1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2009, 2010, 2011, 2013, 2016 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 #include "language/lexer/lexer.h"
33 #include "language/command.h"
34 #include "language/lexer/macro.h"
35 #include "language/lexer/scan.h"
36 #include "language/lexer/segment.h"
37 #include "language/lexer/token.h"
38 #include "libpspp/assertion.h"
39 #include "libpspp/cast.h"
40 #include "libpspp/deque.h"
41 #include "libpspp/i18n.h"
42 #include "libpspp/ll.h"
43 #include "libpspp/message.h"
44 #include "libpspp/misc.h"
45 #include "libpspp/str.h"
46 #include "libpspp/u8-istream.h"
47 #include "output/journal.h"
48 #include "output/output-item.h"
50 #include "gl/c-ctype.h"
51 #include "gl/minmax.h"
52 #include "gl/xalloc.h"
53 #include "gl/xmemdup0.h"
56 #define _(msgid) gettext (msgid)
57 #define N_(msgid) msgid
59 /* A token within a lex_source. */
62 /* The regular token information. */
65 /* For a token obtained through the lexer in an ordinary way, this is the
66 location of the token in terms of the lex_source's buffer.
68 For a token produced through macro expansion, this is the entire macro
71 src->tail <= line_pos <= token_pos <= src->head. */
72 size_t token_pos; /* Start of token. */
73 size_t token_len; /* Length of source for token in bytes. */
74 size_t line_pos; /* Start of line containing token_pos. */
75 int first_line; /* Line number at token_pos. */
77 /* For a token obtained through macro expansion, this is just this token.
79 For a token obtained through the lexer in an ordinary way, these are
81 char *macro_rep; /* The whole macro expansion. */
82 size_t ofs; /* Offset of this token in macro_rep. */
83 size_t len; /* Length of this token in macro_rep. */
84 size_t *ref_cnt; /* Number of lex_tokens that refer to macro_rep. */
88 lex_token_destroy (struct lex_token *t)
90 token_uninit (&t->token);
93 assert (*t->ref_cnt > 0);
103 /* A deque of lex_tokens that comprises one stage in the token pipeline in a
108 struct lex_token **tokens;
111 static void lex_stage_clear (struct lex_stage *);
112 static void lex_stage_uninit (struct lex_stage *);
114 static size_t lex_stage_count (const struct lex_stage *);
115 static bool lex_stage_is_empty (const struct lex_stage *);
117 static struct lex_token *lex_stage_last (struct lex_stage *);
118 static struct lex_token *lex_stage_first (struct lex_stage *);
119 static struct lex_token *lex_stage_nth (struct lex_stage *, size_t ofs);
121 static void lex_stage_push_last (struct lex_stage *, struct lex_token *);
122 static void lex_stage_pop_first (struct lex_stage *);
124 static void lex_stage_shift (struct lex_stage *dst, struct lex_stage *src,
127 /* Deletes all the tokens from STAGE. */
129 lex_stage_clear (struct lex_stage *stage)
131 while (!deque_is_empty (&stage->deque))
132 lex_stage_pop_first (stage);
135 /* Deletes all the tokens from STAGE and frees storage for the deque. */
137 lex_stage_uninit (struct lex_stage *stage)
139 lex_stage_clear (stage);
140 free (stage->tokens);
143 /* Returns true if STAGE contains no tokens, otherwise false. */
145 lex_stage_is_empty (const struct lex_stage *stage)
147 return deque_is_empty (&stage->deque);
150 /* Returns the number of tokens in STAGE. */
152 lex_stage_count (const struct lex_stage *stage)
154 return deque_count (&stage->deque);
157 /* Returns the last token in STAGE, which must be nonempty. The last token is
158 the one accessed with the greatest lookahead. */
159 static struct lex_token *
160 lex_stage_last (struct lex_stage *stage)
162 return stage->tokens[deque_front (&stage->deque, 0)];
165 /* Returns the first token in STAGE, which must be nonempty.
166 The first token is the one accessed with the least lookahead. */
167 static struct lex_token *
168 lex_stage_first (struct lex_stage *stage)
170 return lex_stage_nth (stage, 0);
173 /* Returns the token the given INDEX in STAGE. The first token (with the least
174 lookahead) is 0, the second token is 1, and so on. There must be at least
175 INDEX + 1 tokens in STAGE. */
176 static struct lex_token *
177 lex_stage_nth (struct lex_stage *stage, size_t index)
179 return stage->tokens[deque_back (&stage->deque, index)];
182 /* Adds TOKEN so that it becomes the last token in STAGE. */
184 lex_stage_push_last (struct lex_stage *stage, struct lex_token *token)
186 if (deque_is_full (&stage->deque))
187 stage->tokens = deque_expand (&stage->deque, stage->tokens,
188 sizeof *stage->tokens);
189 stage->tokens[deque_push_front (&stage->deque)] = token;
192 /* Removes the first token from STAGE and uninitializes it. */
194 lex_stage_pop_first (struct lex_stage *stage)
196 lex_token_destroy (stage->tokens[deque_pop_back (&stage->deque)]);
199 /* Removes the first N tokens from SRC, appending them to DST as the last
202 lex_stage_shift (struct lex_stage *dst, struct lex_stage *src, size_t n)
204 for (size_t i = 0; i < n; i++)
206 lex_stage_push_last (dst, lex_stage_first (src));
207 deque_pop_back (&src->deque);
211 /* A source of tokens, corresponding to a syntax file.
213 This is conceptually a lex_reader wrapped with everything needed to convert
214 its UTF-8 bytes into tokens. */
217 struct ll ll; /* In lexer's list of sources. */
218 struct lex_reader *reader;
220 struct segmenter segmenter;
221 bool eof; /* True if T_STOP was read from 'reader'. */
223 /* Buffer of UTF-8 bytes. */
225 size_t allocated; /* Number of bytes allocated. */
226 size_t tail; /* &buffer[0] offset into UTF-8 source. */
227 size_t head; /* &buffer[head - tail] offset into source. */
229 /* Positions in source file, tail <= pos <= head for each member here. */
230 size_t journal_pos; /* First byte not yet output to journal. */
231 size_t seg_pos; /* First byte not yet scanned as token. */
232 size_t line_pos; /* First byte of line containing seg_pos. */
234 int n_newlines; /* Number of new-lines up to seg_pos. */
235 bool suppress_next_newline;
239 This is a pipeline with the following stages. Each token eventually
240 made available to the parser passes through of these stages. The stages
241 are named after the processing that happens in each one.
