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/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/ll.h"
42 #include "libpspp/message.h"
43 #include "libpspp/misc.h"
44 #include "libpspp/str.h"
45 #include "libpspp/u8-istream.h"
46 #include "output/journal.h"
47 #include "output/output-item.h"
49 #include "gl/c-ctype.h"
50 #include "gl/minmax.h"
51 #include "gl/xalloc.h"
52 #include "gl/xmemdup0.h"
55 #define _(msgid) gettext (msgid)
56 #define N_(msgid) msgid
58 /* A token within a lex_source. */
61 /* The regular token information. */
64 /* Location of token in terms of the lex_source's buffer.
65 src->tail <= line_pos <= token_pos <= src->head. */
66 size_t token_pos; /* Start of token. */
67 size_t token_len; /* Length of source for token in bytes. */
68 size_t line_pos; /* Start of line containing token_pos. */
69 int first_line; /* Line number at token_pos. */
72 /* A source of tokens, corresponding to a syntax file.
74 This is conceptually a lex_reader wrapped with everything needed to convert
75 its UTF-8 bytes into tokens. */
78 struct ll ll; /* In lexer's list of sources. */
79 struct lex_reader *reader;
80 struct segmenter segmenter;
81 bool eof; /* True if T_STOP was read from 'reader'. */
83 /* Buffer of UTF-8 bytes. */
85 size_t allocated; /* Number of bytes allocated. */
86 size_t tail; /* &buffer[0] offset into UTF-8 source. */
87 size_t head; /* &buffer[head - tail] offset into source. */
89 /* Positions in source file, tail <= pos <= head for each member here. */
90 size_t journal_pos; /* First byte not yet output to journal. */
91 size_t seg_pos; /* First byte not yet scanned as token. */
92 size_t line_pos; /* First byte of line containing seg_pos. */
94 int n_newlines; /* Number of new-lines up to seg_pos. */
95 bool suppress_next_newline;
98 struct deque deque; /* Indexes into 'tokens'. */
99 struct lex_token *tokens; /* Lookahead tokens for parser. */
102 static struct lex_source *lex_source_create (struct lex_reader *);
103 static void lex_source_destroy (struct lex_source *);
108 struct ll_list sources; /* Contains "struct lex_source"s. */
111 static struct lex_source *lex_source__ (const struct lexer *);
112 static const struct lex_token *lex_next__ (const struct lexer *, int n);
113 static void lex_source_push_endcmd__ (struct lex_source *);
115 static void lex_source_pop__ (struct lex_source *);
116 static bool lex_source_get__ (const struct lex_source *);
117 static void lex_source_error_valist (struct lex_source *, int n0, int n1,
118 const char *format, va_list)
119 PRINTF_FORMAT (4, 0);
120 static const struct lex_token *lex_source_next__ (const struct lex_source *,
123 /* Initializes READER with the specified CLASS and otherwise some reasonable
124 defaults. The caller should fill in the others members as desired. */
126 lex_reader_init (struct lex_reader *reader,
127 const struct lex_reader_class *class)
129 reader->class = class;
130 reader->syntax = SEG_MODE_AUTO;
131 reader->error = LEX_ERROR_CONTINUE;
132 reader->file_name = NULL;
133 reader->encoding = NULL;
134 reader->line_number = 0;
138 /* Frees any file name already in READER and replaces it by a copy of
139 FILE_NAME, or if FILE_NAME is null then clears any existing name. */
141 lex_reader_set_file_name (struct lex_reader *reader, const char *file_name)
143 free (reader->file_name);
144 reader->file_name = xstrdup_if_nonnull (file_name);
147 /* Creates and returns a new lexer. */
151 struct lexer *lexer = xzalloc (sizeof *lexer);
152 ll_init (&lexer->sources);
156 /* Destroys LEXER. */
158 lex_destroy (struct lexer *lexer)
162 struct lex_source *source, *next;
164 ll_for_each_safe (source, next, struct lex_source, ll, &lexer->sources)
165 lex_source_destroy (source);
170 /* Inserts READER into LEXER so that the next token read by LEXER comes from
171 READER. Before the caller, LEXER must either be empty or at a T_ENDCMD
174 lex_include (struct lexer *lexer, struct lex_reader *reader)
176 assert (ll_is_empty (&lexer->sources) || lex_token (lexer) == T_ENDCMD);
177 ll_push_head (&lexer->sources, &lex_source_create (reader)->ll);
180 /* Appends READER to LEXER, so that it will be read after all other current
181 readers have already been read. */
183 lex_append (struct lexer *lexer, struct lex_reader *reader)
185 ll_push_tail (&lexer->sources, &lex_source_create (reader)->ll);
190 static struct lex_token *
191 lex_push_token__ (struct lex_source *src)
193 struct lex_token *token;
195 if (deque_is_full (&src->deque))
196 src->tokens = deque_expand (&src->deque, src->tokens, sizeof *src->tokens);
198 token = &src->tokens[deque_push_front (&src->deque)];
199 token_init (&token->token);
204 lex_source_pop__ (struct lex_source *src)
206 token_uninit (&src->tokens[deque_pop_back (&src->deque)].token);
210 lex_source_pop_front (struct lex_source *src)
212 token_uninit (&src->tokens[deque_pop_front (&src->deque)].token);
215 /* Advances LEXER to the next token, consuming the current token. */
217 lex_get (struct lexer *lexer)
219 struct lex_source *src;
221 src = lex_source__ (lexer);
225 if (!deque_is_empty (&src->deque))
226 lex_source_pop__ (src);
228 while (deque_is_empty (&src->deque))
229 if (!lex_source_get__ (src))
231 lex_source_destroy (src);
232 src = lex_source__ (lexer);
238 /* Issuing errors. */
240 /* Prints a syntax error message containing the current token and
241 given message MESSAGE (if non-null). */
243 lex_error (struct lexer *lexer, const char *format, ...)
