1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000 Free Software Foundation, Inc.
3 Written by Ben Pfaff <blp@gnu.org>.
5 This program is free software; you can redistribute it and/or
6 modify it under the terms of the GNU General Public License as
7 published by the Free Software Foundation; either version 2 of the
8 License, or (at your option) any later version.
10 This program is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 #include <libpspp/message.h>
29 #include <libpspp/alloc.h>
30 #include <libpspp/assertion.h>
31 #include <language/command.h>
32 #include <libpspp/message.h>
33 #include <libpspp/magic.h>
34 #include <data/settings.h>
35 #include <libpspp/getl.h>
36 #include <libpspp/str.h>
41 #define _(msgid) gettext (msgid)
42 #define N_(msgid) msgid
50 struct string line_buffer;
52 struct source_stream *ss;
54 int token; /* Current token. */
55 double tokval; /* T_POS_NUM, T_NEG_NUM: the token's value. */
57 char tokid [LONG_NAME_LEN + 1]; /* T_ID: the identifier. */
59 struct string tokstr; /* T_ID, T_STRING: token string value.
60 For T_ID, this is not truncated as is
63 char *prog; /* Pointer to next token in line_buffer. */
64 bool dot; /* True only if this line ends with a terminal dot. */
65 bool eof; /* True only if the last token returned was T_STOP. */
67 int put_token ; /* If nonzero, next token returned by lex_get().
68 Used only in exceptional circumstances. */
70 struct string put_tokstr;
75 static int parse_id (struct lexer *);
77 /* How a string represents its contents. */
80 CHARACTER_STRING, /* Characters. */
81 BINARY_STRING, /* Binary digits. */
82 OCTAL_STRING, /* Octal digits. */
83 HEX_STRING /* Hexadecimal digits. */
86 static int parse_string (struct lexer *, enum string_type);
89 static void dump_token (void);
94 /* Initializes the lexer. */
96 lex_create (struct source_stream *ss)
98 struct lexer *lexer = xzalloc (sizeof (*lexer));
100 ds_init_empty (&lexer->tokstr);
101 ds_init_empty (&lexer->put_tokstr);
102 ds_init_empty (&lexer->line_buffer);
105 if (!lex_get_line (lexer))
111 struct source_stream *
112 lex_get_source_stream (const struct lexer *lex)
119 lex_destroy (struct lexer *lexer)
123 ds_destroy (&lexer->put_tokstr);
124 ds_destroy (&lexer->tokstr);
125 ds_destroy (&lexer->line_buffer);
132 /* Common functions. */
134 /* Copies put_token, lexer->put_tokstr, put_tokval into token, tokstr,
135 tokval, respectively, and sets tokid appropriately. */
137 restore_token (struct lexer *lexer)
139 assert (lexer->put_token != 0);
140 lexer->token = lexer->put_token;
141 ds_assign_string (&lexer->tokstr, &lexer->put_tokstr);
142 str_copy_trunc (lexer->tokid, sizeof lexer->tokid, ds_cstr (&lexer->tokstr));
143 lexer->tokval = lexer->put_tokval;
144 lexer->put_token = 0;
147 /* Copies token, tokstr, lexer->tokval into lexer->put_token, put_tokstr,
148 put_lexer->tokval respectively. */
150 save_token (struct lexer *lexer)
152 lexer->put_token = lexer->token;
153 ds_assign_string (&lexer->put_tokstr, &lexer->tokstr);
154 lexer->put_tokval = lexer->tokval;
157 /* Parses a single token, setting appropriate global variables to
158 indicate the token's attributes. */
160 lex_get (struct lexer *lexer)
162 /* If a token was pushed ahead, return it. */
163 if (lexer->put_token)
165 restore_token (lexer);
175 /* Skip whitespace. */
178 lexer->token = T_STOP;
184 while (isspace ((unsigned char) *lexer->prog))
198 else if (!lex_get_line (lexer))
201 lexer->token = T_STOP;
208 if (lexer->put_token)
210 restore_token (lexer);
219 /* Actually parse the token. */
220 ds_clear (&lexer->tokstr);
222 switch (*lexer->prog)
225 case '0': case '1': case '2': case '3': case '4':
226 case '5': case '6': case '7': case '8': case '9':
230 /* `-' can introduce a negative number, or it can be a
231 token by itself. If it is not followed by a digit or a
232 decimal point, it is definitely not a number.