243 Initially, tokens come from the segmenter and scanner to 'pp':
245 - pp: Tokens that need to pass through the macro preprocessor to end up
248 - merge: Tokens that need to pass through scan_merge() to end up in
251 - lookahead: Tokens available to the client for parsing. */
253 struct lex_stage merge;
254 struct lex_stage lookahead;
257 static struct lex_source *lex_source_create (struct lexer *,
258 struct lex_reader *);
259 static void lex_source_destroy (struct lex_source *);
264 struct ll_list sources; /* Contains "struct lex_source"s. */
265 struct macro_set *macros;
268 static struct lex_source *lex_source__ (const struct lexer *);
269 static char *lex_source_get_syntax__ (const struct lex_source *,
271 static const struct lex_token *lex_next__ (const struct lexer *, int n);
272 static void lex_source_push_endcmd__ (struct lex_source *);
274 static bool lex_source_get_lookahead (struct lex_source *);
275 static void lex_source_error_valist (struct lex_source *, int n0, int n1,
276 const char *format, va_list)
277 PRINTF_FORMAT (4, 0);
278 static const struct lex_token *lex_source_next__ (const struct lex_source *,
281 /* Initializes READER with the specified CLASS and otherwise some reasonable
282 defaults. The caller should fill in the others members as desired. */
284 lex_reader_init (struct lex_reader *reader,
285 const struct lex_reader_class *class)
287 reader->class = class;
288 reader->syntax = SEG_MODE_AUTO;
289 reader->error = LEX_ERROR_CONTINUE;
290 reader->file_name = NULL;
291 reader->encoding = NULL;
292 reader->line_number = 0;
296 /* Frees any file name already in READER and replaces it by a copy of
297 FILE_NAME, or if FILE_NAME is null then clears any existing name. */
299 lex_reader_set_file_name (struct lex_reader *reader, const char *file_name)
301 free (reader->file_name);
302 reader->file_name = xstrdup_if_nonnull (file_name);
305 /* Creates and returns a new lexer. */
309 struct lexer *lexer = xmalloc (sizeof *lexer);
310 *lexer = (struct lexer) {
311 .sources = LL_INITIALIZER (lexer->sources),
312 .macros = macro_set_create (),
317 /* Destroys LEXER. */
319 lex_destroy (struct lexer *lexer)
323 struct lex_source *source, *next;
325 ll_for_each_safe (source, next, struct lex_source, ll, &lexer->sources)
326 lex_source_destroy (source);
327 macro_set_destroy (lexer->macros);
332 /* Adds M to LEXER's set of macros. M replaces any existing macro with the
333 same name. Takes ownership of M. */
335 lex_define_macro (struct lexer *lexer, struct macro *m)
337 macro_set_add (lexer->macros, m);
340 /* Inserts READER into LEXER so that the next token read by LEXER comes from
341 READER. Before the caller, LEXER must either be empty or at a T_ENDCMD
344 lex_include (struct lexer *lexer, struct lex_reader *reader)
346 assert (ll_is_empty (&lexer->sources) || lex_token (lexer) == T_ENDCMD);
347 ll_push_head (&lexer->sources, &lex_source_create (lexer, reader)->ll);
350 /* Appends READER to LEXER, so that it will be read after all other current
351 readers have already been read. */
353 lex_append (struct lexer *lexer, struct lex_reader *reader)
355 ll_push_tail (&lexer->sources, &lex_source_create (lexer, reader)->ll);
360 /* Advances LEXER to the next token, consuming the current token. */
362 lex_get (struct lexer *lexer)
364 struct lex_source *src;
366 src = lex_source__ (lexer);
370 if (!lex_stage_is_empty (&src->lookahead))
371 lex_stage_pop_first (&src->lookahead);
373 while (lex_stage_is_empty (&src->lookahead))
374 if (!lex_source_get_lookahead (src))
376 lex_source_destroy (src);
377 src = lex_source__ (lexer);
383 /* Issuing errors. */
385 /* Prints a syntax error message containing the current token and
386 given message MESSAGE (if non-null). */
388 lex_error (struct lexer *lexer, const char *format, ...)
392 va_start (args, format);
393 lex_next_error_valist (lexer, 0, 0, format, args);
397 /* Prints a syntax error message containing the current token and
398 given message MESSAGE (if non-null). */
400 lex_error_valist (struct lexer *lexer, const char *format, va_list args)
402 lex_next_error_valist (lexer, 0, 0, format, args);
405 /* Prints a syntax error message containing the current token and
406 given message MESSAGE (if non-null). */
408 lex_next_error (struct lexer *lexer, int n0, int n1, const char *format, ...)
412 va_start (args, format);
413 lex_next_error_valist (lexer, n0, n1, format, args);
417 /* Prints a syntax error message saying that one of the strings provided as
418 varargs, up to the first NULL, is expected. */
420 (lex_error_expecting) (struct lexer *lexer, ...)
424 va_start (args, lexer);
425 lex_error_expecting_valist (lexer, args);
429 /* Prints a syntax error message saying that one of the options provided in
430 ARGS, up to the first NULL, is expected. */
432 lex_error_expecting_valist (struct lexer *lexer, va_list args)
434 enum { MAX_OPTIONS = 9 };
435 const char *options[MAX_OPTIONS];
437 while (n < MAX_OPTIONS)
439 const char *option = va_arg (args, const char *);
443 options[n++] = option;
445 lex_error_expecting_array (lexer, options, n);
449 lex_error_expecting_array (struct lexer *lexer, const char **options, size_t n)
454 lex_error (lexer, NULL);
458 lex_error (lexer, _("expecting %s"), options[0]);
462 lex_error (lexer, _("expecting %s or %s"), options[0], options[1]);
466 lex_error (lexer, _("expecting %s, %s, or %s"), options[0], options[1],
471 lex_error (lexer, _("expecting %s, %s, %s, or %s"),
472 options[0], options[1], options[2], options[3]);
476 lex_error (lexer, _("expecting %s, %s, %s, %s, or %s"),
477 options[0], options[1], options[2], options[3], options[4]);
481 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, or %s"),
482 options[0], options[1], options[2], options[3], options[4],
487 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, %s, or %s"),
488 options[0], options[1], options[2], options[3], options[4],
489 options[5], options[6]);
493 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, %s, %s, or %s"),
494 options[0], options[1], options[2], options[3], options[4],
495 options[5], options[6], options[7]);
499 lex_error (lexer, NULL);
503 /* Reports an error to the effect that subcommand SBC may only be specified
506 This function does not take a lexer as an argument or use lex_error(),
507 because the result would ordinarily just be redundant: "Syntax error at
508 SUBCOMMAND: Subcommand SUBCOMMAND may only be specified once.", which does
509 not help the user find the error. */
511 lex_sbc_only_once (const char *sbc)
513 msg (SE, _("Subcommand %s may only be specified once."), sbc);
516 /* Reports an error to the effect that subcommand SBC is missing.
518 This function does not take a lexer as an argument or use lex_error(),
519 because a missing subcommand can normally be detected only after the whole
520 command has been parsed, and so lex_error() would always report "Syntax
521 error at end of command", which does not help the user find the error. */
523 lex_sbc_missing (const char *sbc)
525 msg (SE, _("Required subcommand %s was not specified."), sbc);
528 /* Reports an error to the effect that specification SPEC may only be specified
529 once within subcommand SBC. */
531 lex_spec_only_once (struct lexer *lexer, const char *sbc, const char *spec)
533 lex_error (lexer, _("%s may only be specified once within subcommand %s"),
537 /* Reports an error to the effect that specification SPEC is missing within
540 lex_spec_missing (struct lexer *lexer, const char *sbc, const char *spec)
542 lex_error (lexer, _("Required %s specification missing from %s subcommand"),
546 /* Prints a syntax error message containing the current token and
547 given message MESSAGE (if non-null). */
549 lex_next_error_valist (struct lexer *lexer, int n0, int n1,
550 const char *format, va_list args)
552 struct lex_source *src = lex_source__ (lexer);
555 lex_source_error_valist (src, n0, n1, format, args);
561 ds_put_format (&s, _("Syntax error at end of input"));
564 ds_put_cstr (&s, ": ");
565 ds_put_vformat (&s, format, args);
567 ds_put_byte (&s, '.');
568 msg (SE, "%s", ds_cstr (&s));
573 /* Checks that we're at end of command.
574 If so, returns a successful command completion code.
575 If not, flags a syntax error and returns an error command
578 lex_end_of_command (struct lexer *lexer)
580 if (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_STOP)
582 lex_error (lexer, _("expecting end of command"));
589 /* Token testing functions. */
591 /* Returns true if the current token is a number. */
593 lex_is_number (const struct lexer *lexer)
595 return lex_next_is_number (lexer, 0);
598 /* Returns true if the current token is a string. */
600 lex_is_string (const struct lexer *lexer)
602 return lex_next_is_string (lexer, 0);
605 /* Returns the value of the current token, which must be a
606 floating point number. */
608 lex_number (const struct lexer *lexer)
610 return lex_next_number (lexer, 0);
613 /* Returns true iff the current token is an integer. */
615 lex_is_integer (const struct lexer *lexer)
617 return lex_next_is_integer (lexer, 0);
620 /* Returns the value of the current token, which must be an
623 lex_integer (const struct lexer *lexer)
625 return lex_next_integer (lexer, 0);
628 /* Token testing functions with lookahead.
