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"
32 #include "language/command.h"
33 #include "language/lexer/macro.h"
34 #include "language/lexer/scan.h"
35 #include "language/lexer/segment.h"
36 #include "language/lexer/token.h"
37 #include "libpspp/assertion.h"
38 #include "libpspp/cast.h"
39 #include "libpspp/deque.h"
40 #include "libpspp/i18n.h"
41 #include "libpspp/intern.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
70 size_t token_pos; /* Offset into src->buffer of token start. */
71 size_t token_len; /* Length of source for token in bytes. */
72 int first_line; /* Line number at token_pos. */
74 /* For a token obtained through macro expansion, this is just this token.
76 For a token obtained through the lexer in an ordinary way, these are
78 char *macro_rep; /* The whole macro expansion. */
79 size_t ofs; /* Offset of this token in macro_rep. */
80 size_t len; /* Length of this token in macro_rep. */
81 size_t *ref_cnt; /* Number of lex_tokens that refer to macro_rep. */
85 lex_token_destroy (struct lex_token *t)
87 token_uninit (&t->token);
90 assert (*t->ref_cnt > 0);
100 /* A deque of lex_tokens that comprises one stage in the token pipeline in a
105 struct lex_token **tokens;
108 static void lex_stage_clear (struct lex_stage *);
109 static void lex_stage_uninit (struct lex_stage *);
111 static size_t lex_stage_count (const struct lex_stage *);
112 static bool lex_stage_is_empty (const struct lex_stage *);
114 static struct lex_token *lex_stage_first (struct lex_stage *);
115 static struct lex_token *lex_stage_nth (struct lex_stage *, size_t ofs);
117 static void lex_stage_push_last (struct lex_stage *, struct lex_token *);
118 static void lex_stage_pop_first (struct lex_stage *);
120 static void lex_stage_shift (struct lex_stage *dst, struct lex_stage *src,
123 /* Deletes all the tokens from STAGE. */
125 lex_stage_clear (struct lex_stage *stage)
127 while (!deque_is_empty (&stage->deque))
128 lex_stage_pop_first (stage);
131 /* Deletes all the tokens from STAGE and frees storage for the deque. */
133 lex_stage_uninit (struct lex_stage *stage)
135 lex_stage_clear (stage);
136 free (stage->tokens);
139 /* Returns true if STAGE contains no tokens, otherwise false. */
141 lex_stage_is_empty (const struct lex_stage *stage)
143 return deque_is_empty (&stage->deque);
146 /* Returns the number of tokens in STAGE. */
148 lex_stage_count (const struct lex_stage *stage)
150 return deque_count (&stage->deque);
153 /* Returns the first token in STAGE, which must be nonempty.
154 The first token is the one accessed with the least lookahead. */
155 static struct lex_token *
156 lex_stage_first (struct lex_stage *stage)
158 return lex_stage_nth (stage, 0);
161 /* Returns the token the given INDEX in STAGE. The first token (with the least
162 lookahead) is 0, the second token is 1, and so on. There must be at least
163 INDEX + 1 tokens in STAGE. */
164 static struct lex_token *
165 lex_stage_nth (struct lex_stage *stage, size_t index)
167 return stage->tokens[deque_back (&stage->deque, index)];
170 /* Adds TOKEN so that it becomes the last token in STAGE. */
172 lex_stage_push_last (struct lex_stage *stage, struct lex_token *token)
174 if (deque_is_full (&stage->deque))
175 stage->tokens = deque_expand (&stage->deque, stage->tokens,
176 sizeof *stage->tokens);
177 stage->tokens[deque_push_front (&stage->deque)] = token;
180 /* Removes and returns the first token from STAGE. */
181 static struct lex_token *
182 lex_stage_take_first (struct lex_stage *stage)
184 return stage->tokens[deque_pop_back (&stage->deque)];
187 /* Removes the first token from STAGE and uninitializes it. */
189 lex_stage_pop_first (struct lex_stage *stage)
191 lex_token_destroy (lex_stage_take_first (stage));
194 /* Removes the first N tokens from SRC, appending them to DST as the last
197 lex_stage_shift (struct lex_stage *dst, struct lex_stage *src, size_t n)
199 for (size_t i = 0; i < n; i++)
200 lex_stage_push_last (dst, lex_stage_take_first (src));
203 /* A source of tokens, corresponding to a syntax file.
205 This is conceptually a lex_reader wrapped with everything needed to convert
206 its UTF-8 bytes into tokens. */
209 struct ll ll; /* In lexer's list of sources. */
210 struct lex_reader *reader;
212 struct segmenter segmenter;
213 bool eof; /* True if T_STOP was read from 'reader'. */
215 /* Buffer of UTF-8 bytes. */
216 char *buffer; /* Source file contents. */
217 size_t length; /* Number of bytes filled. */
218 size_t allocated; /* Number of bytes allocated. */
220 /* Offsets into 'buffer'. */
221 size_t journal_pos; /* First byte not yet output to journal. */
222 size_t seg_pos; /* First byte not yet scanned as token. */
224 int n_newlines; /* Number of new-lines up to seg_pos. */
225 bool suppress_next_newline;
229 This is a pipeline with the following stages. Each token eventually
230 made available to the parser passes through of these stages. The stages
231 are named after the processing that happens in each one.
233 Initially, tokens come from the segmenter and scanner to 'pp':
235 - pp: Tokens that need to pass through the macro preprocessor to end up
238 - merge: Tokens that need to pass through scan_merge() to end up in
241 - parse: Tokens available to the client for parsing.