247 va_start (args, format);
248 lex_next_error_valist (lexer, 0, 0, format, args);
252 /* Prints a syntax error message containing the current token and
253 given message MESSAGE (if non-null). */
255 lex_error_valist (struct lexer *lexer, const char *format, va_list args)
257 lex_next_error_valist (lexer, 0, 0, format, args);
260 /* Prints a syntax error message containing the current token and
261 given message MESSAGE (if non-null). */
263 lex_next_error (struct lexer *lexer, int n0, int n1, const char *format, ...)
267 va_start (args, format);
268 lex_next_error_valist (lexer, n0, n1, format, args);
272 /* Prints a syntax error message saying that one of the strings provided as
273 varargs, up to the first NULL, is expected. */
275 (lex_error_expecting) (struct lexer *lexer, ...)
279 va_start (args, lexer);
280 lex_error_expecting_valist (lexer, args);
284 /* Prints a syntax error message saying that one of the options provided in
285 ARGS, up to the first NULL, is expected. */
287 lex_error_expecting_valist (struct lexer *lexer, va_list args)
289 enum { MAX_OPTIONS = 9 };
290 const char *options[MAX_OPTIONS];
292 while (n < MAX_OPTIONS)
294 const char *option = va_arg (args, const char *);
298 options[n++] = option;
300 lex_error_expecting_array (lexer, options, n);
304 lex_error_expecting_array (struct lexer *lexer, const char **options, size_t n)
309 lex_error (lexer, NULL);
313 lex_error (lexer, _("expecting %s"), options[0]);
317 lex_error (lexer, _("expecting %s or %s"), options[0], options[1]);
321 lex_error (lexer, _("expecting %s, %s, or %s"), options[0], options[1],
326 lex_error (lexer, _("expecting %s, %s, %s, or %s"),
327 options[0], options[1], options[2], options[3]);
331 lex_error (lexer, _("expecting %s, %s, %s, %s, or %s"),
332 options[0], options[1], options[2], options[3], options[4]);
336 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, or %s"),
337 options[0], options[1], options[2], options[3], options[4],
342 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, %s, or %s"),
343 options[0], options[1], options[2], options[3], options[4],
344 options[5], options[6]);
348 lex_error (lexer, _("expecting %s, %s, %s, %s, %s, %s, %s, or %s"),
349 options[0], options[1], options[2], options[3], options[4],
350 options[5], options[6], options[7]);
354 lex_error (lexer, NULL);
358 /* Reports an error to the effect that subcommand SBC may only be specified
361 This function does not take a lexer as an argument or use lex_error(),
362 because the result would ordinarily just be redundant: "Syntax error at
363 SUBCOMMAND: Subcommand SUBCOMMAND may only be specified once.", which does
364 not help the user find the error. */
366 lex_sbc_only_once (const char *sbc)
368 msg (SE, _("Subcommand %s may only be specified once."), sbc);
371 /* Reports an error to the effect that subcommand SBC is missing.
373 This function does not take a lexer as an argument or use lex_error(),
374 because a missing subcommand can normally be detected only after the whole
375 command has been parsed, and so lex_error() would always report "Syntax
376 error at end of command", which does not help the user find the error. */
378 lex_sbc_missing (const char *sbc)
380 msg (SE, _("Required subcommand %s was not specified."), sbc);
383 /* Reports an error to the effect that specification SPEC may only be specified
384 once within subcommand SBC. */
386 lex_spec_only_once (struct lexer *lexer, const char *sbc, const char *spec)
388 lex_error (lexer, _("%s may only be specified once within subcommand %s"),
392 /* Reports an error to the effect that specification SPEC is missing within
395 lex_spec_missing (struct lexer *lexer, const char *sbc, const char *spec)
397 lex_error (lexer, _("Required %s specification missing from %s subcommand"),
401 /* Prints a syntax error message containing the current token and
402 given message MESSAGE (if non-null). */
404 lex_next_error_valist (struct lexer *lexer, int n0, int n1,
405 const char *format, va_list args)
407 struct lex_source *src = lex_source__ (lexer);
410 lex_source_error_valist (src, n0, n1, format, args);
416 ds_put_format (&s, _("Syntax error at end of input"));
419 ds_put_cstr (&s, ": ");
420 ds_put_vformat (&s, format, args);
422 ds_put_byte (&s, '.');
423 msg (SE, "%s", ds_cstr (&s));
428 /* Checks that we're at end of command.
429 If so, returns a successful command completion code.
430 If not, flags a syntax error and returns an error command
433 lex_end_of_command (struct lexer *lexer)
435 if (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_STOP)
437 lex_error (lexer, _("expecting end of command"));
444 /* Token testing functions. */
446 /* Returns true if the current token is a number. */
448 lex_is_number (const struct lexer *lexer)
450 return lex_next_is_number (lexer, 0);
453 /* Returns true if the current token is a string. */
455 lex_is_string (const struct lexer *lexer)
457 return lex_next_is_string (lexer, 0);
460 /* Returns the value of the current token, which must be a
461 floating point number. */
463 lex_number (const struct lexer *lexer)
465 return lex_next_number (lexer, 0);
468 /* Returns true iff the current token is an integer. */
470 lex_is_integer (const struct lexer *lexer)
472 return lex_next_is_integer (lexer, 0);
475 /* Returns the value of the current token, which must be an
478 lex_integer (const struct lexer *lexer)
480 return lex_next_integer (lexer, 0);
483 /* Token testing functions with lookahead.