630 A value of 0 for N as an argument to any of these functions refers to the
631 current token. Lookahead is limited to the current command. Any N greater
632 than the number of tokens remaining in the current command will be treated
633 as referring to a T_ENDCMD token. */
635 /* Returns true if the token N ahead of the current token is a number. */
637 lex_next_is_number (const struct lexer *lexer, int n)
639 return token_is_number (lex_next (lexer, n));
642 /* Returns true if the token N ahead of the current token is a string. */
644 lex_next_is_string (const struct lexer *lexer, int n)
646 return token_is_string (lex_next (lexer, n));
649 /* Returns the value of the token N ahead of the current token, which must be a
650 floating point number. */
652 lex_next_number (const struct lexer *lexer, int n)
654 return token_number (lex_next (lexer, n));
657 /* Returns true if the token N ahead of the current token is an integer. */
659 lex_next_is_integer (const struct lexer *lexer, int n)
661 return token_is_integer (lex_next (lexer, n));
664 /* Returns the value of the token N ahead of the current token, which must be
667 lex_next_integer (const struct lexer *lexer, int n)
669 return token_integer (lex_next (lexer, n));
672 /* Token matching functions. */
674 /* If the current token has the specified TYPE, skips it and returns true.
675 Otherwise, returns false. */
677 lex_match (struct lexer *lexer, enum token_type type)
679 if (lex_token (lexer) == type)
688 /* If the current token matches IDENTIFIER, skips it and returns true.
689 IDENTIFIER may be abbreviated to its first three letters. Otherwise,
692 IDENTIFIER must be an ASCII string. */
694 lex_match_id (struct lexer *lexer, const char *identifier)
696 return lex_match_id_n (lexer, identifier, 3);
699 /* If the current token is IDENTIFIER, skips it and returns true. IDENTIFIER
700 may be abbreviated to its first N letters. Otherwise, returns false.
702 IDENTIFIER must be an ASCII string. */
704 lex_match_id_n (struct lexer *lexer, const char *identifier, size_t n)
706 if (lex_token (lexer) == T_ID
707 && lex_id_match_n (ss_cstr (identifier), lex_tokss (lexer), n))
716 /* If the current token is integer X, skips it and returns true. Otherwise,
719 lex_match_int (struct lexer *lexer, int x)
721 if (lex_is_integer (lexer) && lex_integer (lexer) == x)
730 /* Forced matches. */
732 /* If this token is IDENTIFIER, skips it and returns true. IDENTIFIER may be
733 abbreviated to its first 3 letters. Otherwise, reports an error and returns
736 IDENTIFIER must be an ASCII string. */
738 lex_force_match_id (struct lexer *lexer, const char *identifier)
740 if (lex_match_id (lexer, identifier))
744 lex_error_expecting (lexer, identifier);
749 /* If the current token has the specified TYPE, skips it and returns true.
750 Otherwise, reports an error and returns false. */
752 lex_force_match (struct lexer *lexer, enum token_type type)
754 if (lex_token (lexer) == type)
761 const char *type_string = token_type_to_string (type);
764 char *s = xasprintf ("`%s'", type_string);
765 lex_error_expecting (lexer, s);
769 lex_error_expecting (lexer, token_type_to_name (type));
775 /* If the current token is a string, does nothing and returns true.
776 Otherwise, reports an error and returns false. */
778 lex_force_string (struct lexer *lexer)
780 if (lex_is_string (lexer))
784 lex_error (lexer, _("expecting string"));
789 /* If the current token is a string or an identifier, does nothing and returns
790 true. Otherwise, reports an error and returns false.
792 This is meant for use in syntactic situations where we want to encourage the
793 user to supply a quoted string, but for compatibility we also accept
794 identifiers. (One example of such a situation is file names.) Therefore,
795 the error message issued when the current token is wrong only says that a
796 string is expected and doesn't mention that an identifier would also be
799 lex_force_string_or_id (struct lexer *lexer)
801 return lex_token (lexer) == T_ID || lex_force_string (lexer);
804 /* If the current token is an integer, does nothing and returns true.
805 Otherwise, reports an error and returns false. */
807 lex_force_int (struct lexer *lexer)
809 if (lex_is_integer (lexer))
813 lex_error (lexer, _("expecting integer"));
818 /* If the current token is an integer in the range MIN...MAX (inclusive), does
819 nothing and returns true. Otherwise, reports an error and returns false.
820 If NAME is nonnull, then it is used in the error message. */
822 lex_force_int_range (struct lexer *lexer, const char *name, long min, long max)
824 bool is_number = lex_is_number (lexer);
825 bool is_integer = lex_is_integer (lexer);
826 bool too_small = (is_integer ? lex_integer (lexer) < min
827 : is_number ? lex_number (lexer) < min
829 bool too_big = (is_integer ? lex_integer (lexer) > max
830 : is_number ? lex_number (lexer) > max
832 if (is_integer && !too_small && !too_big)
837 /* Weird, maybe a bug in the caller. Just report that we needed an
840 lex_error (lexer, _("Integer expected for %s."), name);
842 lex_error (lexer, _("Integer expected."));
847 lex_error (lexer, _("Expected %ld for %s."), min, name);
849 lex_error (lexer, _("Expected %ld."), min);
851 else if (min + 1 == max)
854 lex_error (lexer, _("Expected %ld or %ld for %s."), min, min + 1, name);
856 lex_error (lexer, _("Expected %ld or %ld."), min, min + 1);
860 bool report_lower_bound = (min > INT_MIN / 2) || too_small;
861 bool report_upper_bound = (max < INT_MAX / 2) || too_big;
863 if (report_lower_bound && report_upper_bound)
867 _("Expected integer between %ld and %ld for %s."),
870 lex_error (lexer, _("Expected integer between %ld and %ld."),
873 else if (report_lower_bound)
878 lex_error (lexer, _("Expected non-negative integer for %s."),
881 lex_error (lexer, _("Expected non-negative integer."));
886 lex_error (lexer, _("Expected positive integer for %s."),
889 lex_error (lexer, _("Expected positive integer."));
892 else if (report_upper_bound)
896 _("Expected integer less than or equal to %ld for %s."),
899 lex_error (lexer, _("Expected integer less than or equal to %ld."),
905 lex_error (lexer, _("Integer expected for %s."), name);
907 lex_error (lexer, _("Integer expected."));
913 /* If the current token is a number, does nothing and returns true.
914 Otherwise, reports an error and returns false. */
916 lex_force_num (struct lexer *lexer)
918 if (lex_is_number (lexer))
921 lex_error (lexer, _("expecting number"));
925 /* If the current token is an identifier, does nothing and returns true.
926 Otherwise, reports an error and returns false. */
928 lex_force_id (struct lexer *lexer)
930 if (lex_token (lexer) == T_ID)
933 lex_error (lexer, _("expecting identifier"));
937 /* Token accessors. */
939 /* Returns the type of LEXER's current token. */
941 lex_token (const struct lexer *lexer)
943 return lex_next_token (lexer, 0);
946 /* Returns the number in LEXER's current token.
948 Only T_NEG_NUM and T_POS_NUM tokens have meaningful values. For other
949 tokens this function will always return zero. */
951 lex_tokval (const struct lexer *lexer)
953 return lex_next_tokval (lexer, 0);
956 /* Returns the null-terminated string in LEXER's current token, UTF-8 encoded.