243 'pp' and 'merge' store tokens only temporarily until they pass into
244 'parse'. Tokens then live in 'parse' until the command is fully
245 consumed, at which time they are freed together. */
247 struct lex_stage merge;
248 struct lex_token **parse;
249 size_t n_parse, allocated_parse, parse_ofs;
252 static struct lex_source *lex_source_create (struct lexer *,
253 struct lex_reader *);
254 static void lex_source_destroy (struct lex_source *);
259 struct ll_list sources; /* Contains "struct lex_source"s. */
260 struct macro_set *macros;
263 static struct lex_source *lex_source__ (const struct lexer *);
264 static char *lex_source_get_syntax__ (const struct lex_source *,
266 static const struct lex_token *lex_next__ (const struct lexer *, int n);
267 static void lex_source_push_endcmd__ (struct lex_source *);
268 static void lex_source_push_parse (struct lex_source *, struct lex_token *);
269 static void lex_source_clear_parse (struct lex_source *);
271 static bool lex_source_get_parse (struct lex_source *);
272 static void lex_source_error_valist (struct lex_source *, int n0, int n1,
273 const char *format, va_list)
274 PRINTF_FORMAT (4, 0);
275 static const struct lex_token *lex_source_next__ (const struct lex_source *,
278 /* Initializes READER with the specified CLASS and otherwise some reasonable
279 defaults. The caller should fill in the others members as desired. */
281 lex_reader_init (struct lex_reader *reader,
282 const struct lex_reader_class *class)
284 reader->class = class;
285 reader->syntax = SEG_MODE_AUTO;
286 reader->error = LEX_ERROR_CONTINUE;
287 reader->file_name = NULL;
288 reader->encoding = NULL;
289 reader->line_number = 0;
293 /* Frees any file name already in READER and replaces it by a copy of
294 FILE_NAME, or if FILE_NAME is null then clears any existing name. */
296 lex_reader_set_file_name (struct lex_reader *reader, const char *file_name)
298 free (reader->file_name);
299 reader->file_name = xstrdup_if_nonnull (file_name);
302 /* Creates and returns a new lexer. */
306 struct lexer *lexer = xmalloc (sizeof *lexer);
307 *lexer = (struct lexer) {
308 .sources = LL_INITIALIZER (lexer->sources),
309 .macros = macro_set_create (),
314 /* Destroys LEXER. */
316 lex_destroy (struct lexer *lexer)
320 struct lex_source *source, *next;
322 ll_for_each_safe (source, next, struct lex_source, ll, &lexer->sources)
323 lex_source_destroy (source);
324 macro_set_destroy (lexer->macros);
329 /* Adds M to LEXER's set of macros. M replaces any existing macro with the
330 same name. Takes ownership of M. */
332 lex_define_macro (struct lexer *lexer, struct macro *m)
334 macro_set_add (lexer->macros, m);
337 /* Inserts READER into LEXER so that the next token read by LEXER comes from
338 READER. Before the caller, LEXER must either be empty or at a T_ENDCMD
341 lex_include (struct lexer *lexer, struct lex_reader *reader)
343 assert (ll_is_empty (&lexer->sources) || lex_token (lexer) == T_ENDCMD);
344 ll_push_head (&lexer->sources, &lex_source_create (lexer, reader)->ll);
347 /* Appends READER to LEXER, so that it will be read after all other current
348 readers have already been read. */
350 lex_append (struct lexer *lexer, struct lex_reader *reader)
352 ll_push_tail (&lexer->sources, &lex_source_create (lexer, reader)->ll);
357 /* Advances LEXER to the next token, consuming the current token. */
359 lex_get (struct lexer *lexer)
361 struct lex_source *src;
363 src = lex_source__ (lexer);
367 if (src->parse_ofs < src->n_parse)
369 if (src->parse[src->parse_ofs]->token.type == T_ENDCMD)
370 lex_source_clear_parse (src);
375 while (src->parse_ofs == src->n_parse)
376 if (!lex_source_get_parse (src))
378 lex_source_destroy (src);
379 src = lex_source__ (lexer);
385 /* Advances LEXER by N tokens. */
387 lex_get_n (struct lexer *lexer, size_t n)
393 /* Issuing errors. */
395 /* Prints a syntax error message containing the current token and
396 given message MESSAGE (if non-null). */
398 lex_error (struct lexer *lexer, const char *format, ...)
402 va_start (args, format);
403 lex_next_error_valist (lexer, 0, 0, format, args);
407 /* Prints a syntax error message containing the current token and
408 given message MESSAGE (if non-null). */
410 lex_error_valist (struct lexer *lexer, const char *format, va_list args)
412 lex_next_error_valist (lexer, 0, 0, format, args);
415 /* Prints a syntax error message containing the current token and
416 given message MESSAGE (if non-null). */
418 lex_next_error (struct lexer *lexer, int n0, int n1, const char *format, ...)
422 va_start (args, format);
423 lex_next_error_valist (lexer, n0, n1, format, args);
427 /* Prints a syntax error message saying that one of the strings provided as
428 varargs, up to the first NULL, is expected. */
430 (lex_error_expecting) (struct lexer *lexer, ...)
434 va_start (args, lexer);
435 lex_error_expecting_valist (lexer, args);
439 /* Prints a syntax error message saying that one of the options provided in
440 ARGS, up to the first NULL, is expected. */
442 lex_error_expecting_valist (struct lexer *lexer, va_list args)
444 enum { MAX_OPTIONS = 9 };
445 const char *options[MAX_OPTIONS];
447 while (n < MAX_OPTIONS)
449 const char *option = va_arg (args, const char *);
453 options[n++] = option;
455 lex_error_expecting_array (lexer, options, n);
459 lex_error_expecting_array (struct lexer *lexer, const char **options, size_t n)
464 lex_error (lexer, NULL);
468 lex_error (lexer, _("expecting %s"), options[0]);
472 lex_error (lexer, _("expecting %s or %s"), options[0], options[1]);
476 lex_error (lexer, _("expecting %s, %s, or %s"), options[0], options[1],
481 lex_error (lexer, _("expecting %s, %s, %s, or %s"),
482 options[0], options[1], options[2], options[3]);
486 lex_error (lexer, _("expecting %s, %s, %s, %s, or %s"),
487 options[0], options[1], options[2], options[3], options[4]);
491 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, or %s"),
492 options[0], options[1], options[2], options[3], options[4],
497 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, %s, or %s"),
498 options[0], options[1], options[2], options[3], options[4],
499 options[5], options[6]);
503 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, %s, %s, or %s"),
504 options[0], options[1], options[2], options[3], options[4],
505 options[5], options[6], options[7]);
509 lex_error (lexer, NULL);
513 /* Reports an error to the effect that subcommand SBC may only be specified
516 This function does not take a lexer as an argument or use lex_error(),
517 because the result would ordinarily just be redundant: "Syntax error at
518 SUBCOMMAND: Subcommand SUBCOMMAND may only be specified once.", which does
519 not help the user find the error. */
521 lex_sbc_only_once (const char *sbc)
523 msg (SE, _("Subcommand %s may only be specified once."), sbc);
526 /* Reports an error to the effect that subcommand SBC is missing.
528 This function does not take a lexer as an argument or use lex_error(),
529 because a missing subcommand can normally be detected only after the whole
530 command has been parsed, and so lex_error() would always report "Syntax
531 error at end of command", which does not help the user find the error. */
533 lex_sbc_missing (const char *sbc)
535 msg (SE, _("Required subcommand %s was not specified."), sbc);
538 /* Reports an error to the effect that specification SPEC may only be specified
539 once within subcommand SBC. */
541 lex_spec_only_once (struct lexer *lexer, const char *sbc, const char *spec)
543 lex_error (lexer, _("%s may only be specified once within subcommand %s"),
547 /* Reports an error to the effect that specification SPEC is missing within
550 lex_spec_missing (struct lexer *lexer, const char *sbc, const char *spec)
552 lex_error (lexer, _("Required %s specification missing from %s subcommand"),
556 /* Prints a syntax error message containing the current token and
557 given message MESSAGE (if non-null). */
559 lex_next_error_valist (struct lexer *lexer, int n0, int n1,
560 const char *format, va_list args)
562 struct lex_source *src = lex_source__ (lexer);
565 lex_source_error_valist (src, n0, n1, format, args);
571 ds_put_format (&s, _("Syntax error at end of input"));
574 ds_put_cstr (&s, ": ");
575 ds_put_vformat (&s, format, args);
577 if (ds_last (&s) != '.')
578 ds_put_byte (&s, '.');
579 msg (SE, "%s", ds_cstr (&s));
584 /* Checks that we're at end of command.
585 If so, returns a successful command completion code.
586 If not, flags a syntax error and returns an error command
589 lex_end_of_command (struct lexer *lexer)
591 if (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_STOP)
593 lex_error (lexer, _("expecting end of command"));
600 /* Token testing functions. */
602 /* Returns true if the current token is a number. */
604 lex_is_number (const struct lexer *lexer)
606 return lex_next_is_number (lexer, 0);
609 /* Returns true if the current token is a string. */
611 lex_is_string (const struct lexer *lexer)
613 return lex_next_is_string (lexer, 0);
616 /* Returns the value of the current token, which must be a
617 floating point number. */
619 lex_number (const struct lexer *lexer)
621 return lex_next_number (lexer, 0);
624 /* Returns true iff the current token is an integer. */
626 lex_is_integer (const struct lexer *lexer)
628 return lex_next_is_integer (lexer, 0);
631 /* Returns the value of the current token, which must be an
634 lex_integer (const struct lexer *lexer)
636 return lex_next_integer (lexer, 0);
639 /* Token testing functions with lookahead.
