_state->identifier_count < sizeof(cur_state->identifiers))
894 cur_state->identifiers[cur_state->identifier_count] = pg_tolower((unsigned char) yytext[0]);
895 }
897 cur_state->identifier_count++;
899 if (cur_state->identifiers[0] == 'c' &&
900 (cur_state->identifiers[1] == 'f' || cur_state->identifiers[1] == 'p' ||
901 (cur_state->identifiers[1] == 'o' && cur_state->identifiers[2] == 'r' &&
902 (cur_state->identifiers[3] == 'f' || cur_state->identifiers[3] == 'p'))) &&
903 cur_state->paren_depth == 0)
904 {
905 if (pg_strcasecmp(yytext, "begin") == 0)
906 cur_state->begin_depth++;
907 else if (pg_strcasecmp(yytext, "case") == 0)
908 {
909 /*
910 * CASE also ends with END. We only need to track
911 * this if we are already inside a BEGIN.
912 */
913 if (cur_state->begin_depth >= 1)
914 cur_state->begin_depth++;
915 }
916 else if (pg_strcasecmp(yytext, "end") == 0)
917 {
918 if (cur_state->begin_depth > 0)
919 cur_state->begin_depth--;
920 }
921 }
923 ECHO;
924 }
926 {other} {
927 ECHO;
928 }
930 <<EOF>> {
931 if (cur_state->buffer_stack == NULL)
932 {
933 cur_state->start_state = YY_START;
934 return LEXRES_EOL; /* end of input reached */
935 }
937 /*
938 * We were expanding a variable, so pop the inclusion
939 * stack and keep lexing
940 */
941 psqlscan_pop_buffer_stack(cur_state);
942 psqlscan_select_top_buffer(cur_state);
943 }
945 %%
947 /* LCOV_EXCL_STOP */
949 /*
950 * Create a lexer working state struct.
951 *
952 * callbacks is a struct of function pointers that encapsulate some
953 * behavior we need from the surrounding program. This struct must
954 * remain valid for the lifespan of the PsqlScanState.
955 */
956 PsqlScanState
957 psql_scan_create(const PsqlScanCallbacks *callbacks)
958 {
959 PsqlScanState state;
961 state = (PsqlScanStateData *) pg_malloc0(sizeof(PsqlScanStateData));
963 state->callbacks = callbacks;
965 yylex_init(&state->scanner);
967 yyset_extra(state, state->scanner);
969 psql_scan_reset(state);
971 return state;
972 }
974 /*
975 * Destroy a lexer working state struct, releasing all resources.
976 */
977 void
978 psql_scan_destroy(PsqlScanState state)
979 {
980 psql_scan_finish(state);
982 psql_scan_reset(state);
984 yylex_destroy(state->scanner);
986 free(state);
987 }
989 /*
990 * Set the callback passthrough pointer for the lexer.
991 *
992 * This could have been integrated into psql_scan_create, but keeping it
993 * separate allows the application to change the pointer later, which might
994 * be useful.
995 */
996 void
997 psql_scan_set_passthrough(PsqlScanState state, void *passthrough)
998 {
999 state->cb_passthrough = passthrough;
1000 }
1002 /*
1003 * Set up to perform lexing of the given input line.
1004 *
1005 * The text at *line, extending for line_len bytes, will be scanned by
1006 * subsequent calls to the psql_scan routines. psql_scan_finish should
1007 * be called when scanning is complete. Note that the lexer retains
1008 * a pointer to the storage at *line --- this string must not be altered
1009 * or freed until after psql_scan_finish is called.
1010 *
1011 * encoding is the libpq identifier for the character encoding in use,
1012 * and std_strings says whether standard_conforming_strings is on.
1013 */
1014 void
1015 psql_scan_setup(PsqlScanState state,
1016 const char *line, int line_len,
1017 int encoding, bool std_strings)
1018 {
1019 /* Mustn't be scanning already */
1020 Assert(state->scanbufhandle == NULL);
1021 Assert(state->buffer_stack == NULL);
1023 /* Do we need to hack the character set encoding? */
1024 state->encoding = encoding;
1025 state->safe_encoding = pg_valid_server_encoding_id(encoding);
1027 /* Save standard-strings flag as well */
1028 state->std_strings = std_strings;
1030 /* Set up flex input buffer with appropriate translation and padding */
1031 state->scanbufhandle = psqlscan_prepare_buffer(state, line, line_len,
1032 &state->scanbuf);
1033 state->scanline = line;
1035 /* Set lookaside data in case we have to map unsafe encoding */
1036 state->curline = state->scanbuf;
1037 state->refline = state->scanline;
1038 }
1040 /*
1041 * Do lexical analysis of SQL command text.