485 A value of 0 for N as an argument to any of these functions refers to the
486 current token. Lookahead is limited to the current command. Any N greater
487 than the number of tokens remaining in the current command will be treated
488 as referring to a T_ENDCMD token. */
490 /* Returns true if the token N ahead of the current token is a number. */
492 lex_next_is_number (const struct lexer *lexer, int n)
494 enum token_type next_token = lex_next_token (lexer, n);
495 return next_token == T_POS_NUM || next_token == T_NEG_NUM;
498 /* Returns true if the token N ahead of the current token is a string. */
500 lex_next_is_string (const struct lexer *lexer, int n)
502 return lex_next_token (lexer, n) == T_STRING;
505 /* Returns the value of the token N ahead of the current token, which must be a
506 floating point number. */
508 lex_next_number (const struct lexer *lexer, int n)
510 assert (lex_next_is_number (lexer, n));
511 return lex_next_tokval (lexer, n);
514 /* Returns true if the token N ahead of the current token is an integer. */
516 lex_next_is_integer (const struct lexer *lexer, int n)
520 if (!lex_next_is_number (lexer, n))
523 value = lex_next_tokval (lexer, n);
524 return value > LONG_MIN && value <= LONG_MAX && floor (value) == value;
527 /* Returns the value of the token N ahead of the current token, which must be
530 lex_next_integer (const struct lexer *lexer, int n)
532 assert (lex_next_is_integer (lexer, n));
533 return lex_next_tokval (lexer, n);
536 /* Token matching functions. */
538 /* If the current token has the specified TYPE, skips it and returns true.
539 Otherwise, returns false. */
541 lex_match (struct lexer *lexer, enum token_type type)
543 if (lex_token (lexer) == type)
552 /* If the current token matches IDENTIFIER, skips it and returns true.
553 IDENTIFIER may be abbreviated to its first three letters. Otherwise,
556 IDENTIFIER must be an ASCII string. */
558 lex_match_id (struct lexer *lexer, const char *identifier)
560 return lex_match_id_n (lexer, identifier, 3);
563 /* If the current token is IDENTIFIER, skips it and returns true. IDENTIFIER
564 may be abbreviated to its first N letters. Otherwise, returns false.
566 IDENTIFIER must be an ASCII string. */
568 lex_match_id_n (struct lexer *lexer, const char *identifier, size_t n)
570 if (lex_token (lexer) == T_ID
571 && lex_id_match_n (ss_cstr (identifier), lex_tokss (lexer), n))
580 /* If the current token is integer X, skips it and returns true. Otherwise,
583 lex_match_int (struct lexer *lexer, int x)
585 if (lex_is_integer (lexer) && lex_integer (lexer) == x)
594 /* Forced matches. */
596 /* If this token is IDENTIFIER, skips it and returns true. IDENTIFIER may be
597 abbreviated to its first 3 letters. Otherwise, reports an error and returns
600 IDENTIFIER must be an ASCII string. */
602 lex_force_match_id (struct lexer *lexer, const char *identifier)
604 if (lex_match_id (lexer, identifier))
608 lex_error_expecting (lexer, identifier);
613 /* If the current token has the specified TYPE, skips it and returns true.
614 Otherwise, reports an error and returns false. */
616 lex_force_match (struct lexer *lexer, enum token_type type)
618 if (lex_token (lexer) == type)
625 const char *type_string = token_type_to_string (type);
628 char *s = xasprintf ("`%s'", type_string);
629 lex_error_expecting (lexer, s);
633 lex_error_expecting (lexer, token_type_to_name (type));
639 /* If the current token is a string, does nothing and returns true.
640 Otherwise, reports an error and returns false. */
642 lex_force_string (struct lexer *lexer)
644 if (lex_is_string (lexer))
648 lex_error (lexer, _("expecting string"));
653 /* If the current token is a string or an identifier, does nothing and returns
654 true. Otherwise, reports an error and returns false.
656 This is meant for use in syntactic situations where we want to encourage the
657 user to supply a quoted string, but for compatibility we also accept
658 identifiers. (One example of such a situation is file names.) Therefore,
659 the error message issued when the current token is wrong only says that a
660 string is expected and doesn't mention that an identifier would also be
663 lex_force_string_or_id (struct lexer *lexer)
665 return lex_token (lexer) == T_ID || lex_force_string (lexer);
668 /* If the current token is an integer, does nothing and returns true.
669 Otherwise, reports an error and returns false. */
671 lex_force_int (struct lexer *lexer)
673 if (lex_is_integer (lexer))
677 lex_error (lexer, _("expecting integer"));
682 /* If the current token is an integer in the range MIN...MAX (inclusive), does
683 nothing and returns true. Otherwise, reports an error and returns false.
684 If NAME is nonnull, then it is used in the error message. */
686 lex_force_int_range (struct lexer *lexer, const char *name, long min, long max)
688 bool is_integer = lex_is_integer (lexer);
689 bool too_small = is_integer && lex_integer (lexer) < min;
690 bool too_big = is_integer && lex_integer (lexer) > max;
691 if (is_integer && !too_small && !too_big)
696 /* Weird, maybe a bug in the caller. Just report that we needed an
699 lex_error (lexer, _("Integer expected for %s."), name);
701 lex_error (lexer, _("Integer expected."));
706 lex_error (lexer, _("Expected %ld for %s."), min, name);
708 lex_error (lexer, _("Expected %ld."), min);
710 else if (min + 1 == max)
713 lex_error (lexer, _("Expected %ld or %ld for %s."), min, min + 1, name);
715 lex_error (lexer, _("Expected %ld or %ld."), min, min + 1);
719 bool report_lower_bound = (min > INT_MIN / 2) || too_small;
720 bool report_upper_bound = (max < INT_MAX / 2) || too_big;
722 if (report_lower_bound && report_upper_bound)
726 _("Expected integer between %ld and %ld for %s."),
729 lex_error (lexer, _("Expected integer between %ld and %ld."),
732 else if (report_lower_bound)
737 lex_error (lexer, _("Expected non-negative integer for %s."),
740 lex_error (lexer, _("Expected non-negative integer."));
745 lex_error (lexer, _("Expected positive integer for %s."),
748 lex_error (lexer, _("Expected positive integer."));
751 else if (report_upper_bound)
755 _("Expected integer less than or equal to %ld for %s."),
758 lex_error (lexer, _("Expected integer less than or equal to %ld."),
764 lex_error (lexer, _("Integer expected for %s."), name);
766 lex_error (lexer, _("Integer expected."));
772 /* If the current token is a number, does nothing and returns true.