641 A value of 0 for N as an argument to any of these functions refers to the
642 current token. Lookahead is limited to the current command. Any N greater
643 than the number of tokens remaining in the current command will be treated
644 as referring to a T_ENDCMD token. */
646 /* Returns true if the token N ahead of the current token is a number. */
648 lex_next_is_number (const struct lexer *lexer, int n)
650 return token_is_number (lex_next (lexer, n));
653 /* Returns true if the token N ahead of the current token is a string. */
655 lex_next_is_string (const struct lexer *lexer, int n)
657 return token_is_string (lex_next (lexer, n));
660 /* Returns the value of the token N ahead of the current token, which must be a
661 floating point number. */
663 lex_next_number (const struct lexer *lexer, int n)
665 return token_number (lex_next (lexer, n));
668 /* Returns true if the token N ahead of the current token is an integer. */
670 lex_next_is_integer (const struct lexer *lexer, int n)
672 return token_is_integer (lex_next (lexer, n));
675 /* Returns the value of the token N ahead of the current token, which must be
678 lex_next_integer (const struct lexer *lexer, int n)
680 return token_integer (lex_next (lexer, n));
683 /* Token matching functions. */
685 /* If the current token has the specified TYPE, skips it and returns true.
686 Otherwise, returns false. */
688 lex_match (struct lexer *lexer, enum token_type type)
690 if (lex_token (lexer) == type)
699 /* If the current token matches IDENTIFIER, skips it and returns true.
700 IDENTIFIER may be abbreviated to its first three letters. Otherwise,
703 IDENTIFIER must be an ASCII string. */
705 lex_match_id (struct lexer *lexer, const char *identifier)
707 return lex_match_id_n (lexer, identifier, 3);
710 /* If the current token is IDENTIFIER, skips it and returns true. IDENTIFIER
711 may be abbreviated to its first N letters. Otherwise, returns false.
713 IDENTIFIER must be an ASCII string. */
715 lex_match_id_n (struct lexer *lexer, const char *identifier, size_t n)
717 if (lex_token (lexer) == T_ID
718 && lex_id_match_n (ss_cstr (identifier), lex_tokss (lexer), n))
727 /* If the current token is integer X, skips it and returns true. Otherwise,
730 lex_match_int (struct lexer *lexer, int x)
732 if (lex_is_integer (lexer) && lex_integer (lexer) == x)
741 /* Forced matches. */
743 /* If this token is IDENTIFIER, skips it and returns true. IDENTIFIER may be
744 abbreviated to its first 3 letters. Otherwise, reports an error and returns
747 IDENTIFIER must be an ASCII string. */
749 lex_force_match_id (struct lexer *lexer, const char *identifier)
751 if (lex_match_id (lexer, identifier))
755 lex_error_expecting (lexer, identifier);
760 /* If the current token has the specified TYPE, skips it and returns true.
761 Otherwise, reports an error and returns false. */
763 lex_force_match (struct lexer *lexer, enum token_type type)
765 if (lex_token (lexer) == type)
772 const char *type_string = token_type_to_string (type);
775 char *s = xasprintf ("`%s'", type_string);
776 lex_error_expecting (lexer, s);
780 lex_error_expecting (lexer, token_type_to_name (type));
786 /* If the current token is a string, does nothing and returns true.
787 Otherwise, reports an error and returns false. */
789 lex_force_string (struct lexer *lexer)
791 if (lex_is_string (lexer))
795 lex_error (lexer, _("expecting string"));
800 /* If the current token is a string or an identifier, does nothing and returns
801 true. Otherwise, reports an error and returns false.
803 This is meant for use in syntactic situations where we want to encourage the
804 user to supply a quoted string, but for compatibility we also accept
805 identifiers. (One example of such a situation is file names.) Therefore,
806 the error message issued when the current token is wrong only says that a
807 string is expected and doesn't mention that an identifier would also be
810 lex_force_string_or_id (struct lexer *lexer)
812 return lex_token (lexer) == T_ID || lex_force_string (lexer);
815 /* If the current token is an integer, does nothing and returns true.
816 Otherwise, reports an error and returns false. */
818 lex_force_int (struct lexer *lexer)
820 if (lex_is_integer (lexer))
824 lex_error (lexer, _("expecting integer"));
829 /* If the current token is an integer in the range MIN...MAX (inclusive), does
830 nothing and returns true. Otherwise, reports an error and returns false.
831 If NAME is nonnull, then it is used in the error message. */
833 lex_force_int_range (struct lexer *lexer, const char *name, long min, long max)
835 bool is_integer = lex_is_integer (lexer);
836 bool too_small = is_integer && lex_integer (lexer) < min;
837 bool too_big = is_integer && lex_integer (lexer) > max;
838 if (is_integer && !too_small && !too_big)
843 /* Weird, maybe a bug in the caller. Just report that we needed an
846 lex_error (lexer, _("Integer expected for %s."), name);
848 lex_error (lexer, _("Integer expected."));
853 lex_error (lexer, _("Expected %ld for %s."), min, name);
855 lex_error (lexer, _("Expected %ld."), min);
857 else if (min + 1 == max)
860 lex_error (lexer, _("Expected %ld or %ld for %s."), min, min + 1, name);
862 lex_error (lexer, _("Expected %ld or %ld."), min, min + 1);
866 bool report_lower_bound = (min > INT_MIN / 2) || too_small;
867 bool report_upper_bound = (max < INT_MAX / 2) || too_big;
869 if (report_lower_bound && report_upper_bound)
873 _("Expected integer between %ld and %ld for %s."),
876 lex_error (lexer, _("Expected integer between %ld and %ld."),
879 else if (report_lower_bound)
884 lex_error (lexer, _("Expected non-negative integer for %s."),
887 lex_error (lexer, _("Expected non-negative integer."));
892 lex_error (lexer, _("Expected positive integer for %s."),
895 lex_error (lexer, _("Expected positive integer."));
898 else if (report_upper_bound)
902 _("Expected integer less than or equal to %ld for %s."),
905 lex_error (lexer, _("Expected integer less than or equal to %ld."),
911 lex_error (lexer, _("Integer expected for %s."), name);
913 lex_error (lexer, _("Integer expected."));
919 /* If the current token is a number, does nothing and returns true.
920 Otherwise, reports an error and returns false. */
922 lex_force_num (struct lexer *lexer)
924 if (lex_is_number (lexer))
927 lex_error (lexer, _("expecting number"));
931 /* If the current token is an identifier, does nothing and returns true.
932 Otherwise, reports an error and returns false. */
934 lex_force_id (struct lexer *lexer)
936 if (lex_token (lexer) == T_ID)
939 lex_error (lexer, _("expecting identifier"));
943 /* Token accessors. */
945 /* Returns the type of LEXER's current token. */
947 lex_token (const struct lexer *lexer)
949 return lex_next_token (lexer, 0);
952 /* Returns the number in LEXER's current token.
954 Only T_NEG_NUM and T_POS_NUM tokens have meaningful values. For other
955 tokens this function will always return zero. */
957 lex_tokval (const struct lexer *lexer)
959 return lex_next_tokval (lexer, 0);
962 /* Returns the null-terminated string in LEXER's current token, UTF-8 encoded.
964 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
965 this functions this function will always return NULL.
967 The UTF-8 encoding of the returned string is correct for variable names and
968 other identifiers. Use filename_to_utf8() to use it as a filename. Use
969 data_in() to use it in a "union value". */
971 lex_tokcstr (const struct lexer *lexer)
973 return lex_next_tokcstr (lexer, 0);
976 /* Returns the string in LEXER's current token, UTF-8 encoded. The string is
977 null-terminated (but the null terminator is not included in the returned
978 substring's 'length').