1042 *
1043 * The text previously passed to psql_scan_setup is scanned, and appended
1044 * (possibly with transformation) to query_buf.
1045 *
1046 * The return value indicates the condition that stopped scanning:
1047 *
1048 * PSCAN_SEMICOLON: found a command-ending semicolon. (The semicolon is
1049 * transferred to query_buf.) The command accumulated in query_buf should
1050 * be executed, then clear query_buf and call again to scan the remainder
1051 * of the line.
1052 *
1053 * PSCAN_BACKSLASH: found a backslash that starts a special command.
1054 * Any previous data on the line has been transferred to query_buf.
1055 * The caller will typically next apply a separate flex lexer to scan
1056 * the special command.
1057 *
1058 * PSCAN_INCOMPLETE: the end of the line was reached, but we have an
1059 * incomplete SQL command. *prompt is set to the appropriate prompt type.
1060 *
1061 * PSCAN_EOL: the end of the line was reached, and there is no lexical
1062 * reason to consider the command incomplete. The caller may or may not
1063 * choose to send it. *prompt is set to the appropriate prompt type if
1064 * the caller chooses to collect more input.
1065 *
1066 * In the PSCAN_INCOMPLETE and PSCAN_EOL cases, psql_scan_finish() should
1067 * be called next, then the cycle may be repeated with a fresh input line.
1068 *
1069 * In all cases, *prompt is set to an appropriate prompt type code for the
1070 * next line-input operation.
1071 */
1072 PsqlScanResult
1073 psql_scan(PsqlScanState state,
1074 PQExpBuffer query_buf,
1075 promptStatus_t *prompt)
1076 {
1077 PsqlScanResult result;
1078 int lexresult;
1080 /* Must be scanning already */
1081 Assert(state->scanbufhandle != NULL);
1083 /* Set current output target */
1084 state->output_buf = query_buf;
1086 /* Set input source */
1087 if (state->buffer_stack != NULL)
1088 yy_switch_to_buffer(state->buffer_stack->buf, state->scanner);
1089 else
1090 yy_switch_to_buffer(state->scanbufhandle, state->scanner);
1092 /* And lex. */
1093 lexresult = yylex(NULL, state->scanner);
1095 /*
1096 * Check termination state and return appropriate result info.
1097 */
1098 switch (lexresult)
1099 {
1100 case LEXRES_EOL: /* end of input */
1101 switch (state->start_state)
1102 {
1103 case INITIAL:
1104 case xqs: /* we treat this like INITIAL */
1105 if (state->paren_depth > 0)
1106 {
1107 result = PSCAN_INCOMPLETE;
1108 *prompt = PROMPT_PAREN;
1109 }
1110 else if (state->begin_depth > 0)
1111 {
1112 result = PSCAN_INCOMPLETE;
1113 *prompt = PROMPT_CONTINUE;
1114 }
1115 else if (query_buf->len > 0)
1116 {
1117 result = PSCAN_EOL;
1118 *prompt = PROMPT_CONTINUE;
1119 }
1120 else
1121 {
1122 /* never bother to send an empty buffer */
1123 result = PSCAN_INCOMPLETE;
1124 *prompt = PROMPT_READY;
1125 }
1126 break;
1127 case xb:
1128 result = PSCAN_INCOMPLETE;
1129 *prompt = PROMPT_SINGLEQUOTE;
1130 break;
1131 case xc:
1132 result = PSCAN_INCOMPLETE;
1133 *prompt = PROMPT_COMMENT;
1134 break;
1135 case xd:
1136 result = PSCAN_INCOMPLETE;
1137 *prompt = PROMPT_DOUBLEQUOTE;
1138 break;
1139 case xh:
1140 result = PSCAN_INCOMPLETE;
1141 *prompt = PROMPT_SINGLEQUOTE;
1142 break;
1143 case xe:
1144 result = PSCAN_INCOMPLETE;
1145 *prompt = PROMPT_SINGLEQUOTE;
1146 break;
1147 case xq:
1148 result = PSCAN_INCOMPLETE;
1149 *prompt = PROMPT_SINGLEQUOTE;
1150 break;
1151 case xdolq:
1152 result = PSCAN_INCOMPLETE;
1153 *prompt = PROMPT_DOLLARQUOTE;
1154 break;
1155 case xui:
1156 result = PSCAN_INCOMPLETE;
1157 *prompt = PROMPT_DOUBLEQUOTE;
1158 break;
1159 case xus:
1160 result = PSCAN_INCOMPLETE;