958 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
959 this functions this function will always return NULL.
961 The UTF-8 encoding of the returned string is correct for variable names and
962 other identifiers. Use filename_to_utf8() to use it as a filename. Use
963 data_in() to use it in a "union value". */
965 lex_tokcstr (const struct lexer *lexer)
967 return lex_next_tokcstr (lexer, 0);
970 /* Returns the string in LEXER's current token, UTF-8 encoded. The string is
971 null-terminated (but the null terminator is not included in the returned
972 substring's 'length').
974 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
975 this functions this function will always return NULL.
977 The UTF-8 encoding of the returned string is correct for variable names and
978 other identifiers. Use filename_to_utf8() to use it as a filename. Use
979 data_in() to use it in a "union value". */
981 lex_tokss (const struct lexer *lexer)
983 return lex_next_tokss (lexer, 0);
988 A value of 0 for N as an argument to any of these functions refers to the
989 current token. Lookahead is limited to the current command. Any N greater
990 than the number of tokens remaining in the current command will be treated
991 as referring to a T_ENDCMD token. */
993 static const struct lex_token *
994 lex_next__ (const struct lexer *lexer_, int n)
996 struct lexer *lexer = CONST_CAST (struct lexer *, lexer_);
997 struct lex_source *src = lex_source__ (lexer);
1000 return lex_source_next__ (src, n);
1003 static const struct lex_token stop_token = { .token = { .type = T_STOP } };
1008 static const struct lex_token *
1009 lex_source_next__ (const struct lex_source *src_, int n)
1011 struct lex_source *src = CONST_CAST (struct lex_source *, src_);
1012 while (lex_stage_count (&src->lookahead) <= n)
1014 if (!lex_stage_is_empty (&src->lookahead))
1016 const struct lex_token *t = lex_stage_last (&src->lookahead);
1017 if (t->token.type == T_STOP || t->token.type == T_ENDCMD)
1021 lex_source_get_lookahead (src);
1024 return lex_stage_nth (&src->lookahead, n);
1027 /* Returns the "struct token" of the token N after the current one in LEXER.
1028 The returned pointer can be invalidated by pretty much any succeeding call
1029 into the lexer, although the string pointer within the returned token is
1030 only invalidated by consuming the token (e.g. with lex_get()). */
1031 const struct token *
1032 lex_next (const struct lexer *lexer, int n)
1034 return &lex_next__ (lexer, n)->token;
1037 /* Returns the type of the token N after the current one in LEXER. */
1039 lex_next_token (const struct lexer *lexer, int n)
1041 return lex_next (lexer, n)->type;
1044 /* Returns the number in the tokn N after the current one in LEXER.
1046 Only T_NEG_NUM and T_POS_NUM tokens have meaningful values. For other
1047 tokens this function will always return zero. */
1049 lex_next_tokval (const struct lexer *lexer, int n)
1051 return token_number (lex_next (lexer, n));
1054 /* Returns the null-terminated string in the token N after the current one, in
1057 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
1058 this functions this function will always return NULL.
1060 The UTF-8 encoding of the returned string is correct for variable names and
1061 other identifiers. Use filename_to_utf8() to use it as a filename. Use
1062 data_in() to use it in a "union value". */
1064 lex_next_tokcstr (const struct lexer *lexer, int n)
1066 return lex_next_tokss (lexer, n).string;
1069 /* Returns the string in the token N after the current one, in UTF-8 encoding.
1070 The string is null-terminated (but the null terminator is not included in
1071 the returned substring's 'length').
1073 Only T_ID, T_MACRO_ID, T_STRING tokens have meaningful strings. For other
1074 tokens this functions this function will always return NULL.
1076 The UTF-8 encoding of the returned string is correct for variable names and
1077 other identifiers. Use filename_to_utf8() to use it as a filename. Use
1078 data_in() to use it in a "union value". */
1080 lex_next_tokss (const struct lexer *lexer, int n)
1082 return lex_next (lexer, n)->string;
1085 /* Returns the text of the syntax in tokens N0 ahead of the current one,
1086 through N1 ahead of the current one, inclusive. (For example, if N0 and N1
1087 are both zero, this requests the syntax for the current token.) The caller
1088 must eventually free the returned string (with free()). The syntax is
1089 encoded in UTF-8 and in the original form supplied to the lexer so that, for
1090 example, it may include comments, spaces, and new-lines if it spans multiple
1091 tokens. Macro expansion, however, has already been performed. */
1093 lex_next_representation (const struct lexer *lexer, int n0, int n1)
1095 return lex_source_get_syntax__ (lex_source__ (lexer), n0, n1);
1098 /* Returns true if the token N ahead of the current one was produced by macro
1099 expansion, false otherwise. */
1101 lex_next_is_from_macro (const struct lexer *lexer, int n)
1103 return lex_next__ (lexer, n)->macro_rep != NULL;
1107 lex_tokens_match (const struct token *actual, const struct token *expected)
1109 if (actual->type != expected->type)
1112 switch (actual->type)
1116 return actual->number == expected->number;
1119 return lex_id_match (expected->string, actual->string);
1122 return (actual->string.length == expected->string.length
1123 && !memcmp (actual->string.string, expected->string.string,
1124 actual->string.length));
1131 /* If LEXER is positioned at the sequence of tokens that may be parsed from S,
1132 skips it and returns true. Otherwise, returns false.
1134 S may consist of an arbitrary sequence of tokens, e.g. "KRUSKAL-WALLIS",
1135 "2SLS", or "END INPUT PROGRAM". Identifiers may be abbreviated to their
1136 first three letters. */
1138 lex_match_phrase (struct lexer *lexer, const char *s)
1140 struct string_lexer slex;
1145 string_lexer_init (&slex, s, strlen (s), SEG_MODE_INTERACTIVE, true);
1146 while (string_lexer_next (&slex, &token))
1148 bool match = lex_tokens_match (lex_next (lexer, i++), &token);
1149 token_uninit (&token);
1160 count_newlines (char *s, size_t length)
1165 while ((newline = memchr (s, '\n', length)) != NULL)
1168 length -= (newline + 1) - s;
1176 lex_token_get_last_line_number (const struct lex_source *src,
1177 const struct lex_token *token)
1179 if (token->first_line == 0)
1183 char *token_str = &src->buffer[token->token_pos - src->tail];
1184 return token->first_line + count_newlines (token_str, token->token_len) + 1;
1189 count_columns (const char *s_, size_t length)
1191 const uint8_t *s = CHAR_CAST (const uint8_t *, s_);
1197 for (ofs = 0; ofs < length; ofs += mblen)
1201 mblen = u8_mbtouc (&uc, s + ofs, length - ofs);
1204 int width = uc_width (uc, "UTF-8");
1209 columns = ROUND_UP (columns + 1, 8);
1216 lex_token_get_first_column (const struct lex_source *src,
1217 const struct lex_token *token)
1219 return count_columns (&src->buffer[token->line_pos - src->tail],
1220 token->token_pos - token->line_pos);
1224 lex_token_get_last_column (const struct lex_source *src,
1225 const struct lex_token *token)
1227 char *start, *end, *newline;
1229 start = &src->buffer[token->line_pos - src->tail];
1230 end = &src->buffer[(token->token_pos + token->token_len) - src->tail];
1231 newline = memrchr (start, '\n', end - start);
1232 if (newline != NULL)
1233 start = newline + 1;
1234 return count_columns (start, end - start);
1237 static struct msg_location
1238 lex_token_location (const struct lex_source *src,
1239 const struct lex_token *t0,
1240 const struct lex_token *t1)
1242 return (struct msg_location) {
1243 .file_name = src->reader->file_name,
1244 .first_line = t0->first_line,
1245 .last_line = lex_token_get_last_line_number (src, t1),
1246 .first_column = lex_token_get_first_column (src, t0),
1247 .last_column = lex_token_get_last_column (src, t1),
1251 static struct msg_location *
1252 lex_token_location_rw (const struct lex_source *src,
1253 const struct lex_token *t0,
1254 const struct lex_token *t1)
1256 struct msg_location location = lex_token_location (src, t0, t1);
1257 return msg_location_dup (&location);
1260 static struct msg_location *
1261 lex_source_get_location (const struct lex_source *src, int n0, int n1)
1263 return lex_token_location_rw (src,
1264 lex_source_next__ (src, n0),
1265 lex_source_next__ (src, n1));
1268 /* Returns the 1-based line number of the start of the syntax that represents
1269 the token N after the current one in LEXER. Returns 0 for a T_STOP token or
1270 if the token is drawn from a source that does not have line numbers. */
1272 lex_get_first_line_number (const struct lexer *lexer, int n)
1274 const struct lex_source *src = lex_source__ (lexer);
1275 return src ? lex_source_next__ (src, n)->first_line : 0;
1278 /* Returns the 1-based line number of the end of the syntax that represents the
1279 token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
1280 token or if the token is drawn from a source that does not have line
1283 Most of the time, a single token is wholly within a single line of syntax,
1284 but there are two exceptions: a T_STRING token can be made up of multiple
1285 segments on adjacent lines connected with "+" punctuators, and a T_NEG_NUM
1286 token can consist of a "-" on one line followed by the number on the next.