980 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
981 this functions this function will always return NULL.
983 The UTF-8 encoding of the returned string is correct for variable names and
984 other identifiers. Use filename_to_utf8() to use it as a filename. Use
985 data_in() to use it in a "union value". */
987 lex_tokss (const struct lexer *lexer)
989 return lex_next_tokss (lexer, 0);
994 A value of 0 for N as an argument to any of these functions refers to the
995 current token. Lookahead is limited to the current command. Any N greater
996 than the number of tokens remaining in the current command will be treated
997 as referring to a T_ENDCMD token. */
999 static const struct lex_token *
1000 lex_next__ (const struct lexer *lexer_, int n)
1002 struct lexer *lexer = CONST_CAST (struct lexer *, lexer_);
1003 struct lex_source *src = lex_source__ (lexer);
1006 return lex_source_next__ (src, n);
1009 static const struct lex_token stop_token = { .token = { .type = T_STOP } };
1014 static const struct lex_token *
1015 lex_source_next__ (const struct lex_source *src_, int n)
1017 struct lex_source *src = CONST_CAST (struct lex_source *, src_);
1021 if (-n <= src->parse_ofs)
1022 return src->parse[src->parse_ofs - (-n)];
1025 static const struct lex_token endcmd_token
1026 = { .token = { .type = T_ENDCMD } };
1027 return &endcmd_token;
1031 while (src->n_parse - src->parse_ofs <= n)
1033 if (src->n_parse > 0)
1035 const struct lex_token *t = src->parse[src->n_parse - 1];
1036 if (t->token.type == T_STOP || t->token.type == T_ENDCMD)
1040 lex_source_get_parse (src);
1043 return src->parse[src->parse_ofs + n];
1046 /* Returns the "struct token" of the token N after the current one in LEXER.
1047 The returned pointer can be invalidated by pretty much any succeeding call
1048 into the lexer, although the string pointer within the returned token is
1049 only invalidated by consuming the token (e.g. with lex_get()). */
1050 const struct token *
1051 lex_next (const struct lexer *lexer, int n)
1053 return &lex_next__ (lexer, n)->token;
1056 /* Returns the type of the token N after the current one in LEXER. */
1058 lex_next_token (const struct lexer *lexer, int n)
1060 return lex_next (lexer, n)->type;
1063 /* Returns the number in the tokn N after the current one in LEXER.
1065 Only T_NEG_NUM and T_POS_NUM tokens have meaningful values. For other
1066 tokens this function will always return zero. */
1068 lex_next_tokval (const struct lexer *lexer, int n)
1070 return token_number (lex_next (lexer, n));
1073 /* Returns the null-terminated string in the token N after the current one, in
1076 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
1077 this functions this function will always return NULL.
1079 The UTF-8 encoding of the returned string is correct for variable names and
1080 other identifiers. Use filename_to_utf8() to use it as a filename. Use
1081 data_in() to use it in a "union value". */
1083 lex_next_tokcstr (const struct lexer *lexer, int n)
1085 return lex_next_tokss (lexer, n).string;
1088 /* Returns the string in the token N after the current one, in UTF-8 encoding.
1089 The string is null-terminated (but the null terminator is not included in
1090 the returned substring's 'length').
1092 Only T_ID, T_MACRO_ID, T_STRING tokens have meaningful strings. For other
1093 tokens this functions this function will always return NULL.
1095 The UTF-8 encoding of the returned string is correct for variable names and
1096 other identifiers. Use filename_to_utf8() to use it as a filename. Use
1097 data_in() to use it in a "union value". */
1099 lex_next_tokss (const struct lexer *lexer, int n)
1101 return lex_next (lexer, n)->string;
1104 /* Returns the text of the syntax in tokens N0 ahead of the current one,
1105 through N1 ahead of the current one, inclusive. (For example, if N0 and N1
1106 are both zero, this requests the syntax for the current token.) The caller
1107 must eventually free the returned string (with free()). The syntax is
1108 encoded in UTF-8 and in the original form supplied to the lexer so that, for
1109 example, it may include comments, spaces, and new-lines if it spans multiple
1110 tokens. Macro expansion, however, has already been performed. */
1112 lex_next_representation (const struct lexer *lexer, int n0, int n1)
1114 return lex_source_get_syntax__ (lex_source__ (lexer), n0, n1);
1117 /* Returns true if the token N ahead of the current one was produced by macro
1118 expansion, false otherwise. */
1120 lex_next_is_from_macro (const struct lexer *lexer, int n)
1122 return lex_next__ (lexer, n)->macro_rep != NULL;
1126 lex_tokens_match (const struct token *actual, const struct token *expected)
1128 if (actual->type != expected->type)
1131 switch (actual->type)
1135 return actual->number == expected->number;
1138 return lex_id_match (expected->string, actual->string);
1141 return (actual->string.length == expected->string.length
1142 && !memcmp (actual->string.string, expected->string.string,
1143 actual->string.length));
1151 lex_at_phrase__ (struct lexer *lexer, const char *s)
1153 struct string_lexer slex;
1157 string_lexer_init (&slex, s, strlen (s), SEG_MODE_INTERACTIVE, true);
1158 while (string_lexer_next (&slex, &token))
1160 bool match = lex_tokens_match (lex_next (lexer, i++), &token);
1161 token_uninit (&token);
1168 /* If LEXER is positioned at the sequence of tokens that may be parsed from S,
1169 returns true. Otherwise, returns false.
1171 S may consist of an arbitrary sequence of tokens, e.g. "KRUSKAL-WALLIS",
1172 "2SLS", or "END INPUT PROGRAM". Identifiers may be abbreviated to their
1173 first three letters. */
1175 lex_at_phrase (struct lexer *lexer, const char *s)
1177 return lex_at_phrase__ (lexer, s) > 0;
1180 /* If LEXER is positioned at the sequence of tokens that may be parsed from S,
1181 skips it and returns true. Otherwise, returns false.
1183 S may consist of an arbitrary sequence of tokens, e.g. "KRUSKAL-WALLIS",
1184 "2SLS", or "END INPUT PROGRAM". Identifiers may be abbreviated to their
1185 first three letters. */
1187 lex_match_phrase (struct lexer *lexer, const char *s)
1189 size_t n = lex_at_phrase__ (lexer, s);
1191 lex_get_n (lexer, n);
1196 count_newlines (char *s, size_t length)
1201 while ((newline = memchr (s, '\n', length)) != NULL)
1204 length -= (newline + 1) - s;
1212 lex_token_get_last_line_number (const struct lex_source *src,
1213 const struct lex_token *token)
1215 if (token->first_line == 0)
1219 char *token_str = &src->buffer[token->token_pos];
1220 return token->first_line + count_newlines (token_str, token->token_len) + 1;
1225 lex_token_get_column__ (const struct lex_source *src, size_t offset)
1227 const char *newline = memrchr (src->buffer, '\n', offset);
1228 size_t line_ofs = newline ? newline - src->buffer + 1 : 0;
1229 return utf8_count_columns (&src->buffer[line_ofs], offset - line_ofs) + 1;
1233 lex_token_get_first_column (const struct lex_source *src,
1234 const struct lex_token *token)
1236 return lex_token_get_column__ (src, token->token_pos);
1240 lex_token_get_last_column (const struct lex_source *src,
1241 const struct lex_token *token)
1243 return lex_token_get_column__ (src, token->token_pos + token->token_len);
1246 static struct msg_location
1247 lex_token_location (const struct lex_source *src,
1248 const struct lex_token *t0,
1249 const struct lex_token *t1)
1251 int first_column = lex_token_get_first_column (src, t0);
1252 int last_line = lex_token_get_last_line_number (src, t1) - 1;
1253 int last_column = lex_token_get_last_column (src, t1) - 1;
1254 return (struct msg_location) {
1255 .file_name = intern_new_if_nonnull (src->reader->file_name),
1256 .p[0] = { .line = t0->first_line, .column = first_column },
1257 .p[1] = { .line = last_line, .column = last_column },
1261 static struct msg_location *
1262 lex_token_location_rw (const struct lex_source *src,
1263 const struct lex_token *t0,
1264 const struct lex_token *t1)
1266 struct msg_location location = lex_token_location (src, t0, t1);
1267 return msg_location_dup (&location);
1270 static struct msg_location *
1271 lex_source_get_location (const struct lex_source *src, int n0, int n1)
1273 return lex_token_location_rw (src,
1274 lex_source_next__ (src, n0),
1275 lex_source_next__ (src, n1));
1278 /* Returns the 1-based line number of the start of the syntax that represents
1279 the token N after the current one in LEXER. Returns 0 for a T_STOP token or
1280 if the token is drawn from a source that does not have line numbers. */
1282 lex_get_first_line_number (const struct lexer *lexer, int n)
1284 const struct lex_source *src = lex_source__ (lexer);
1285 return src ? lex_source_next__ (src, n)->first_line : 0;
1288 /* Returns the 1-based line number of the end of the syntax that represents the
1289 token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
1290 token or if the token is drawn from a source that does not have line
1293 Most of the time, a single token is wholly within a single line of syntax,
1294 but there are two exceptions: a T_STRING token can be made up of multiple
1295 segments on adjacent lines connected with "+" punctuators, and a T_NEG_NUM
1296 token can consist of a "-" on one line followed by the number on the next.