773 Otherwise, reports an error and returns false. */
775 lex_force_num (struct lexer *lexer)
777 if (lex_is_number (lexer))
780 lex_error (lexer, _("expecting number"));
784 /* If the current token is an identifier, does nothing and returns true.
785 Otherwise, reports an error and returns false. */
787 lex_force_id (struct lexer *lexer)
789 if (lex_token (lexer) == T_ID)
792 lex_error (lexer, _("expecting identifier"));
796 /* Token accessors. */
798 /* Returns the type of LEXER's current token. */
800 lex_token (const struct lexer *lexer)
802 return lex_next_token (lexer, 0);
805 /* Returns the number in LEXER's current token.
807 Only T_NEG_NUM and T_POS_NUM tokens have meaningful values. For other
808 tokens this function will always return zero. */
810 lex_tokval (const struct lexer *lexer)
812 return lex_next_tokval (lexer, 0);
815 /* Returns the null-terminated string in LEXER's current token, UTF-8 encoded.
817 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
818 this functions this function will always return NULL.
820 The UTF-8 encoding of the returned string is correct for variable names and
821 other identifiers. Use filename_to_utf8() to use it as a filename. Use
822 data_in() to use it in a "union value". */
824 lex_tokcstr (const struct lexer *lexer)
826 return lex_next_tokcstr (lexer, 0);
829 /* Returns the string in LEXER's current token, UTF-8 encoded. The string is
830 null-terminated (but the null terminator is not included in the returned
831 substring's 'length').
833 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
834 this functions this function will always return NULL.
836 The UTF-8 encoding of the returned string is correct for variable names and
837 other identifiers. Use filename_to_utf8() to use it as a filename. Use
838 data_in() to use it in a "union value". */
840 lex_tokss (const struct lexer *lexer)
842 return lex_next_tokss (lexer, 0);
847 A value of 0 for N as an argument to any of these functions refers to the
848 current token. Lookahead is limited to the current command. Any N greater
849 than the number of tokens remaining in the current command will be treated
850 as referring to a T_ENDCMD token. */
852 static const struct lex_token *
853 lex_next__ (const struct lexer *lexer_, int n)
855 struct lexer *lexer = CONST_CAST (struct lexer *, lexer_);
856 struct lex_source *src = lex_source__ (lexer);
859 return lex_source_next__ (src, n);
862 static const struct lex_token stop_token =
863 { TOKEN_INITIALIZER (T_STOP, 0.0, ""), 0, 0, 0, 0 };
869 static const struct lex_token *
870 lex_source_next__ (const struct lex_source *src, int n)
872 while (deque_count (&src->deque) <= n)
874 if (!deque_is_empty (&src->deque))
876 struct lex_token *front;
878 front = &src->tokens[deque_front (&src->deque, 0)];
879 if (front->token.type == T_STOP || front->token.type == T_ENDCMD)
883 lex_source_get__ (src);
886 return &src->tokens[deque_back (&src->deque, n)];
889 /* Returns the "struct token" of the token N after the current one in LEXER.
890 The returned pointer can be invalidated by pretty much any succeeding call
891 into the lexer, although the string pointer within the returned token is
892 only invalidated by consuming the token (e.g. with lex_get()). */
894 lex_next (const struct lexer *lexer, int n)
896 return &lex_next__ (lexer, n)->token;
899 /* Returns the type of the token N after the current one in LEXER. */
901 lex_next_token (const struct lexer *lexer, int n)
903 return lex_next (lexer, n)->type;
906 /* Returns the number in the tokn N after the current one in LEXER.
908 Only T_NEG_NUM and T_POS_NUM tokens have meaningful values. For other
909 tokens this function will always return zero. */
911 lex_next_tokval (const struct lexer *lexer, int n)
913 const struct token *token = lex_next (lexer, n);
914 return token->number;
917 /* Returns the null-terminated string in the token N after the current one, in
920 Only T_ID and T_STRING tokens have meaningful strings. For other tokens
921 this functions this function will always return NULL.
923 The UTF-8 encoding of the returned string is correct for variable names and
924 other identifiers. Use filename_to_utf8() to use it as a filename. Use
925 data_in() to use it in a "union value". */
927 lex_next_tokcstr (const struct lexer *lexer, int n)
929 return lex_next_tokss (lexer, n).string;
932 /* Returns the string in the token N after the current one, in UTF-8 encoding.
933 The string is null-terminated (but the null terminator is not included in
934 the returned substring's 'length').
936 Only T_ID, T_MACRO_ID, T_STRING tokens have meaningful strings. For other
937 tokens this functions this function will always return NULL.
939 The UTF-8 encoding of the returned string is correct for variable names and
940 other identifiers. Use filename_to_utf8() to use it as a filename. Use
941 data_in() to use it in a "union value". */
943 lex_next_tokss (const struct lexer *lexer, int n)
945 return lex_next (lexer, n)->string;
949 lex_tokens_match (const struct token *actual, const struct token *expected)
951 if (actual->type != expected->type)
954 switch (actual->type)
958 return actual->number == expected->number;
961 return lex_id_match (expected->string, actual->string);
964 return (actual->string.length == expected->string.length
965 && !memcmp (actual->string.string, expected->string.string,
966 actual->string.length));
973 /* If LEXER is positioned at the sequence of tokens that may be parsed from S,
974 skips it and returns true. Otherwise, returns false.