1289 lex_get_last_line_number (const struct lexer *lexer, int n)
1291 const struct lex_source *src = lex_source__ (lexer);
1292 return src ? lex_token_get_last_line_number (src,
1293 lex_source_next__ (src, n)) : 0;
1296 /* Returns the 1-based column number of the start of the syntax that represents
1297 the token N after the current one in LEXER. Returns 0 for a T_STOP
1300 Column numbers are measured according to the width of characters as shown in
1301 a typical fixed-width font, in which CJK characters have width 2 and
1302 combining characters have width 0. */
1304 lex_get_first_column (const struct lexer *lexer, int n)
1306 const struct lex_source *src = lex_source__ (lexer);
1307 return src ? lex_token_get_first_column (src, lex_source_next__ (src, n)) : 0;
1310 /* Returns the 1-based column number of the end of the syntax that represents
1311 the token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
1314 Column numbers are measured according to the width of characters as shown in
1315 a typical fixed-width font, in which CJK characters have width 2 and
1316 combining characters have width 0. */
1318 lex_get_last_column (const struct lexer *lexer, int n)
1320 const struct lex_source *src = lex_source__ (lexer);
1321 return src ? lex_token_get_last_column (src, lex_source_next__ (src, n)) : 0;
1324 /* Returns the name of the syntax file from which the current command is drawn.
1325 Returns NULL for a T_STOP token or if the command's source does not have
1328 There is no version of this function that takes an N argument because
1329 lookahead only works to the end of a command and any given command is always
1330 within a single syntax file. */
1332 lex_get_file_name (const struct lexer *lexer)
1334 struct lex_source *src = lex_source__ (lexer);
1335 return src == NULL ? NULL : src->reader->file_name;
1338 /* Returns a newly allocated msg_location for the syntax that represents tokens
1339 with 0-based offsets N0...N1, inclusive, from the current token. The caller
1340 must eventually free the location (with msg_location_destroy()). */
1341 struct msg_location *
1342 lex_get_location (const struct lexer *lexer, int n0, int n1)
1344 struct msg_location *loc = lex_get_lines (lexer, n0, n1);
1345 loc->first_column = lex_get_first_column (lexer, n0);
1346 loc->last_column = lex_get_last_column (lexer, n1);
1350 /* Returns a newly allocated msg_location for the syntax that represents tokens
1351 with 0-based offsets N0...N1, inclusive, from the current token. The
1352 location only covers the tokens' lines, not the columns. The caller must
1353 eventually free the location (with msg_location_destroy()). */
1354 struct msg_location *
1355 lex_get_lines (const struct lexer *lexer, int n0, int n1)
1357 struct msg_location *loc = xmalloc (sizeof *loc);
1358 *loc = (struct msg_location) {
1359 .file_name = xstrdup_if_nonnull (lex_get_file_name (lexer)),
1360 .first_line = lex_get_first_line_number (lexer, n0),
1361 .last_line = lex_get_last_line_number (lexer, n1),
1367 lex_get_encoding (const struct lexer *lexer)
1369 struct lex_source *src = lex_source__ (lexer);
1370 return src == NULL ? NULL : src->reader->encoding;
1373 /* Returns the syntax mode for the syntax file from which the current drawn is
1374 drawn. Returns SEG_MODE_AUTO for a T_STOP token or if the command's source
1375 does not have line numbers.
1377 There is no version of this function that takes an N argument because
1378 lookahead only works to the end of a command and any given command is always
1379 within a single syntax file. */
1381 lex_get_syntax_mode (const struct lexer *lexer)
1383 struct lex_source *src = lex_source__ (lexer);
1384 return src == NULL ? SEG_MODE_AUTO : src->reader->syntax;
1387 /* Returns the error mode for the syntax file from which the current drawn is
1388 drawn. Returns LEX_ERROR_TERMINAL for a T_STOP token or if the command's
1389 source does not have line numbers.
1391 There is no version of this function that takes an N argument because
1392 lookahead only works to the end of a command and any given command is always
1393 within a single syntax file. */
1395 lex_get_error_mode (const struct lexer *lexer)
1397 struct lex_source *src = lex_source__ (lexer);
1398 return src == NULL ? LEX_ERROR_TERMINAL : src->reader->error;
1401 /* If the source that LEXER is currently reading has error mode
1402 LEX_ERROR_TERMINAL, discards all buffered input and tokens, so that the next
1403 token to be read comes directly from whatever is next read from the stream.