1299 lex_get_last_line_number (const struct lexer *lexer, int n)
1301 const struct lex_source *src = lex_source__ (lexer);
1302 return src ? lex_token_get_last_line_number (src,
1303 lex_source_next__ (src, n)) : 0;
1306 /* Returns the 1-based column number of the start of the syntax that represents
1307 the token N after the current one in LEXER. Returns 0 for a T_STOP
1310 Column numbers are measured according to the width of characters as shown in
1311 a typical fixed-width font, in which CJK characters have width 2 and
1312 combining characters have width 0. */
1314 lex_get_first_column (const struct lexer *lexer, int n)
1316 const struct lex_source *src = lex_source__ (lexer);
1317 return src ? lex_token_get_first_column (src, lex_source_next__ (src, n)) : 0;
1320 /* Returns the 1-based column number of the end of the syntax that represents
1321 the token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
1324 Column numbers are measured according to the width of characters as shown in
1325 a typical fixed-width font, in which CJK characters have width 2 and
1326 combining characters have width 0. */
1328 lex_get_last_column (const struct lexer *lexer, int n)
1330 const struct lex_source *src = lex_source__ (lexer);
1331 return src ? lex_token_get_last_column (src, lex_source_next__ (src, n)) : 0;
1334 /* Returns the name of the syntax file from which the current command is drawn.
1335 Returns NULL for a T_STOP token or if the command's source does not have
1338 There is no version of this function that takes an N argument because
1339 lookahead only works to the end of a command and any given command is always
1340 within a single syntax file. */
1342 lex_get_file_name (const struct lexer *lexer)
1344 struct lex_source *src = lex_source__ (lexer);
1345 return src == NULL ? NULL : src->reader->file_name;
1348 /* Returns a newly allocated msg_location for the syntax that represents tokens
1349 with 0-based offsets N0...N1, inclusive, from the current token. The caller
1350 must eventually free the location (with msg_location_destroy()). */
1351 struct msg_location *
1352 lex_get_location (const struct lexer *lexer, int n0, int n1)
1354 struct msg_location *loc = lex_get_lines (lexer, n0, n1);
1355 loc->p[0].column = lex_get_first_column (lexer, n0);
1356 loc->p[1].column = lex_get_last_column (lexer, n1) - 1;
1360 /* Returns a newly allocated msg_location for the syntax that represents tokens
1361 with 0-based offsets N0...N1, inclusive, from the current token. The
1362 location only covers the tokens' lines, not the columns. The caller must
1363 eventually free the location (with msg_location_destroy()). */
1364 struct msg_location *
1365 lex_get_lines (const struct lexer *lexer, int n0, int n1)
1367 struct msg_location *loc = xmalloc (sizeof *loc);
1368 int first_line = lex_get_first_line_number (lexer, n0);
1369 int last_line = lex_get_last_line_number (lexer, n1) - 1;
1370 *loc = (struct msg_location) {
1371 .file_name = intern_new_if_nonnull (lex_get_file_name (lexer)),
1372 .p[0] = { .line = first_line },
1373 .p[1] = { .line = last_line },
1379 lex_extend_location (const struct lexer *lexer, int n, struct msg_location *loc)
1381 struct msg_location *new = lex_get_location (lexer, n, n);
1382 msg_location_merge (loc, new);
1383 msg_location_destroy (new);
1387 lex_get_encoding (const struct lexer *lexer)
1389 struct lex_source *src = lex_source__ (lexer);
1390 return src == NULL ? NULL : src->reader->encoding;
1393 /* Returns the syntax mode for the syntax file from which the current drawn is
1394 drawn. Returns SEG_MODE_AUTO for a T_STOP token or if the command's source
1395 does not have line numbers.
1397 There is no version of this function that takes an N argument because
1398 lookahead only works to the end of a command and any given command is always
1399 within a single syntax file. */
1401 lex_get_syntax_mode (const struct lexer *lexer)
1403 struct lex_source *src = lex_source__ (lexer);
1404 return src == NULL ? SEG_MODE_AUTO : src->reader->syntax;
1407 /* Returns the error mode for the syntax file from which the current drawn is
1408 drawn. Returns LEX_ERROR_TERMINAL for a T_STOP token or if the command's
1409 source does not have line numbers.
1411 There is no version of this function that takes an N argument because
1412 lookahead only works to the end of a command and any given command is always
1413 within a single syntax file. */
1415 lex_get_error_mode (const struct lexer *lexer)
1417 struct lex_source *src = lex_source__ (lexer);
1418 return src == NULL ? LEX_ERROR_TERMINAL : src->reader->error;
1421 /* If the source that LEXER is currently reading has error mode
1422 LEX_ERROR_TERMINAL, discards all buffered input and tokens, so that the next
1423 token to be read comes directly from whatever is next read from the stream.