1161 *prompt = PROMPT_SINGLEQUOTE;
1162 break;
1163 default:
1164 /* can't get here */
1165 fprintf(stderr, "invalid YY_START\n");
1166 exit(1);
1167 }
1168 break;
1169 case LEXRES_SEMI: /* semicolon */
1170 result = PSCAN_SEMICOLON;
1171 *prompt = PROMPT_READY;
1172 break;
1173 case LEXRES_BACKSLASH: /* backslash */
1174 result = PSCAN_BACKSLASH;
1175 *prompt = PROMPT_READY;
1176 break;
1177 default:
1178 /* can't get here */
1179 fprintf(stderr, "invalid yylex result\n");
1180 exit(1);
1181 }
1183 return result;
1184 }
1186 /*
1187 * Clean up after scanning a string. This flushes any unread input and
1188 * releases resources (but not the PsqlScanState itself). Note however
1189 * that this does not reset the lexer scan state; that can be done by
1190 * psql_scan_reset(), which is an orthogonal operation.
1191 *
1192 * It is legal to call this when not scanning anything (makes it easier
1193 * to deal with error recovery).
1194 */
1195 void
1196 psql_scan_finish(PsqlScanState state)
1197 {
1198 /* Drop any incomplete variable expansions. */
1199 while (state->buffer_stack != NULL)
1200 psqlscan_pop_buffer_stack(state);
1202 /* Done with the outer scan buffer, too */
1203 if (state->scanbufhandle)
1204 yy_delete_buffer(state->scanbufhandle, state->scanner);
1205 state->scanbufhandle = NULL;
1206 if (state->scanbuf)
1207 free(state->scanbuf);
1208 state->scanbuf = NULL;
1209 }
1211 /*
1212 * Reset lexer scanning state to start conditions. This is appropriate
1213 * for executing \r psql commands (or any other time that we discard the
1214 * prior contents of query_buf). It is not, however, necessary to do this
1215 * when we execute and clear the buffer after getting a PSCAN_SEMICOLON or
1216 * PSCAN_EOL scan result, because the scan state must be INITIAL when those
1217 * conditions are returned.
1218 *
1219 * Note that this is unrelated to flushing unread input; that task is
1220 * done by psql_scan_finish().
1221 */
1222 void
1223 psql_scan_reset(PsqlScanState state)
1224 {
1225 state->start_state = INITIAL;
1226 state->paren_depth = 0;
1227 state->xcdepth = 0; /* not really necessary */
1228 if (state->dolqstart)
1229 free(state->dolqstart);
1230 state->dolqstart = NULL;
1231 state->identifier_count = 0;
1232 state->begin_depth = 0;
1233 }
1235 /*
1236 * Reselect this lexer (psqlscan.l) after using another one.
1237 *
1238 * Currently and for foreseeable uses, it's sufficient to reset to INITIAL
1239 * state, because we'd never switch to another lexer in a different state.
1240 * However, we don't want to reset e.g. paren_depth, so this can't be
1241 * the same as psql_scan_reset().
1242 *
1243 * Note: psql setjmp error recovery just calls psql_scan_reset(), so that
1244 * must be a superset of this.
1245 *
1246 * Note: it seems likely that other lexers could just assign INITIAL for
1247 * themselves, since that probably has the value zero in every flex-generated
1248 * lexer. But let's not assume that.
1249 */
1250 void
1251 psql_scan_reselect_sql_lexer(PsqlScanState state)
1252 {
1253 state->start_state = INITIAL;
1254 }
1256 /*
1257 * Return true if lexer is currently in an "inside quotes" state.
1258 *
1259 * This is pretty grotty but is needed to preserve the old behavior
1260 * that mainloop.c drops blank lines not inside quotes without even
1261 * echoing them.
1262 */
1263 bool
1264 psql_scan_in_quote(PsqlScanState state)
1265 {
1266 return state->start_state != INITIAL &&
1267 state->start_state != xqs;
1268 }
1270 /*
1271 * Push the given string onto the stack of stuff to scan.