1405 It makes sense to call this function after encountering an error in a
1406 command entered on the console, because usually the user would prefer not to
1407 have cascading errors. */
1409 lex_interactive_reset (struct lexer *lexer)
1411 struct lex_source *src = lex_source__ (lexer);
1412 if (src != NULL && src->reader->error == LEX_ERROR_TERMINAL)
1414 src->head = src->tail = 0;
1415 src->journal_pos = src->seg_pos = src->line_pos = 0;
1416 src->n_newlines = 0;
1417 src->suppress_next_newline = false;
1418 src->segmenter = segmenter_init (segmenter_get_mode (&src->segmenter),
1420 lex_stage_clear (&src->pp);
1421 lex_stage_clear (&src->merge);
1422 lex_stage_clear (&src->lookahead);
1423 lex_source_push_endcmd__ (src);
1427 /* Advances past any tokens in LEXER up to a T_ENDCMD or T_STOP. */
1429 lex_discard_rest_of_command (struct lexer *lexer)
1431 while (lex_token (lexer) != T_STOP && lex_token (lexer) != T_ENDCMD)
1435 /* Discards all lookahead tokens in LEXER, then discards all input sources
1436 until it encounters one with error mode LEX_ERROR_TERMINAL or until it
1437 runs out of input sources. */
1439 lex_discard_noninteractive (struct lexer *lexer)
1441 struct lex_source *src = lex_source__ (lexer);
1445 lex_stage_clear (&src->pp);
1446 lex_stage_clear (&src->merge);
1447 lex_stage_clear (&src->lookahead);
1449 for (; src != NULL && src->reader->error != LEX_ERROR_TERMINAL;
1450 src = lex_source__ (lexer))
1451 lex_source_destroy (src);
1456 lex_source_max_tail__ (const struct lex_source *src_)
1458 struct lex_source *src = CONST_CAST (struct lex_source *, src_);
1460 assert (src->seg_pos >= src->line_pos);
1461 size_t max_tail = MIN (src->journal_pos, src->line_pos);
1463 /* Use the oldest token also. */
1464 struct lex_stage *stages[] = { &src->lookahead, &src->merge, &src->pp };
1465 for (size_t i = 0; i < sizeof stages / sizeof *stages; i++)
1466 if (!lex_stage_is_empty (stages[i]))
1468 struct lex_token *first = lex_stage_first (stages[i]);
1469 assert (first->token_pos >= first->line_pos);
1470 return MIN (max_tail, first->line_pos);
1477 lex_source_expand__ (struct lex_source *src)
1479 if (src->head - src->tail >= src->allocated)
1481 size_t max_tail = lex_source_max_tail__ (src);
1482 if (max_tail > src->tail)
1484 /* Advance the tail, freeing up room at the head. */
1485 memmove (src->buffer, src->buffer + (max_tail - src->tail),
1486 src->head - max_tail);
1487 src->tail = max_tail;
1491 /* Buffer is completely full. Expand it. */
1492 src->buffer = x2realloc (src->buffer, &src->allocated);
1497 /* There's space available at the head of the buffer. Nothing to do. */
1502 lex_source_read__ (struct lex_source *src)
1506 lex_source_expand__ (src);
1508 size_t head_ofs = src->head - src->tail;
1509 size_t space = src->allocated - head_ofs;
1510 enum prompt_style prompt = segmenter_get_prompt (&src->segmenter);
1511 size_t n = src->reader->class->read (src->reader, &src->buffer[head_ofs],
1513 assert (n <= space);
1518 src->reader->eof = true;
1519 lex_source_expand__ (src);
1525 while (!memchr (&src->buffer[src->seg_pos - src->tail], '\n',
1526 src->head - src->seg_pos));
1529 static struct lex_source *
1530 lex_source__ (const struct lexer *lexer)
1532 return (ll_is_empty (&lexer->sources) ? NULL
1533 : ll_data (ll_head (&lexer->sources), struct lex_source, ll));
1536 /* Returns the text of the syntax in SRC for tokens N0 ahead of the current
1537 one, through N1 ahead of the current one, inclusive. (For example, if N0
1538 and N1 are both zero, this requests the syntax for the current token.) The
1539 caller must eventually free the returned string (with free()). The syntax
1540 is encoded in UTF-8 and in the original form supplied to the lexer so that,
1541 for example, it may include comments, spaces, and new-lines if it spans
1542 multiple tokens. Macro expansion, however, has already been performed. */
1544 lex_source_get_syntax__ (const struct lex_source *src, int n0, int n1)
1546 struct string s = DS_EMPTY_INITIALIZER;
1547 for (size_t i = n0; i <= n1; )
1549 /* Find [I,J) as the longest sequence of tokens not produced by macro
1550 expansion, or otherwise the longest sequence expanded from a single
1552 const struct lex_token *first = lex_source_next__ (src, i);
1554 for (j = i + 1; j <= n1; j++)
1556 const struct lex_token *cur = lex_source_next__ (src, j);
1557 if ((first->macro_rep != NULL) != (cur->macro_rep != NULL)
1558 || first->macro_rep != cur->macro_rep)
1561 const struct lex_token *last = lex_source_next__ (src, j - 1);
1563 /* Now add the syntax for this sequence of tokens to SRC. */
1564 if (!ds_is_empty (&s))
1565 ds_put_byte (&s, ' ');
1566 if (!first->macro_rep)
1568 size_t start = first->token_pos;
1569 size_t end = last->token_pos + last->token_len;
1570 ds_put_substring (&s, ss_buffer (&src->buffer[start - src->tail],
1575 size_t start = first->ofs;
1576 size_t end = last->ofs + last->len;
1577 ds_put_substring (&s, ss_buffer (first->macro_rep + start,
1583 return ds_steal_cstr (&s);
1587 lex_source_contains_macro_call (struct lex_source *src, int n0, int n1)
1589 for (size_t i = n0; i <= n1; i++)
1590 if (lex_source_next__ (src, i)->macro_rep)
1595 /* If tokens N0...N1 (inclusive) in SRC contains a macro call, this returns the
1596 raw UTF-8 syntax for the macro call (not for the expansion) and for any
1597 other tokens included in that range. The syntax is encoded in UTF-8 and in
1598 the original form supplied to the lexer so that, for example, it may include
1599 comments, spaces, and new-lines if it spans multiple tokens.
1601 Returns an empty string if the token range doesn't include a macro call.
1603 The caller must not modify or free the returned string. */
1604 static struct substring
1605 lex_source_get_macro_call (struct lex_source *src, int n0, int n1)
1607 if (!lex_source_contains_macro_call (src, n0, n1))
1610 const struct lex_token *token0 = lex_source_next__ (src, n0);
1611 const struct lex_token *token1 = lex_source_next__ (src, MAX (n0, n1));
1612 size_t start = token0->token_pos;
1613 size_t end = token1->token_pos + token1->token_len;
1615 return ss_buffer (&src->buffer[start - src->tail], end - start);
1619 lex_source_error_valist (struct lex_source *src, int n0, int n1,
1620 const char *format, va_list args)
1622 const struct lex_token *token;
1627 token = lex_source_next__ (src, n0);
1628 if (token->token.type == T_ENDCMD)
1629 ds_put_cstr (&s, _("Syntax error at end of command"));
1632 /* Get the syntax that caused the error. */
1633 char *raw_syntax = lex_source_get_syntax__ (src, n0, n1);
1635 str_ellipsize (ss_cstr (raw_syntax), syntax, sizeof syntax);
1638 /* Get the macro call(s) that expanded to the syntax that caused the
1641 str_ellipsize (lex_source_get_macro_call (src, n0, n1),
1648 _("Syntax error at `%s' (in expansion of `%s')"),
1651 ds_put_format (&s, _("Syntax error at `%s'"), syntax);
1656 ds_put_format (&s, _("Syntax error in syntax expanded from `%s'"),
1659 ds_put_cstr (&s, _("Syntax error"));
1665 ds_put_cstr (&s, ": ");
1666 ds_put_vformat (&s, format, args);
1668 if (ds_last (&s) != '.')