1425 It makes sense to call this function after encountering an error in a
1426 command entered on the console, because usually the user would prefer not to
1427 have cascading errors. */
1429 lex_interactive_reset (struct lexer *lexer)
1431 struct lex_source *src = lex_source__ (lexer);
1432 if (src != NULL && src->reader->error == LEX_ERROR_TERMINAL)
1435 src->journal_pos = src->seg_pos = 0;
1436 src->n_newlines = 0;
1437 src->suppress_next_newline = false;
1438 src->segmenter = segmenter_init (segmenter_get_mode (&src->segmenter),
1440 lex_stage_clear (&src->pp);
1441 lex_stage_clear (&src->merge);
1442 lex_source_clear_parse (src);
1443 lex_source_push_endcmd__ (src);
1447 /* Advances past any tokens in LEXER up to a T_ENDCMD or T_STOP. */
1449 lex_discard_rest_of_command (struct lexer *lexer)
1451 while (lex_token (lexer) != T_STOP && lex_token (lexer) != T_ENDCMD)
1455 /* Discards all lookahead tokens in LEXER, then discards all input sources
1456 until it encounters one with error mode LEX_ERROR_TERMINAL or until it
1457 runs out of input sources. */
1459 lex_discard_noninteractive (struct lexer *lexer)
1461 struct lex_source *src = lex_source__ (lexer);
1465 lex_stage_clear (&src->pp);
1466 lex_stage_clear (&src->merge);
1467 lex_source_clear_parse (src);
1469 for (; src != NULL && src->reader->error != LEX_ERROR_TERMINAL;
1470 src = lex_source__ (lexer))
1471 lex_source_destroy (src);
1476 lex_source_expand__ (struct lex_source *src)
1478 if (src->length >= src->allocated)
1479 src->buffer = x2realloc (src->buffer, &src->allocated);
1483 lex_source_read__ (struct lex_source *src)
1487 lex_source_expand__ (src);
1489 size_t space = src->allocated - src->length;
1490 enum prompt_style prompt = segmenter_get_prompt (&src->segmenter);
1491 size_t n = src->reader->class->read (src->reader,
1492 &src->buffer[src->length],
1494 assert (n <= space);
1499 src->reader->eof = true;
1500 lex_source_expand__ (src);
1506 while (!memchr (&src->buffer[src->seg_pos], '\n',
1507 src->length - src->seg_pos));
1510 static struct lex_source *
1511 lex_source__ (const struct lexer *lexer)
1513 return (ll_is_empty (&lexer->sources) ? NULL
1514 : ll_data (ll_head (&lexer->sources), struct lex_source, ll));
1517 /* Returns the text of the syntax in SRC for tokens N0 ahead of the current
1518 one, through N1 ahead of the current one, inclusive. (For example, if N0
1519 and N1 are both zero, this requests the syntax for the current token.) The
1520 caller must eventually free the returned string (with free()). The syntax
1521 is encoded in UTF-8 and in the original form supplied to the lexer so that,
1522 for example, it may include comments, spaces, and new-lines if it spans
1523 multiple tokens. Macro expansion, however, has already been performed. */
1525 lex_source_get_syntax__ (const struct lex_source *src, int n0, int n1)
1527 struct string s = DS_EMPTY_INITIALIZER;
1528 for (size_t i = n0; i <= n1; )
1530 /* Find [I,J) as the longest sequence of tokens not produced by macro
1531 expansion, or otherwise the longest sequence expanded from a single
1533 const struct lex_token *first = lex_source_next__ (src, i);
1535 for (j = i + 1; j <= n1; j++)
1537 const struct lex_token *cur = lex_source_next__ (src, j);
1538 if ((first->macro_rep != NULL) != (cur->macro_rep != NULL)
1539 || first->macro_rep != cur->macro_rep)
1542 const struct lex_token *last = lex_source_next__ (src, j - 1);
1544 /* Now add the syntax for this sequence of tokens to SRC. */
1545 if (!ds_is_empty (&s))
1546 ds_put_byte (&s, ' ');
1547 if (!first->macro_rep)
1549 size_t start = first->token_pos;
1550 size_t end = last->token_pos + last->token_len;
1551 ds_put_substring (&s, ss_buffer (&src->buffer[start], end - start));
1555 size_t start = first->ofs;
1556 size_t end = last->ofs + last->len;
1557 ds_put_substring (&s, ss_buffer (first->macro_rep + start,
1563 return ds_steal_cstr (&s);
1567 lex_source_contains_macro_call (struct lex_source *src, int n0, int n1)
1569 for (size_t i = n0; i <= n1; i++)
1570 if (lex_source_next__ (src, i)->macro_rep)
1575 /* If tokens N0...N1 (inclusive) in SRC contains a macro call, this returns the
1576 raw UTF-8 syntax for the macro call (not for the expansion) and for any
1577 other tokens included in that range. The syntax is encoded in UTF-8 and in
1578 the original form supplied to the lexer so that, for example, it may include
1579 comments, spaces, and new-lines if it spans multiple tokens.
1581 Returns an empty string if the token range doesn't include a macro call.
1583 The caller must not modify or free the returned string. */
1584 static struct substring
1585 lex_source_get_macro_call (struct lex_source *src, int n0, int n1)
1587 if (!lex_source_contains_macro_call (src, n0, n1))
1590 const struct lex_token *token0 = lex_source_next__ (src, n0);
1591 const struct lex_token *token1 = lex_source_next__ (src, MAX (n0, n1));
1592 size_t start = token0->token_pos;
1593 size_t end = token1->token_pos + token1->token_len;
1595 return ss_buffer (&src->buffer[start], end - start);
1599 lex_source_error_valist (struct lex_source *src, int n0, int n1,
1600 const char *format, va_list args)
1602 const struct lex_token *token;
1607 token = lex_source_next__ (src, n0);
1608 if (token->token.type == T_ENDCMD)
1609 ds_put_cstr (&s, _("Syntax error at end of command"));
1612 /* Get the syntax that caused the error. */
1613 char *raw_syntax = lex_source_get_syntax__ (src, n0, n1);
1615 str_ellipsize (ss_cstr (raw_syntax), syntax, sizeof syntax);
1618 /* Get the macro call(s) that expanded to the syntax that caused the
1621 str_ellipsize (lex_source_get_macro_call (src, n0, n1),
1628 _("Syntax error at `%s' (in expansion of `%s')"),
1631 ds_put_format (&s, _("Syntax error at `%s'"), syntax);
1636 ds_put_format (&s, _("Syntax error in syntax expanded from `%s'"),
1639 ds_put_cstr (&s, _("Syntax error"));
1645 ds_put_cstr (&s, ": ");
1646 ds_put_vformat (&s, format, args);
1648 if (ds_last (&s) != '.')
1649 ds_put_byte (&s, '.');
1651 struct msg *m = xmalloc (sizeof *m);
1653 .category = MSG_C_SYNTAX,
1654 .severity = MSG_S_ERROR,
1655 .location = lex_source_get_location (src, n0, n1),
1656 .text = ds_steal_cstr (&s),
1662 lex_get_error (struct lex_source *src, const struct lex_token *token)
1665 str_ellipsize (ss_buffer (&src->buffer[token->token_pos], token->token_len),
1666 syntax, sizeof syntax);
1668 struct string s = DS_EMPTY_INITIALIZER;
1669 ds_put_format (&s, _("Syntax error at `%s'"), syntax);
1670 ds_put_format (&s, ": %s", token->token.string.string);
1672 struct msg *m = xmalloc (sizeof *m);
1674 .category = MSG_C_SYNTAX,
1675 .severity = MSG_S_ERROR,
1676 .location = lex_token_location_rw (src, token, token),
1677 .text = ds_steal_cstr (&s),
1682 /* Attempts to append an additional token to 'pp' in SRC, reading more from the
1683 underlying lex_reader if necessary. Returns true if a new token was added
1684 to SRC's deque, false otherwise. The caller should retry failures unless
1685 SRC's 'eof' marker was set to true indicating that there will be no more
1686 tokens from this source. */
1688 lex_source_try_get_pp (struct lex_source *src)
1690 /* Append a new token to SRC and initialize it. */
1691 struct lex_token *token = xmalloc (sizeof *token);
1692 token->token = (struct token) { .type = T_STOP };
1693 token->macro_rep = NULL;
1694 token->ref_cnt = NULL;
1695 token->token_pos = src->seg_pos;
1696 if (src->reader->line_number > 0)
1697 token->first_line = src->reader->line_number + src->n_newlines;
1699 token->first_line = 0;
1701 /* Extract a segment. */
1702 const char *segment;
1703 enum segment_type seg_type;
1707 segment = &src->buffer[src->seg_pos];
1708 seg_len = segmenter_push (&src->segmenter, segment,
1709 src->length - src->seg_pos,
1710 src->reader->eof, &seg_type);
1714 /* The segmenter needs more input to produce a segment. */
1715 assert (!src->reader->eof);
1716 lex_source_read__ (src);
1719 /* Update state based on the segment. */
1720 token->token_len = seg_len;
1721 src->seg_pos += seg_len;
1722 if (seg_type == SEG_NEWLINE)
1725 /* Get a token from the segment. */
1726 enum tokenize_result result = token_from_segment (
1727 seg_type, ss_buffer (segment, seg_len), &token->token);
1729 /* If we've reached the end of a line, or the end of a command, then pass
1730 the line to the output engine as a syntax text item. */
1731 int n_lines = seg_type == SEG_NEWLINE;
1732 if (seg_type == SEG_END_COMMAND && !src->suppress_next_newline)
1735 src->suppress_next_newline = true;
1737 else if (n_lines > 0 && src->suppress_next_newline)
1740 src->suppress_next_newline = false;
1742 for (int i = 0; i < n_lines; i++)
1744 /* Beginning of line. */
1745 const char *line = &src->buffer[src->journal_pos];
1747 /* Calculate line length, including \n or \r\n end-of-line if present.