1272 *
1273 * NOTE SIDE EFFECT: the new buffer is made the active flex input buffer.
1274 */
1275 void
1276 psqlscan_push_new_buffer(PsqlScanState state, const char *newstr,
1277 const char *varname)
1278 {
1279 StackElem *stackelem;
1281 stackelem = (StackElem *) pg_malloc(sizeof(StackElem));
1283 /*
1284 * In current usage, the passed varname points at the current flex input
1285 * buffer; we must copy it before calling psqlscan_prepare_buffer()
1286 * because that will change the buffer state.
1287 */
1288 stackelem->varname = varname ? pg_strdup(varname) : NULL;
1290 stackelem->buf = psqlscan_prepare_buffer(state, newstr, strlen(newstr),
1291 &stackelem->bufstring);
1292 state->curline = stackelem->bufstring;
1293 if (state->safe_encoding)
1294 {
1295 stackelem->origstring = NULL;
1296 state->refline = stackelem->bufstring;
1297 }
1298 else
1299 {
1300 stackelem->origstring = pg_strdup(newstr);
1301 state->refline = stackelem->origstring;
1302 }
1303 stackelem->next = state->buffer_stack;
1304 state->buffer_stack = stackelem;
1305 }
1307 /*
1308 * Pop the topmost buffer stack item (there must be one!)
1309 *
1310 * NB: after this, the flex input state is unspecified; caller must
1311 * switch to an appropriate buffer to continue lexing.
1312 * See psqlscan_select_top_buffer().
1313 */
1314 void
1315 psqlscan_pop_buffer_stack(PsqlScanState state)
1316 {
1317 StackElem *stackelem = state->buffer_stack;
1319 state->buffer_stack = stackelem->next;
1320 yy_delete_buffer(stackelem->buf, state->scanner);
1321 free(stackelem->bufstring);
1322 if (stackelem->origstring)
1323 free(stackelem->origstring);
1324 if (stackelem->varname)
1325 free(stackelem->varname);
1326 free(stackelem);
1327 }
1329 /*
1330 * Select the topmost surviving buffer as the active input.
1331 */
1332 void
1333 psqlscan_select_top_buffer(PsqlScanState state)
1334 {
1335 StackElem *stackelem = state->buffer_stack;
1337 if (stackelem != NULL)
1338 {
1339 yy_switch_to_buffer(stackelem->buf, state->scanner);
1340 state->curline = stackelem->bufstring;
1341 state->refline = stackelem->origstring ? stackelem->origstring : stackelem->bufstring;
1342 }
1343 else
1344 {
1345 yy_switch_to_buffer(state->scanbufhandle, state->scanner);
1346 state->curline = state->scanbuf;
1347 state->refline = state->scanline;
1348 }
1349 }
1351 /*
1352 * Check if specified variable name is the source for any string
1353 * currently being scanned
1354 */
1355 bool
1356 psqlscan_var_is_current_source(PsqlScanState state, const char *varname)
1357 {
1358 StackElem *stackelem;
1360 for (stackelem = state->buffer_stack;
1361 stackelem != NULL;
1362 stackelem = stackelem->next)
1363 {
1364 if (stackelem->varname && strcmp(stackelem->varname, varname) == 0)
1365 return true;
1366 }
1367 return false;
1368 }
1370 /*
1371 * Set up a flex input buffer to scan the given data. We always make a
1372 * copy of the data. If working in an unsafe encoding, the copy has
1373 * multibyte sequences replaced by FFs to avoid fooling the lexer rules.
1374 *
1375 * NOTE SIDE EFFECT: the new buffer is made the active flex input buffer.