976 S may consist of an arbitrary sequence of tokens, e.g. "KRUSKAL-WALLIS",
977 "2SLS", or "END INPUT PROGRAM". Identifiers may be abbreviated to their
978 first three letters. */
980 lex_match_phrase (struct lexer *lexer, const char *s)
982 struct string_lexer slex;
987 string_lexer_init (&slex, s, strlen (s), SEG_MODE_INTERACTIVE);
988 while (string_lexer_next (&slex, &token))
989 if (token.type != SCAN_SKIP)
991 bool match = lex_tokens_match (lex_next (lexer, i++), &token);
992 token_uninit (&token);
1003 lex_source_get_first_line_number (const struct lex_source *src, int n)
1005 return lex_source_next__ (src, n)->first_line;
1009 count_newlines (char *s, size_t length)
1014 while ((newline = memchr (s, '\n', length)) != NULL)
1017 length -= (newline + 1) - s;
1025 lex_source_get_last_line_number (const struct lex_source *src, int n)
1027 const struct lex_token *token = lex_source_next__ (src, n);
1029 if (token->first_line == 0)
1033 char *token_str = &src->buffer[token->token_pos - src->tail];
1034 return token->first_line + count_newlines (token_str, token->token_len) + 1;
1039 count_columns (const char *s_, size_t length)
1041 const uint8_t *s = CHAR_CAST (const uint8_t *, s_);
1047 for (ofs = 0; ofs < length; ofs += mblen)
1051 mblen = u8_mbtouc (&uc, s + ofs, length - ofs);
1054 int width = uc_width (uc, "UTF-8");
1059 columns = ROUND_UP (columns + 1, 8);
1066 lex_source_get_first_column (const struct lex_source *src, int n)
1068 const struct lex_token *token = lex_source_next__ (src, n);
1069 return count_columns (&src->buffer[token->line_pos - src->tail],
1070 token->token_pos - token->line_pos);
1074 lex_source_get_last_column (const struct lex_source *src, int n)
1076 const struct lex_token *token = lex_source_next__ (src, n);
1077 char *start, *end, *newline;
1079 start = &src->buffer[token->line_pos - src->tail];
1080 end = &src->buffer[(token->token_pos + token->token_len) - src->tail];
1081 newline = memrchr (start, '\n', end - start);
1082 if (newline != NULL)
1083 start = newline + 1;
1084 return count_columns (start, end - start);
1087 /* Returns the 1-based line number of the start of the syntax that represents
1088 the token N after the current one in LEXER. Returns 0 for a T_STOP token or
1089 if the token is drawn from a source that does not have line numbers. */
1091 lex_get_first_line_number (const struct lexer *lexer, int n)
1093 const struct lex_source *src = lex_source__ (lexer);
1094 return src != NULL ? lex_source_get_first_line_number (src, n) : 0;
1097 /* Returns the 1-based line number of the end of the syntax that represents the
1098 token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
1099 token or if the token is drawn from a source that does not have line
1102 Most of the time, a single token is wholly within a single line of syntax,
1103 but there are two exceptions: a T_STRING token can be made up of multiple
1104 segments on adjacent lines connected with "+" punctuators, and a T_NEG_NUM
1105 token can consist of a "-" on one line followed by the number on the next.
1108 lex_get_last_line_number (const struct lexer *lexer, int n)
1110 const struct lex_source *src = lex_source__ (lexer);
1111 return src != NULL ? lex_source_get_last_line_number (src, n) : 0;
1114 /* Returns the 1-based column number of the start of the syntax that represents
1115 the token N after the current one in LEXER. Returns 0 for a T_STOP
1118 Column numbers are measured according to the width of characters as shown in
1119 a typical fixed-width font, in which CJK characters have width 2 and
1120 combining characters have width 0. */
1122 lex_get_first_column (const struct lexer *lexer, int n)
1124 const struct lex_source *src = lex_source__ (lexer);
1125 return src != NULL ? lex_source_get_first_column (src, n) : 0;
1128 /* Returns the 1-based column number of the end of the syntax that represents
1129 the token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
1132 Column numbers are measured according to the width of characters as shown in
1133 a typical fixed-width font, in which CJK characters have width 2 and
1134 combining characters have width 0. */
1136 lex_get_last_column (const struct lexer *lexer, int n)
1138 const struct lex_source *src = lex_source__ (lexer);
1139 return src != NULL ? lex_source_get_last_column (src, n) : 0;
1142 /* Returns the name of the syntax file from which the current command is drawn.
1143 Returns NULL for a T_STOP token or if the command's source does not have
1146 There is no version of this function that takes an N argument because
1147 lookahead only works to the end of a command and any given command is always
1148 within a single syntax file. */
1150 lex_get_file_name (const struct lexer *lexer)
1152 struct lex_source *src = lex_source__ (lexer);
1153 return src == NULL ? NULL : src->reader->file_name;
1157 lex_get_encoding (const struct lexer *lexer)
1159 struct lex_source *src = lex_source__ (lexer);
1160 return src == NULL ? NULL : src->reader->encoding;
1164 /* Returns the syntax mode for the syntax file from which the current drawn is
1165 drawn. Returns SEG_MODE_AUTO for a T_STOP token or if the command's source
1166 does not have line numbers.
1168 There is no version of this function that takes an N argument because
1169 lookahead only works to the end of a command and any given command is always
1170 within a single syntax file. */
1172 lex_get_syntax_mode (const struct lexer *lexer)
1174 struct lex_source *src = lex_source__ (lexer);
1175 return src == NULL ? SEG_MODE_AUTO : src->reader->syntax;
1178 /* Returns the error mode for the syntax file from which the current drawn is
1179 drawn. Returns LEX_ERROR_TERMINAL for a T_STOP token or if the command's
1180 source does not have line numbers.
1182 There is no version of this function that takes an N argument because
1183 lookahead only works to the end of a command and any given command is always
1184 within a single syntax file. */
1186 lex_get_error_mode (const struct lexer *lexer)
1188 struct lex_source *src = lex_source__ (lexer);
1189 return src == NULL ? LEX_ERROR_TERMINAL : src->reader->error;
1192 /* If the source that LEXER is currently reading has error mode
1193 LEX_ERROR_TERMINAL, discards all buffered input and tokens, so that the next
1194 token to be read comes directly from whatever is next read from the stream.