1749 We use src->head even though that may be beyond what we've actually
1750 converted to tokens (which is only through line_pos). That's because,
1751 if we're emitting the line due to SEG_END_COMMAND, we want to take the
1752 whole line through the newline, not just through the '.'. */
1753 size_t max_len = src->length - src->journal_pos;
1754 const char *newline = memchr (line, '\n', max_len);
1755 size_t line_len = newline ? newline - line + 1 : max_len;
1757 /* Calculate line length excluding end-of-line. */
1758 size_t copy_len = line_len;
1759 if (copy_len > 0 && line[copy_len - 1] == '\n')
1761 if (copy_len > 0 && line[copy_len - 1] == '\r')
1764 /* Submit the line as syntax. */
1765 output_item_submit (text_item_create_nocopy (TEXT_ITEM_SYNTAX,
1766 xmemdup0 (line, copy_len),
1769 src->journal_pos += line_len;
1774 case TOKENIZE_ERROR:
1775 lex_get_error (src, token);
1777 case TOKENIZE_EMPTY:
1778 lex_token_destroy (token);
1781 case TOKENIZE_TOKEN:
1782 if (token->token.type == T_STOP)
1784 token->token.type = T_ENDCMD;
1787 lex_stage_push_last (&src->pp, token);
1793 /* Attempts to append a new token to SRC. Returns true if successful, false on
1794 failure. On failure, the end of SRC has been reached and no more tokens
1795 will be forthcoming from it.
1797 Does not make the new token available for lookahead yet; the caller must
1798 adjust SRC's 'middle' pointer to do so. */
1800 lex_source_get_pp (struct lex_source *src)
1803 if (lex_source_try_get_pp (src))
1809 lex_source_try_get_merge (const struct lex_source *src_)
1811 struct lex_source *src = CONST_CAST (struct lex_source *, src_);
1813 if (lex_stage_is_empty (&src->pp) && !lex_source_get_pp (src))
1816 if (!settings_get_mexpand ())
1818 lex_stage_shift (&src->merge, &src->pp, lex_stage_count (&src->pp));
1822 /* Now pass tokens one-by-one to the macro expander.
1824 In the common case where there is no macro to expand, the loop is not
1826 struct macro_call *mc;
1827 int n_call = macro_call_create (src->lexer->macros,
1828 &lex_stage_first (&src->pp)->token, &mc);
1829 for (int ofs = 1; !n_call; ofs++)
1831 if (lex_stage_count (&src->pp) <= ofs && !lex_source_get_pp (src))
1833 /* This should not be reachable because we always get a T_ENDCMD at
1834 the end of an input file (transformed from T_STOP by
1835 lex_source_try_get_pp()) and the macro_expander should always
1836 terminate expansion on T_ENDCMD. */
1840 const struct lex_token *t = lex_stage_nth (&src->pp, ofs);
1841 size_t start = t->token_pos;
1842 size_t end = t->token_pos + t->token_len;
1843 const struct macro_token mt = {
1845 .syntax = ss_buffer (&src->buffer[start], end - start),
1847 const struct msg_location loc = lex_token_location (src, t, t);
1848 n_call = macro_call_add (mc, &mt, &loc);
1852 /* False alarm: no macro expansion after all. Use first token as
1853 lookahead. We'll retry macro expansion from the second token next
1855 macro_call_destroy (mc);
1856 lex_stage_shift (&src->merge, &src->pp, 1);
1860 /* The first 'n_call' tokens in 'pp', which we bracket as C0...C1, inclusive,
1861 are a macro call. (These are likely to be the only tokens in 'pp'.)
1863 const struct lex_token *c0 = lex_stage_first (&src->pp);
1864 const struct lex_token *c1 = lex_stage_nth (&src->pp, n_call - 1);
1865 struct macro_tokens expansion = { .n = 0 };
1866 struct msg_location loc = lex_token_location (src, c0, c1);
1867 macro_call_expand (mc, src->reader->syntax, &loc, &expansion);
1868 macro_call_destroy (mc);
1870 /* Convert the macro expansion into syntax for possible error messages
1872 size_t *ofs = xnmalloc (expansion.n, sizeof *ofs);
1873 size_t *len = xnmalloc (expansion.n, sizeof *len);
1874 struct string s = DS_EMPTY_INITIALIZER;
1875 macro_tokens_to_syntax (&expansion, &s, ofs, len);
1877 if (settings_get_mprint ())
1878 output_item_submit (text_item_create (TEXT_ITEM_LOG, ds_cstr (&s),
1879 _("Macro Expansion")));
1881 /* Append the macro expansion tokens to the lookahead. */
1882 if (expansion.n > 0)
1884 char *macro_rep = ds_steal_cstr (&s);
1885 size_t *ref_cnt = xmalloc (sizeof *ref_cnt);
1886 *ref_cnt = expansion.n;
1887 for (size_t i = 0; i < expansion.n; i++)
1889 struct lex_token *token = xmalloc (sizeof *token);
1890 *token = (struct lex_token) {
1891 .token = expansion.mts[i].token,
1892 .token_pos = c0->token_pos,
1893 .token_len = (c1->token_pos + c1->token_len) - c0->token_pos,
1894 .first_line = c0->first_line,
1895 .macro_rep = macro_rep,
1900 lex_stage_push_last (&src->merge, token);
1902 ss_dealloc (&expansion.mts[i].syntax);
1907 free (expansion.mts);
1911 /* Destroy the tokens for the call. */
1912 for (size_t i = 0; i < n_call; i++)
1913 lex_stage_pop_first (&src->pp);
1915 return expansion.n > 0;
1918 /* Attempts to obtain at least one new token into 'merge' in SRC.
1920 Returns true if successful, false on failure. In the latter case, SRC is
1921 exhausted and 'src->eof' is now true. */
1923 lex_source_get_merge (struct lex_source *src)
1926 if (lex_source_try_get_merge (src))
1931 /* Attempts to obtain at least one new token into 'lookahead' in SRC.