1669 ds_put_byte (&s, '.');
1671 struct msg *m = xmalloc (sizeof *m);
1673 .category = MSG_C_SYNTAX,
1674 .severity = MSG_S_ERROR,
1675 .location = lex_source_get_location (src, n0, n1),
1676 .text = ds_steal_cstr (&s),
1682 lex_get_error (struct lex_source *src, const struct lex_token *token)
1685 str_ellipsize (ss_buffer (&src->buffer[token->token_pos - src->tail],
1687 syntax, sizeof syntax);
1689 struct string s = DS_EMPTY_INITIALIZER;
1690 ds_put_format (&s, _("Syntax error at `%s'"), syntax);
1691 ds_put_format (&s, ": %s", token->token.string.string);
1693 struct msg *m = xmalloc (sizeof *m);
1695 .category = MSG_C_SYNTAX,
1696 .severity = MSG_S_ERROR,
1697 .location = lex_token_location_rw (src, token, token),
1698 .text = ds_steal_cstr (&s),
1703 /* Attempts to append an additional token to 'pp' in SRC, reading more from the
1704 underlying lex_reader if necessary. Returns true if a new token was added
1705 to SRC's deque, false otherwise. The caller should retry failures unless
1706 SRC's 'eof' marker was set to true indicating that there will be no more
1707 tokens from this source. */
1709 lex_source_try_get_pp (struct lex_source *src)
1711 /* Append a new token to SRC and initialize it. */
1712 struct lex_token *token = xmalloc (sizeof *token);
1713 token->token = (struct token) { .type = T_STOP };
1714 token->macro_rep = NULL;
1715 token->ref_cnt = NULL;
1716 token->line_pos = src->line_pos;
1717 token->token_pos = src->seg_pos;
1718 if (src->reader->line_number > 0)
1719 token->first_line = src->reader->line_number + src->n_newlines;
1721 token->first_line = 0;
1723 /* Extract a segment. */
1724 const char *segment;
1725 enum segment_type seg_type;
1729 segment = &src->buffer[src->seg_pos - src->tail];
1730 seg_len = segmenter_push (&src->segmenter, segment,
1731 src->head - src->seg_pos,
1732 src->reader->eof, &seg_type);
1736 /* The segmenter needs more input to produce a segment. */
1737 assert (!src->reader->eof);
1738 lex_source_read__ (src);
1741 /* Update state based on the segment. */
1742 token->token_len = seg_len;
1743 src->seg_pos += seg_len;
1744 if (seg_type == SEG_NEWLINE)
1746 src->line_pos = src->seg_pos;
1750 /* Get a token from the segment. */
1751 enum tokenize_result result = token_from_segment (
1752 seg_type, ss_buffer (segment, seg_len), &token->token);
1754 /* If we've reached the end of a line, or the end of a command, then pass
1755 the line to the output engine as a syntax text item. */
1756 int n_lines = seg_type == SEG_NEWLINE;
1757 if (seg_type == SEG_END_COMMAND && !src->suppress_next_newline)
1760 src->suppress_next_newline = true;
1762 else if (n_lines > 0 && src->suppress_next_newline)
1765 src->suppress_next_newline = false;
1767 for (int i = 0; i < n_lines; i++)
1769 /* Beginning of line. */
1770 const char *line = &src->buffer[src->journal_pos - src->tail];
1772 /* Calculate line length, including \n or \r\n end-of-line if present.
1774 We use src->head even though that may be beyond what we've actually
1775 converted to tokens (which is only through line_pos). That's because,
1776 if we're emitting the line due to SEG_END_COMMAND, we want to take the
1777 whole line through the newline, not just through the '.'. */
1778 size_t max_len = src->head - src->journal_pos;
1779 const char *newline = memchr (line, '\n', max_len);
1780 size_t line_len = newline ? newline - line + 1 : max_len;
1782 /* Calculate line length excluding end-of-line. */
1783 size_t copy_len = line_len;
1784 if (copy_len > 0 && line[copy_len - 1] == '\n')
1786 if (copy_len > 0 && line[copy_len - 1] == '\r')
1789 /* Submit the line as syntax. */
1790 output_item_submit (text_item_create_nocopy (TEXT_ITEM_SYNTAX,
1791 xmemdup0 (line, copy_len),
1794 src->journal_pos += line_len;
1799 case TOKENIZE_ERROR:
1800 lex_get_error (src, token);
1802 case TOKENIZE_EMPTY:
1803 lex_token_destroy (token);
1806 case TOKENIZE_TOKEN:
1807 if (token->token.type == T_STOP)
1809 token->token.type = T_ENDCMD;
1812 lex_stage_push_last (&src->pp, token);
1818 /* Attempts to append a new token to SRC. Returns true if successful, false on
1819 failure. On failure, the end of SRC has been reached and no more tokens
1820 will be forthcoming from it.
1822 Does not make the new token available for lookahead yet; the caller must
1823 adjust SRC's 'middle' pointer to do so. */
1825 lex_source_get_pp (struct lex_source *src)
1828 if (lex_source_try_get_pp (src))
1834 lex_source_try_get_merge (const struct lex_source *src_)
1836 struct lex_source *src = CONST_CAST (struct lex_source *, src_);
1838 if (lex_stage_is_empty (&src->pp) && !lex_source_get_pp (src))
1841 if (!settings_get_mexpand ())
1843 lex_stage_shift (&src->merge, &src->pp, lex_stage_count (&src->pp));
1847 /* Now pass tokens one-by-one to the macro expander.
1849 In the common case where there is no macro to expand, the loop is not
1851 struct macro_call *mc;
1852 int n_call = macro_call_create (src->lexer->macros,
1853 &lex_stage_first (&src->pp)->token, &mc);
1854 for (int ofs = 1; !n_call; ofs++)
1856 if (lex_stage_count (&src->pp) <= ofs && !lex_source_get_pp (src))
1858 /* This should not be reachable because we always get a T_ENDCMD at
1859 the end of an input file (transformed from T_STOP by
1860 lex_source_try_get_pp()) and the macro_expander should always
1861 terminate expansion on T_ENDCMD. */
1865 const struct lex_token *t = lex_stage_nth (&src->pp, ofs);
1866 size_t start = t->token_pos;
1867 size_t end = t->token_pos + t->token_len;
1868 const struct macro_token mt = {
1870 .syntax = ss_buffer (&src->buffer[start - src->tail], end - start),
1872 const struct msg_location loc = lex_token_location (src, t, t);
1873 n_call = macro_call_add (mc, &mt, &loc);
1877 /* False alarm: no macro expansion after all. Use first token as
1878 lookahead. We'll retry macro expansion from the second token next
1880 macro_call_destroy (mc);
1881 lex_stage_shift (&src->merge, &src->pp, 1);
1885 /* The first 'n_call' tokens in 'pp', which we bracket as C0...C1, inclusive,
1886 are a macro call. (These are likely to be the only tokens in 'pp'.)
1888 const struct lex_token *c0 = lex_stage_first (&src->pp);
1889 const struct lex_token *c1 = lex_stage_nth (&src->pp, n_call - 1);
1890 struct macro_tokens expansion = { .n = 0 };
1891 struct msg_location loc = lex_token_location (src, c0, c1);
1892 macro_call_expand (mc, src->reader->syntax, &loc, &expansion);
1893 macro_call_destroy (mc);
1895 /* Convert the macro expansion into syntax for possible error messages
1897 size_t *ofs = xnmalloc (expansion.n, sizeof *ofs);
1898 size_t *len = xnmalloc (expansion.n, sizeof *len);
1899 struct string s = DS_EMPTY_INITIALIZER;
1900 macro_tokens_to_syntax (&expansion, &s, ofs, len);
1902 if (settings_get_mprint ())
1903 output_item_submit (text_item_create (TEXT_ITEM_LOG, ds_cstr (&s),
1904 _("Macro Expansion")));
1906 /* Append the macro expansion tokens to the lookahead. */
1907 if (expansion.n > 0)
1909 char *macro_rep = ds_steal_cstr (&s);
1910 size_t *ref_cnt = xmalloc (sizeof *ref_cnt);
1911 *ref_cnt = expansion.n;
1912 for (size_t i = 0; i < expansion.n; i++)
1914 struct lex_token *token = xmalloc (sizeof *token);
1915 *token = (struct lex_token) {
1916 .token = expansion.mts[i].token,
1917 .token_pos = c0->token_pos,
1918 .token_len = (c1->token_pos + c1->token_len) - c0->token_pos,
1919 .line_pos = c0->line_pos,
1920 .first_line = c0->first_line,
1921 .macro_rep = macro_rep,
1926 lex_stage_push_last (&src->merge, token);
1928 ss_dealloc (&expansion.mts[i].syntax);
1933 free (expansion.mts);
1937 /* Destroy the tokens for the call. */
1938 for (size_t i = 0; i < n_call; i++)
1939 lex_stage_pop_first (&src->pp);
1941 return expansion.n > 0;
1944 /* Attempts to obtain at least one new token into 'merge' in SRC.
1946 Returns true if successful, false on failure. In the latter case, SRC is
1947 exhausted and 'src->eof' is now true. */
1949 lex_source_get_merge (struct lex_source *src)
1952 if (lex_source_try_get_merge (src))
1957 /* Attempts to obtain at least one new token into 'lookahead' in SRC.