1196 It makes sense to call this function after encountering an error in a
1197 command entered on the console, because usually the user would prefer not to
1198 have cascading errors. */
1200 lex_interactive_reset (struct lexer *lexer)
1202 struct lex_source *src = lex_source__ (lexer);
1203 if (src != NULL && src->reader->error == LEX_ERROR_TERMINAL)
1205 src->head = src->tail = 0;
1206 src->journal_pos = src->seg_pos = src->line_pos = 0;
1207 src->n_newlines = 0;
1208 src->suppress_next_newline = false;
1209 segmenter_init (&src->segmenter, segmenter_get_mode (&src->segmenter));
1210 while (!deque_is_empty (&src->deque))
1211 lex_source_pop__ (src);
1212 lex_source_push_endcmd__ (src);
1216 /* Advances past any tokens in LEXER up to a T_ENDCMD or T_STOP. */
1218 lex_discard_rest_of_command (struct lexer *lexer)
1220 while (lex_token (lexer) != T_STOP && lex_token (lexer) != T_ENDCMD)
1224 /* Discards all lookahead tokens in LEXER, then discards all input sources
1225 until it encounters one with error mode LEX_ERROR_TERMINAL or until it
1226 runs out of input sources. */
1228 lex_discard_noninteractive (struct lexer *lexer)
1230 struct lex_source *src = lex_source__ (lexer);
1234 while (!deque_is_empty (&src->deque))
1235 lex_source_pop__ (src);
1237 for (; src != NULL && src->reader->error != LEX_ERROR_TERMINAL;
1238 src = lex_source__ (lexer))
1239 lex_source_destroy (src);
1244 lex_source_max_tail__ (const struct lex_source *src)
1246 const struct lex_token *token;
1249 assert (src->seg_pos >= src->line_pos);
1250 max_tail = MIN (src->journal_pos, src->line_pos);
1252 /* Use the oldest token also. (We know that src->deque cannot be empty
1253 because we are in the process of adding a new token, which is already
1254 initialized enough to use here.) */
1255 token = &src->tokens[deque_back (&src->deque, 0)];
1256 assert (token->token_pos >= token->line_pos);
1257 max_tail = MIN (max_tail, token->line_pos);
1263 lex_source_expand__ (struct lex_source *src)
1265 if (src->head - src->tail >= src->allocated)
1267 size_t max_tail = lex_source_max_tail__ (src);
1268 if (max_tail > src->tail)
1270 /* Advance the tail, freeing up room at the head. */
1271 memmove (src->buffer, src->buffer + (max_tail - src->tail),
1272 src->head - max_tail);
1273 src->tail = max_tail;
1277 /* Buffer is completely full. Expand it. */
1278 src->buffer = x2realloc (src->buffer, &src->allocated);
1283 /* There's space available at the head of the buffer. Nothing to do. */
1288 lex_source_read__ (struct lex_source *src)
1292 lex_source_expand__ (src);
1294 size_t head_ofs = src->head - src->tail;
1295 size_t space = src->allocated - head_ofs;
1296 enum prompt_style prompt = segmenter_get_prompt (&src->segmenter);
1297 size_t n = src->reader->class->read (src->reader, &src->buffer[head_ofs],
1299 assert (n <= space);
1304 src->reader->eof = true;
1305 lex_source_expand__ (src);
1311 while (!memchr (&src->buffer[src->seg_pos - src->tail], '\n',
1312 src->head - src->seg_pos));
1315 static struct lex_source *
1316 lex_source__ (const struct lexer *lexer)
1318 return (ll_is_empty (&lexer->sources) ? NULL
1319 : ll_data (ll_head (&lexer->sources), struct lex_source, ll));
1322 static struct substring
1323 lex_source_get_syntax__ (const struct lex_source *src, int n0, int n1)
1325 const struct lex_token *token0 = lex_source_next__ (src, n0);
1326 const struct lex_token *token1 = lex_source_next__ (src, MAX (n0, n1));
1327 size_t start = token0->token_pos;
1328 size_t end = token1->token_pos + token1->token_len;
1330 return ss_buffer (&src->buffer[start - src->tail], end - start);
1334 lex_ellipsize__ (struct substring in, char *out, size_t out_size)
1340 assert (out_size >= 16);
1341 out_maxlen = out_size - 1;
1342 if (in.length > out_maxlen - 3)
1345 for (out_len = 0; out_len < in.length; out_len += mblen)
1347 if (in.string[out_len] == '\n'
1348 || in.string[out_len] == '\0'
1349 || (in.string[out_len] == '\r'
1350 && out_len + 1 < in.length
1351 && in.string[out_len + 1] == '\n'))
1354 mblen = u8_mblen (CHAR_CAST (const uint8_t *, in.string + out_len),
1355 in.length - out_len);
1360 if (out_len + mblen > out_maxlen)
1364 memcpy (out, in.string, out_len);
1365 strcpy (&out[out_len], out_len < in.length ? "..." : "");
1369 lex_source_error_valist (struct lex_source *src, int n0, int n1,
1370 const char *format, va_list args)
1372 const struct lex_token *token;
1377 token = lex_source_next__ (src, n0);
1378 if (token->token.type == T_ENDCMD)
1379 ds_put_cstr (&s, _("Syntax error at end of command"));
1382 struct substring syntax = lex_source_get_syntax__ (src, n0, n1);
1383 if (!ss_is_empty (syntax))
1385 char syntax_cstr[64];
1387 lex_ellipsize__ (syntax, syntax_cstr, sizeof syntax_cstr);
1388 ds_put_format (&s, _("Syntax error at `%s'"), syntax_cstr);
1391 ds_put_cstr (&s, _("Syntax error"));
1396 ds_put_cstr (&s, ": ");
1397 ds_put_vformat (&s, format, args);
1399 if (ds_last (&s) != '.')
1400 ds_put_byte (&s, '.');
1403 .category = MSG_C_SYNTAX,
1404 .severity = MSG_S_ERROR,
1405 .file_name = src->reader->file_name,
1406 .first_line = lex_source_get_first_line_number (src, n0),
1407 .last_line = lex_source_get_last_line_number (src, n1),
1408 .first_column = lex_source_get_first_column (src, n0),
1409 .last_column = lex_source_get_last_column (src, n1),
1410 .text = ds_steal_cstr (&s),
1415 static void PRINTF_FORMAT (2, 3)
1416 lex_get_error (struct lex_source *src, const char *format, ...)