1376 */
1377 YY_BUFFER_STATE
1378 psqlscan_prepare_buffer(PsqlScanState state, const char *txt, int len,
1379 char **txtcopy)
1380 {
1381 char *newtxt;
1383 /* Flex wants two \0 characters after the actual data */
1384 newtxt = pg_malloc(len + 2);
1385 *txtcopy = newtxt;
1386 newtxt[len] = newtxt[len + 1] = YY_END_OF_BUFFER_CHAR;
1388 if (state->safe_encoding)
1389 memcpy(newtxt, txt, len);
1390 else
1391 {
1392 /* Gotta do it the hard way */
1393 int i = 0;
1395 while (i < len)
1396 {
1397 int thislen = PQmblen(txt + i, state->encoding);
1399 /* first byte should always be okay... */
1400 newtxt[i] = txt[i];
1401 i++;
1402 while (--thislen > 0 && i < len)
1403 newtxt[i++] = (char) 0xFF;
1404 }
1405 }
1407 return yy_scan_buffer(newtxt, len + 2, state->scanner);
1408 }
1410 /*
1411 * psqlscan_emit() --- body for ECHO macro
1412 *
1413 * NB: this must be used for ALL and ONLY the text copied from the flex
1414 * input data. If you pass it something that is not part of the yytext
1415 * string, you are making a mistake. Internally generated text can be
1416 * appended directly to state->output_buf.
1417 */
1418 void
1419 psqlscan_emit(PsqlScanState state, const char *txt, int len)
1420 {
1421 PQExpBuffer output_buf = state->output_buf;
1423 if (state->safe_encoding)
1424 appendBinaryPQExpBuffer(output_buf, txt, len);
1425 else
1426 {
1427 /* Gotta do it the hard way */
1428 const char *reference = state->refline;
1429 int i;
1431 reference += (txt - state->curline);
1433 for (i = 0; i < len; i++)
1434 {
1435 char ch = txt[i];
1437 if (ch == (char) 0xFF)
1438 ch = reference[i];
1439 appendPQExpBufferChar(output_buf, ch);
1440 }
1441 }
1442 }
1444 /*
1445 * psqlscan_extract_substring --- fetch value of (part of) the current token
1446 *
1447 * This is like psqlscan_emit(), except that the data is returned as a
1448 * malloc'd string rather than being pushed directly to state->output_buf.
1449 */
1450 char *
1451 psqlscan_extract_substring(PsqlScanState state, const char *txt, int len)
1452 {
1453 char *result = (char *) pg_malloc(len + 1);
1455 if (state->safe_encoding)
1456 memcpy(result, txt, len);
1457 else
1458 {
1459 /* Gotta do it the hard way */
1460 const char *reference = state->refline;
1461 int i;
1463 reference += (txt - state->curline);
1465 for (i = 0; i < len; i++)
1466 {
1467 char ch = txt[i];
1469 if (ch == (char) 0xFF)
1470 ch = reference[i];
1471 result[i] = ch;
1472 }
1473 }
1474 result[len] = '\0';
1475 return result;
1476 }
1478 /*
1479 * psqlscan_escape_variable --- process :'VARIABLE' or :"VARIABLE"
1480 *
1481 * If the variable name is found, escape its value using the appropriate
1482 * quoting method and emit the value to output_buf. (Since the result is
1483 * surely quoted, there is never any reason to rescan it.) If we don't
1484 * find the variable or escaping fails, emit the token as-is.
1485 */
1486 void
1487 psqlscan_escape_variable(PsqlScanState state, const char *txt, int len,
1488 PsqlScanQuoteType quote)
1489 {
1490 char *varname;
1491 char *value;
1493 /* Variable lookup. */
1494 varname = psqlscan_extract_substring(state, txt + 2, len - 3);
1495 if (state->callbacks->get_variable)
1496 value = state->callbacks->get_variable(varname, quote,
1497 state->cb_passthrough);
1498 else
1499 value = NULL;
1500 free(varname);
1502 if (value)
1503 {
1504 /* Emit the suitably-escaped value */
1505 appendPQExpBufferStr(state->output_buf, value);
1506 free(value);
1507 }
1508 else
1509 {
1510 /* Emit original token as-is */
1511 psqlscan_emit(state, txt, len);
1512 }
1513 }
1515 void
1516 psqlscan_test_variable(PsqlScanState state, const char *txt, int len)
1517 {
1518 char *varname;
1519 char *value;
1521 varname = psqlscan_extract_substring(state, txt + 3, len - 4);
1522 if (state->callbacks->get_variable)
1523 value = state->callbacks->get_variable(varname, PQUOTE_PLAIN,
1524 state->cb_passthrough);
1525 else
1526 value = NULL;
1527 free(varname);
1529 if (value != NULL)
1530 {
1531 psqlscan_emit(state, "TRUE", 4);
1532 free(value);
1533 }
1534 else
1535 {
1536 psqlscan_emit(state, "FALSE", 5);
1537 }
1538 }