233 Otherwise, it might be either, but most of the time we
234 want it as a number. When the syntax calls for a `-'
235 token, lex_negative_to_dash() must be used to break
236 negative numbers into two tokens. */
237 if (*lexer->prog == '-')
239 ds_put_char (&lexer->tokstr, *lexer->prog++);
240 while (isspace ((unsigned char) *lexer->prog))
243 if (!isdigit ((unsigned char) *lexer->prog) && *lexer->prog != '.')
248 lexer->token = T_NEG_NUM;
251 lexer->token = T_POS_NUM;
253 /* Parse the number, copying it into tokstr. */
254 while (isdigit ((unsigned char) *lexer->prog))
255 ds_put_char (&lexer->tokstr, *lexer->prog++);
256 if (*lexer->prog == '.')
258 ds_put_char (&lexer->tokstr, *lexer->prog++);
259 while (isdigit ((unsigned char) *lexer->prog))
260 ds_put_char (&lexer->tokstr, *lexer->prog++);
262 if (*lexer->prog == 'e' || *lexer->prog == 'E')
264 ds_put_char (&lexer->tokstr, *lexer->prog++);
265 if (*lexer->prog == '+' || *lexer->prog == '-')
266 ds_put_char (&lexer->tokstr, *lexer->prog++);
267 while (isdigit ((unsigned char) *lexer->prog))
268 ds_put_char (&lexer->tokstr, *lexer->prog++);
271 /* Parse as floating point. */
272 lexer->tokval = strtod (ds_cstr (&lexer->tokstr), &tail);
275 msg (SE, _("%s does not form a valid number."),
276 ds_cstr (&lexer->tokstr));
279 ds_clear (&lexer->tokstr);
280 ds_put_char (&lexer->tokstr, '0');
287 lexer->token = parse_string (lexer, CHARACTER_STRING);
290 case '(': case ')': case ',': case '=': case '+': case '/':
291 lexer->token = *lexer->prog++;
295 if (*++lexer->prog == '*')
298 lexer->token = T_EXP;
305 if (*++lexer->prog == '=')
310 else if (*lexer->prog == '>')
320 if (*++lexer->prog == '=')
330 if (*++lexer->prog == '=')
336 lexer->token = T_NOT;
341 lexer->token = T_AND;
350 if (lexer->prog[1] == '\'' || lexer->prog[1] == '"')
351 lexer->token = parse_string (lexer, BINARY_STRING);
353 lexer->token = parse_id (lexer);
357 if (lexer->prog[1] == '\'' || lexer->prog[1] == '"')
358 lexer->token = parse_string (lexer, OCTAL_STRING);
360 lexer->token = parse_id (lexer);
364 if (lexer->prog[1] == '\'' || lexer->prog[1] == '"')
365 lexer->token = parse_string (lexer, HEX_STRING);
367 lexer->token = parse_id (lexer);
371 if (lex_is_id1 (*lexer->prog))
373 lexer->token = parse_id (lexer);
378 if (isgraph ((unsigned char) *lexer->prog))
379 msg (SE, _("Bad character in input: `%c'."), *lexer->prog++);
381 msg (SE, _("Bad character in input: `\\%o'."), *lexer->prog++);
393 /* Parses an identifier at the current position into tokid and
395 Returns the correct token type. */
397 parse_id (struct lexer *lexer)
399 struct substring rest_of_line
400 = ss_substr (ds_ss (&lexer->line_buffer),
401 ds_pointer_to_position (&lexer->line_buffer, lexer->prog),
403 struct substring id = ss_head (rest_of_line,
404 lex_id_get_length (rest_of_line));
405 lexer->prog += ss_length (id);
407 ds_assign_substring (&lexer->tokstr, id);
408 str_copy_trunc (lexer->tokid, sizeof lexer->tokid, ds_cstr (&lexer->tokstr));
409 return lex_id_to_token (id);
412 /* Reports an error to the effect that subcommand SBC may only be
415 lex_sbc_only_once (const char *sbc)
417 msg (SE, _("Subcommand %s may only be specified once."), sbc);
420 /* Reports an error to the effect that subcommand SBC is
423 lex_sbc_missing (struct lexer *lexer, const char *sbc)
425 lex_error (lexer, _("missing required subcommand %s"), sbc);
428 /* Prints a syntax error message containing the current token and
429 given message MESSAGE (if non-null). */
431 lex_error (struct lexer *lexer, const char *message, ...)