1421 va_start (args, format);
1423 n = deque_count (&src->deque) - 1;
1424 lex_source_error_valist (src, n, n, format, args);
1425 lex_source_pop_front (src);
1430 /* Attempts to append an additional token into SRC's deque, reading more from
1431 the underlying lex_reader if necessary. Returns true if successful, false
1432 if the deque already represents (a suffix of) the whole lex_reader's
1435 lex_source_get__ (const struct lex_source *src_)
1437 struct lex_source *src = CONST_CAST (struct lex_source *, src_);
1441 /* State maintained while scanning tokens. Usually we only need a single
1442 state, but scanner_push() can return SCAN_SAVE to indicate that the state
1443 needs to be saved and possibly restored later with SCAN_BACK. */
1446 struct segmenter segmenter;
1447 enum segment_type last_segment;
1448 int newlines; /* Number of newlines encountered so far. */
1449 /* Maintained here so we can update lex_source's similar members when we
1455 /* Initialize state. */
1456 struct state state =
1458 .segmenter = src->segmenter,
1460 .seg_pos = src->seg_pos,
1461 .line_pos = src->line_pos,
1463 struct state saved = state;
1465 /* Append a new token to SRC and initialize it. */
1466 struct lex_token *token = lex_push_token__ (src);
1467 struct scanner scanner;
1468 scanner_init (&scanner, &token->token);
1469 token->line_pos = src->line_pos;
1470 token->token_pos = src->seg_pos;
1471 if (src->reader->line_number > 0)
1472 token->first_line = src->reader->line_number + src->n_newlines;
1474 token->first_line = 0;
1476 /* Extract segments and pass them through the scanner until we obtain a
1480 /* Extract a segment. */
1481 const char *segment = &src->buffer[state.seg_pos - src->tail];
1482 size_t seg_maxlen = src->head - state.seg_pos;
1483 enum segment_type type;
1484 int seg_len = segmenter_push (&state.segmenter, segment, seg_maxlen,
1485 src->reader->eof, &type);
1488 /* The segmenter needs more input to produce a segment. */
1489 assert (!src->reader->eof);
1490 lex_source_read__ (src);
1494 /* Update state based on the segment. */
1495 state.last_segment = type;
1496 state.seg_pos += seg_len;
1497 if (type == SEG_NEWLINE)
1500 state.line_pos = state.seg_pos;
1503 /* Pass the segment into the scanner and try to get a token out. */
1504 enum scan_result result = scanner_push (&scanner, type,
1505 ss_buffer (segment, seg_len),
1507 if (result == SCAN_SAVE)
1509 else if (result == SCAN_BACK)
1514 else if (result == SCAN_DONE)
1518 /* If we've reached the end of a line, or the end of a command, then pass
1519 the line to the output engine as a syntax text item. */
1520 int n_lines = state.newlines;
1521 if (state.last_segment == SEG_END_COMMAND && !src->suppress_next_newline)
1524 src->suppress_next_newline = true;
1526 else if (n_lines > 0 && src->suppress_next_newline)
1529 src->suppress_next_newline = false;
1531 for (int i = 0; i < n_lines; i++)
1533 /* Beginning of line. */
1534 const char *line = &src->buffer[src->journal_pos - src->tail];
1536 /* Calculate line length, including \n or \r\n end-of-line if present.
1538 We use src->head even though that may be beyond what we've actually
1539 converted to tokens (which is only through state.line_pos). That's
1540 because, if we're emitting the line due to SEG_END_COMMAND, we want to
1541 take the whole line through the newline, not just through the '.'. */
1542 size_t max_len = src->head - src->journal_pos;
1543 const char *newline = memchr (line, '\n', max_len);
1544 size_t line_len = newline ? newline - line + 1 : max_len;
1546 /* Calculate line length excluding end-of-line. */
1547 size_t copy_len = line_len;
1548 if (copy_len > 0 && line[copy_len - 1] == '\n')
1550 if (copy_len > 0 && line[copy_len - 1] == '\r')
1553 /* Submit the line as syntax. */
1554 output_item_submit (text_item_create_nocopy (TEXT_ITEM_SYNTAX,
1555 xmemdup0 (line, copy_len),
1558 src->journal_pos += line_len;
1561 token->token_len = state.seg_pos - src->seg_pos;
1563 src->segmenter = state.segmenter;
1564 src->seg_pos = state.seg_pos;
1565 src->line_pos = state.line_pos;
1566 src->n_newlines += state.newlines;
1568 switch (token->token.type)
1574 token->token.type = T_ENDCMD;
1578 case SCAN_BAD_HEX_LENGTH:
1579 lex_get_error (src, _("String of hex digits has %d characters, which "
1580 "is not a multiple of 2"),
1581 (int) token->token.number);
1584 case SCAN_BAD_HEX_DIGIT:
1585 case SCAN_BAD_UNICODE_DIGIT:
1586 lex_get_error (src, _("`%c' is not a valid hex digit"),
1587 (int) token->token.number);
1590 case SCAN_BAD_UNICODE_LENGTH:
1591 lex_get_error (src, _("Unicode string contains %d bytes, which is "
1592 "not in the valid range of 1 to 8 bytes"),
1593 (int) token->token.number);
1596 case SCAN_BAD_UNICODE_CODE_POINT:
1597 lex_get_error (src, _("U+%04X is not a valid Unicode code point"),
1598 (int) token->token.number);
1601 case SCAN_EXPECTED_QUOTE:
1602 lex_get_error (src, _("Unterminated string constant"));
1605 case SCAN_EXPECTED_EXPONENT:
1606 lex_get_error (src, _("Missing exponent following `%s'"),
1607 token->token.string.string);
1610 case SCAN_UNEXPECTED_CHAR:
1613 lex_get_error (src, _("Bad character %s in input"),
1614 uc_name (token->token.number, c_name));
1619 lex_source_pop_front (src);
1627 lex_source_push_endcmd__ (struct lex_source *src)
1629 struct lex_token *token = lex_push_token__ (src);
1630 token->token.type = T_ENDCMD;
1631 token->token_pos = 0;
1632 token->token_len = 0;
1633 token->line_pos = 0;
1634 token->first_line = 0;
1637 static struct lex_source *
1638 lex_source_create (struct lex_reader *reader)
1640 struct lex_source *src;
1642 src = xzalloc (sizeof *src);
1643 src->reader = reader;
1644 segmenter_init (&src->segmenter, reader->syntax);
1645 src->tokens = deque_init (&src->deque, 4, sizeof *src->tokens);
1647 lex_source_push_endcmd__ (src);
1653 lex_source_destroy (struct lex_source *src)
1655 char *file_name = src->reader->file_name;
1656 char *encoding = src->reader->encoding;
1657 if (src->reader->class->destroy != NULL)
1658 src->reader->class->destroy (src->reader);
1662 while (!deque_is_empty (&src->deque))
1663 lex_source_pop__ (src);
1665 ll_remove (&src->ll);
1669 struct lex_file_reader
1671 struct lex_reader reader;
1672 struct u8_istream *istream;
1675 static struct lex_reader_class lex_file_reader_class;
1677 /* Creates and returns a new lex_reader that will read from file FILE_NAME (or
1678 from stdin if FILE_NAME is "-"). The file is expected to be encoded with
1679 ENCODING, which should take one of the forms accepted by
1680 u8_istream_for_file(). SYNTAX and ERROR become the syntax mode and error
1681 mode of the new reader, respectively.