1933 Returns true if successful, false on failure. In the latter case, SRC is
1934 exhausted and 'src->eof' is now true. */
1936 lex_source_get_parse (struct lex_source *src)
1938 struct merger m = MERGER_INIT;
1940 for (size_t i = 0; ; i++)
1942 while (lex_stage_count (&src->merge) <= i && !lex_source_get_merge (src))
1944 /* We always get a T_ENDCMD at the end of an input file
1945 (transformed from T_STOP by lex_source_try_get_pp()) and
1946 merger_add() should never return -1 on T_ENDCMD. */
1947 assert (lex_stage_is_empty (&src->merge));
1951 int retval = merger_add (&m, &lex_stage_nth (&src->merge, i)->token,
1955 lex_source_push_parse (src, lex_stage_take_first (&src->merge));
1958 else if (retval > 0)
1960 /* Add a token that merges all the tokens together. */
1961 const struct lex_token *first = lex_stage_first (&src->merge);
1962 const struct lex_token *last = lex_stage_nth (&src->merge,
1964 bool macro = first->macro_rep && first->macro_rep == last->macro_rep;
1965 struct lex_token *t = xmalloc (sizeof *t);
1966 *t = (struct lex_token) {
1968 .token_pos = first->token_pos,
1969 .token_len = (last->token_pos - first->token_pos) + last->token_len,
1970 .first_line = first->first_line,
1972 /* This works well if all the tokens were not expanded from macros,
1973 or if they came from the same macro expansion. It just gives up
1974 in the other (corner) cases. */
1975 .macro_rep = macro ? first->macro_rep : NULL,
1976 .ofs = macro ? first->ofs : 0,
1977 .len = macro ? (last->ofs - first->ofs) + last->len : 0,
1978 .ref_cnt = macro ? first->ref_cnt : NULL,
1982 lex_source_push_parse (src, t);
1984 for (int i = 0; i < retval; i++)
1985 lex_stage_pop_first (&src->merge);
1992 lex_source_push_endcmd__ (struct lex_source *src)
1994 assert (src->n_parse == 0);
1996 struct lex_token *token = xmalloc (sizeof *token);
1997 *token = (struct lex_token) { .token = { .type = T_ENDCMD } };
1998 lex_source_push_parse (src, token);
2002 lex_source_push_parse (struct lex_source *src, struct lex_token *token)
2004 if (src->n_parse >= src->allocated_parse)
2005 src->parse = x2nrealloc (src->parse, &src->allocated_parse,
2006 sizeof *src->parse);
2007 src->parse[src->n_parse++] = token;
2011 lex_source_clear_parse (struct lex_source *src)
2013 for (size_t i = 0; i < src->n_parse; i++)
2014 lex_token_destroy (src->parse[i]);
2015 src->n_parse = src->parse_ofs = 0;
2018 static struct lex_source *
2019 lex_source_create (struct lexer *lexer, struct lex_reader *reader)
2021 struct lex_source *src = xmalloc (sizeof *src);
2022 *src = (struct lex_source) {
2024 .segmenter = segmenter_init (reader->syntax, false),
2028 lex_source_push_endcmd__ (src);
2034 lex_source_destroy (struct lex_source *src)
2036 char *file_name = src->reader->file_name;
2037 char *encoding = src->reader->encoding;
2038 if (src->reader->class->destroy != NULL)
2039 src->reader->class->destroy (src->reader);
2043 lex_stage_uninit (&src->pp);
2044 lex_stage_uninit (&src->merge);
2045 lex_source_clear_parse (src);
2047 ll_remove (&src->ll);
2051 struct lex_file_reader
2053 struct lex_reader reader;
2054 struct u8_istream *istream;
2057 static struct lex_reader_class lex_file_reader_class;
2059 /* Creates and returns a new lex_reader that will read from file FILE_NAME (or
2060 from stdin if FILE_NAME is "-"). The file is expected to be encoded with
2061 ENCODING, which should take one of the forms accepted by
2062 u8_istream_for_file(). SYNTAX and ERROR become the syntax mode and error
2063 mode of the new reader, respectively.
2065 Returns a null pointer if FILE_NAME cannot be opened. */
2067 lex_reader_for_file (const char *file_name, const char *encoding,
2068 enum segmenter_mode syntax,
2069 enum lex_error_mode error)
2071 struct lex_file_reader *r;
2072 struct u8_istream *istream;
2074 istream = (!strcmp(file_name, "-")
2075 ? u8_istream_for_fd (encoding, STDIN_FILENO)
2076 : u8_istream_for_file (encoding, file_name, O_RDONLY));
2077 if (istream == NULL)
2079 msg (ME, _("Opening `%s': %s."), file_name, strerror (errno));
2083 r = xmalloc (sizeof *r);
2084 lex_reader_init (&r->reader, &lex_file_reader_class);
2085 r->reader.syntax = syntax;
2086 r->reader.error = error;
2087 r->reader.file_name = xstrdup (file_name);
2088 r->reader.encoding = xstrdup_if_nonnull (encoding);
2089 r->reader.line_number = 1;
2090 r->istream = istream;
2095 static struct lex_file_reader *
2096 lex_file_reader_cast (struct lex_reader *r)
2098 return UP_CAST (r, struct lex_file_reader, reader);
2102 lex_file_read (struct lex_reader *r_, char *buf, size_t n,
2103 enum prompt_style prompt_style UNUSED)
2105 struct lex_file_reader *r = lex_file_reader_cast (r_);
2106 ssize_t n_read = u8_istream_read (r->istream, buf, n);
2109 msg (ME, _("Error reading `%s': %s."), r_->file_name, strerror (errno));
2116 lex_file_close (struct lex_reader *r_)
2118 struct lex_file_reader *r = lex_file_reader_cast (r_);
2120 if (u8_istream_fileno (r->istream) != STDIN_FILENO)
2122 if (u8_istream_close (r->istream) != 0)
2123 msg (ME, _("Error closing `%s': %s."), r_->file_name, strerror (errno));
2126 u8_istream_free (r->istream);
2131 static struct lex_reader_class lex_file_reader_class =
2137 struct lex_string_reader
2139 struct lex_reader reader;
2144 static struct lex_reader_class lex_string_reader_class;
2146 /* Creates and returns a new lex_reader for the contents of S, which must be
2147 encoded in the given ENCODING. The new reader takes ownership of S and will free it
2148 with ss_dealloc() when it is closed. */
2150 lex_reader_for_substring_nocopy (struct substring s, const char *encoding)
2152 struct lex_string_reader *r;
2154 r = xmalloc (sizeof *r);
2155 lex_reader_init (&r->reader, &lex_string_reader_class);
2156 r->reader.syntax = SEG_MODE_AUTO;
2157 r->reader.encoding = xstrdup_if_nonnull (encoding);
2164 /* Creates and returns a new lex_reader for a copy of null-terminated string S,
2165 which must be encoded in ENCODING. The caller retains ownership of S. */
2167 lex_reader_for_string (const char *s, const char *encoding)
2169 struct substring ss;
2170 ss_alloc_substring (&ss, ss_cstr (s));
2171 return lex_reader_for_substring_nocopy (ss, encoding);
2174 /* Formats FORMAT as a printf()-like format string and creates and returns a
2175 new lex_reader for the formatted result. */
2177 lex_reader_for_format (const char *format, const char *encoding, ...)
2179 struct lex_reader *r;
2182 va_start (args, encoding);
2183 r = lex_reader_for_substring_nocopy (ss_cstr (xvasprintf (format, args)), encoding);
2189 static struct lex_string_reader *
2190 lex_string_reader_cast (struct lex_reader *r)
2192 return UP_CAST (r, struct lex_string_reader, reader);
2196 lex_string_read (struct lex_reader *r_, char *buf, size_t n,
2197 enum prompt_style prompt_style UNUSED)
2199 struct lex_string_reader *r = lex_string_reader_cast (r_);
2202 chunk = MIN (n, r->s.length - r->offset);
2203 memcpy (buf, r->s.string + r->offset, chunk);
2210 lex_string_close (struct lex_reader *r_)
2212 struct lex_string_reader *r = lex_string_reader_cast (r_);
2218 static struct lex_reader_class lex_string_reader_class =