1959 Returns true if successful, false on failure. In the latter case, SRC is
1960 exhausted and 'src->eof' is now true. */
1962 lex_source_get_lookahead (struct lex_source *src)
1964 struct merger m = MERGER_INIT;
1966 for (size_t i = 0; ; i++)
1968 while (lex_stage_count (&src->merge) <= i && !lex_source_get_merge (src))
1970 /* We always get a T_ENDCMD at the end of an input file
1971 (transformed from T_STOP by lex_source_try_get_pp()) and
1972 merger_add() should never return -1 on T_ENDCMD. */
1973 assert (lex_stage_is_empty (&src->merge));
1977 int retval = merger_add (&m, &lex_stage_nth (&src->merge, i)->token,
1981 lex_stage_shift (&src->lookahead, &src->merge, 1);
1984 else if (retval > 0)
1986 /* Add a token that merges all the tokens together. */
1987 const struct lex_token *first = lex_stage_first (&src->merge);
1988 const struct lex_token *last = lex_stage_nth (&src->merge,
1990 bool macro = first->macro_rep && first->macro_rep == last->macro_rep;
1991 struct lex_token *t = xmalloc (sizeof *t);
1992 *t = (struct lex_token) {
1994 .token_pos = first->token_pos,
1995 .token_len = (last->token_pos - first->token_pos) + last->token_len,
1996 .line_pos = first->line_pos,
1997 .first_line = first->first_line,
1999 /* This works well if all the tokens were not expanded from macros,
2000 or if they came from the same macro expansion. It just gives up
2001 in the other (corner) cases. */
2002 .macro_rep = macro ? first->macro_rep : NULL,
2003 .ofs = macro ? first->ofs : 0,
2004 .len = macro ? (last->ofs - first->ofs) + last->len : 0,
2005 .ref_cnt = macro ? first->ref_cnt : NULL,
2009 lex_stage_push_last (&src->lookahead, t);
2011 for (int i = 0; i < retval; i++)
2012 lex_stage_pop_first (&src->merge);
2019 lex_source_push_endcmd__ (struct lex_source *src)
2021 assert (lex_stage_is_empty (&src->lookahead));
2022 struct lex_token *token = xmalloc (sizeof *token);
2023 *token = (struct lex_token) { .token = { .type = T_ENDCMD } };
2024 lex_stage_push_last (&src->lookahead, token);
2027 static struct lex_source *
2028 lex_source_create (struct lexer *lexer, struct lex_reader *reader)
2030 struct lex_source *src = xmalloc (sizeof *src);
2031 *src = (struct lex_source) {
2033 .segmenter = segmenter_init (reader->syntax, false),
2037 lex_source_push_endcmd__ (src);
2043 lex_source_destroy (struct lex_source *src)
2045 char *file_name = src->reader->file_name;
2046 char *encoding = src->reader->encoding;
2047 if (src->reader->class->destroy != NULL)
2048 src->reader->class->destroy (src->reader);
2052 lex_stage_uninit (&src->pp);
2053 lex_stage_uninit (&src->merge);
2054 lex_stage_uninit (&src->lookahead);
2055 ll_remove (&src->ll);
2059 struct lex_file_reader
2061 struct lex_reader reader;
2062 struct u8_istream *istream;
2065 static struct lex_reader_class lex_file_reader_class;
2067 /* Creates and returns a new lex_reader that will read from file FILE_NAME (or
2068 from stdin if FILE_NAME is "-"). The file is expected to be encoded with
2069 ENCODING, which should take one of the forms accepted by
2070 u8_istream_for_file(). SYNTAX and ERROR become the syntax mode and error
2071 mode of the new reader, respectively.
2073 Returns a null pointer if FILE_NAME cannot be opened. */
2075 lex_reader_for_file (const char *file_name, const char *encoding,
2076 enum segmenter_mode syntax,
2077 enum lex_error_mode error)
2079 struct lex_file_reader *r;
2080 struct u8_istream *istream;
2082 istream = (!strcmp(file_name, "-")
2083 ? u8_istream_for_fd (encoding, STDIN_FILENO)
2084 : u8_istream_for_file (encoding, file_name, O_RDONLY));
2085 if (istream == NULL)
2087 msg (ME, _("Opening `%s': %s."), file_name, strerror (errno));
2091 r = xmalloc (sizeof *r);
2092 lex_reader_init (&r->reader, &lex_file_reader_class);
2093 r->reader.syntax = syntax;
2094 r->reader.error = error;
2095 r->reader.file_name = xstrdup (file_name);
2096 r->reader.encoding = xstrdup_if_nonnull (encoding);
2097 r->reader.line_number = 1;
2098 r->istream = istream;
2103 static struct lex_file_reader *
2104 lex_file_reader_cast (struct lex_reader *r)
2106 return UP_CAST (r, struct lex_file_reader, reader);
2110 lex_file_read (struct lex_reader *r_, char *buf, size_t n,
2111 enum prompt_style prompt_style UNUSED)
2113 struct lex_file_reader *r = lex_file_reader_cast (r_);
2114 ssize_t n_read = u8_istream_read (r->istream, buf, n);
2117 msg (ME, _("Error reading `%s': %s."), r_->file_name, strerror (errno));
2124 lex_file_close (struct lex_reader *r_)
2126 struct lex_file_reader *r = lex_file_reader_cast (r_);
2128 if (u8_istream_fileno (r->istream) != STDIN_FILENO)
2130 if (u8_istream_close (r->istream) != 0)
2131 msg (ME, _("Error closing `%s': %s."), r_->file_name, strerror (errno));
2134 u8_istream_free (r->istream);
2139 static struct lex_reader_class lex_file_reader_class =
2145 struct lex_string_reader
2147 struct lex_reader reader;
2152 static struct lex_reader_class lex_string_reader_class;
2154 /* Creates and returns a new lex_reader for the contents of S, which must be
2155 encoded in the given ENCODING. The new reader takes ownership of S and will free it
2156 with ss_dealloc() when it is closed. */
2158 lex_reader_for_substring_nocopy (struct substring s, const char *encoding)
2160 struct lex_string_reader *r;
2162 r = xmalloc (sizeof *r);
2163 lex_reader_init (&r->reader, &lex_string_reader_class);
2164 r->reader.syntax = SEG_MODE_AUTO;
2165 r->reader.encoding = xstrdup_if_nonnull (encoding);
2172 /* Creates and returns a new lex_reader for a copy of null-terminated string S,
2173 which must be encoded in ENCODING. The caller retains ownership of S. */
2175 lex_reader_for_string (const char *s, const char *encoding)
2177 struct substring ss;
2178 ss_alloc_substring (&ss, ss_cstr (s));
2179 return lex_reader_for_substring_nocopy (ss, encoding);
2182 /* Formats FORMAT as a printf()-like format string and creates and returns a
2183 new lex_reader for the formatted result. */
2185 lex_reader_for_format (const char *format, const char *encoding, ...)
2187 struct lex_reader *r;
2190 va_start (args, encoding);
2191 r = lex_reader_for_substring_nocopy (ss_cstr (xvasprintf (format, args)), encoding);
2197 static struct lex_string_reader *
2198 lex_string_reader_cast (struct lex_reader *r)
2200 return UP_CAST (r, struct lex_string_reader, reader);
2204 lex_string_read (struct lex_reader *r_, char *buf, size_t n,
2205 enum prompt_style prompt_style UNUSED)
2207 struct lex_string_reader *r = lex_string_reader_cast (r_);
2210 chunk = MIN (n, r->s.length - r->offset);
2211 memcpy (buf, r->s.string + r->offset, chunk);
2218 lex_string_close (struct lex_reader *r_)
2220 struct lex_string_reader *r = lex_string_reader_cast (r_);
2226 static struct lex_reader_class lex_string_reader_class =