1683 Returns a null pointer if FILE_NAME cannot be opened. */
1685 lex_reader_for_file (const char *file_name, const char *encoding,
1686 enum segmenter_mode syntax,
1687 enum lex_error_mode error)
1689 struct lex_file_reader *r;
1690 struct u8_istream *istream;
1692 istream = (!strcmp(file_name, "-")
1693 ? u8_istream_for_fd (encoding, STDIN_FILENO)
1694 : u8_istream_for_file (encoding, file_name, O_RDONLY));
1695 if (istream == NULL)
1697 msg (ME, _("Opening `%s': %s."), file_name, strerror (errno));
1701 r = xmalloc (sizeof *r);
1702 lex_reader_init (&r->reader, &lex_file_reader_class);
1703 r->reader.syntax = syntax;
1704 r->reader.error = error;
1705 r->reader.file_name = xstrdup (file_name);
1706 r->reader.encoding = xstrdup_if_nonnull (encoding);
1707 r->reader.line_number = 1;
1708 r->istream = istream;
1713 static struct lex_file_reader *
1714 lex_file_reader_cast (struct lex_reader *r)
1716 return UP_CAST (r, struct lex_file_reader, reader);
1720 lex_file_read (struct lex_reader *r_, char *buf, size_t n,
1721 enum prompt_style prompt_style UNUSED)
1723 struct lex_file_reader *r = lex_file_reader_cast (r_);
1724 ssize_t n_read = u8_istream_read (r->istream, buf, n);
1727 msg (ME, _("Error reading `%s': %s."), r_->file_name, strerror (errno));
1734 lex_file_close (struct lex_reader *r_)
1736 struct lex_file_reader *r = lex_file_reader_cast (r_);
1738 if (u8_istream_fileno (r->istream) != STDIN_FILENO)
1740 if (u8_istream_close (r->istream) != 0)
1741 msg (ME, _("Error closing `%s': %s."), r_->file_name, strerror (errno));
1744 u8_istream_free (r->istream);
1749 static struct lex_reader_class lex_file_reader_class =
1755 struct lex_string_reader
1757 struct lex_reader reader;
1762 static struct lex_reader_class lex_string_reader_class;
1764 /* Creates and returns a new lex_reader for the contents of S, which must be
1765 encoded in the given ENCODING. The new reader takes ownership of S and will free it
1766 with ss_dealloc() when it is closed. */
1768 lex_reader_for_substring_nocopy (struct substring s, const char *encoding)
1770 struct lex_string_reader *r;
1772 r = xmalloc (sizeof *r);
1773 lex_reader_init (&r->reader, &lex_string_reader_class);
1774 r->reader.syntax = SEG_MODE_AUTO;
1775 r->reader.encoding = xstrdup_if_nonnull (encoding);
1782 /* Creates and returns a new lex_reader for a copy of null-terminated string S,
1783 which must be encoded in ENCODING. The caller retains ownership of S. */
1785 lex_reader_for_string (const char *s, const char *encoding)
1787 struct substring ss;
1788 ss_alloc_substring (&ss, ss_cstr (s));
1789 return lex_reader_for_substring_nocopy (ss, encoding);
1792 /* Formats FORMAT as a printf()-like format string and creates and returns a
1793 new lex_reader for the formatted result. */
1795 lex_reader_for_format (const char *format, const char *encoding, ...)
1797 struct lex_reader *r;
1800 va_start (args, encoding);
1801 r = lex_reader_for_substring_nocopy (ss_cstr (xvasprintf (format, args)), encoding);
1807 static struct lex_string_reader *
1808 lex_string_reader_cast (struct lex_reader *r)
1810 return UP_CAST (r, struct lex_string_reader, reader);
1814 lex_string_read (struct lex_reader *r_, char *buf, size_t n,
1815 enum prompt_style prompt_style UNUSED)
1817 struct lex_string_reader *r = lex_string_reader_cast (r_);
1820 chunk = MIN (n, r->s.length - r->offset);
1821 memcpy (buf, r->s.string + r->offset, chunk);
1828 lex_string_close (struct lex_reader *r_)
1830 struct lex_string_reader *r = lex_string_reader_cast (r_);
1836 static struct lex_reader_class lex_string_reader_class =