436 token_rep = lex_token_representation (lexer);
437 if (lexer->token == T_STOP)
438 strcpy (where, "end of file");
439 else if (lexer->token == '.')
440 strcpy (where, "end of command");
442 snprintf (where, sizeof where, "`%s'", token_rep);
450 va_start (args, message);
451 vsnprintf (buf, 1024, message, args);
454 msg (SE, _("Syntax error %s at %s."), buf, where);
457 msg (SE, _("Syntax error at %s."), where);
460 /* Checks that we're at end of command.
461 If so, returns a successful command completion code.
462 If not, flags a syntax error and returns an error command
465 lex_end_of_command (struct lexer *lexer)
467 if (lexer->token != '.')
469 lex_error (lexer, _("expecting end of command"));
476 /* Token testing functions. */
478 /* Returns true if the current token is a number. */
480 lex_is_number (struct lexer *lexer)
482 return lexer->token == T_POS_NUM || lexer->token == T_NEG_NUM;
485 /* Returns the value of the current token, which must be a
486 floating point number. */
488 lex_number (struct lexer *lexer)
490 assert (lex_is_number (lexer));
491 return lexer->tokval;
494 /* Returns true iff the current token is an integer. */
496 lex_is_integer (struct lexer *lexer)
498 return (lex_is_number (lexer)
499 && lexer->tokval != NOT_LONG
500 && lexer->tokval >= LONG_MIN
501 && lexer->tokval <= LONG_MAX
502 && floor (lexer->tokval) == lexer->tokval);
505 /* Returns the value of the current token, which must be an
508 lex_integer (struct lexer *lexer)
510 assert (lex_is_integer (lexer));
511 return lexer->tokval;
514 /* Token matching functions. */
516 /* If TOK is the current token, skips it and returns true
517 Otherwise, returns false. */
519 lex_match (struct lexer *lexer, int t)
521 if (lexer->token == t)
530 /* If the current token is the identifier S, skips it and returns
531 true. The identifier may be abbreviated to its first three
533 Otherwise, returns false. */
535 lex_match_id (struct lexer *lexer, const char *s)
537 if (lexer->token == T_ID
538 && lex_id_match (ss_cstr (s), ss_cstr (lexer->tokid)))
547 /* If the current token is integer N, skips it and returns true.
548 Otherwise, returns false. */
550 lex_match_int (struct lexer *lexer, int x)
552 if (lex_is_integer (lexer) && lex_integer (lexer) == x)
561 /* Forced matches. */
563 /* If this token is identifier S, fetches the next token and returns
565 Otherwise, reports an error and returns zero. */
567 lex_force_match_id (struct lexer *lexer, const char *s)
569 if (lex_match_id (lexer, s))
573 lex_error (lexer, _("expecting `%s'"), s);
578 /* If the current token is T, skips the token. Otherwise, reports an
579 error and returns from the current function with return value false. */
581 lex_force_match (struct lexer *lexer, int t)
583 if (lexer->token == t)
590 lex_error (lexer, _("expecting `%s'"), lex_token_name (t));
595 /* If this token is a string, does nothing and returns true.
596 Otherwise, reports an error and returns false. */
598 lex_force_string (struct lexer *lexer)
600 if (lexer->token == T_STRING)
604 lex_error (lexer, _("expecting string"));
609 /* If this token is an integer, does nothing and returns true.
610 Otherwise, reports an error and returns false. */
612 lex_force_int (struct lexer *lexer)
614 if (lex_is_integer (lexer))
618 lex_error (lexer, _("expecting integer"));
623 /* If this token is a number, does nothing and returns true.
624 Otherwise, reports an error and returns false. */
626 lex_force_num (struct lexer *lexer)
628 if (lex_is_number (lexer))
631 lex_error (lexer, _("expecting number"));
635 /* If this token is an identifier, does nothing and returns true.
636 Otherwise, reports an error and returns false. */
638 lex_force_id (struct lexer *lexer)
640 if (lexer->token == T_ID)
643 lex_error (lexer, _("expecting identifier"));
647 /* Weird token functions. */
649 /* Returns the first character of the next token, except that if the
650 next token is not an identifier, the character returned will not be
651 a character that can begin an identifier. Specifically, the
652 hexstring lead-in X' causes lookahead() to return '. Note that an
653 alphanumeric return value doesn't guarantee an ID token, it could
654 also be a reserved-word token. */
656 lex_look_ahead (struct lexer *lexer)
658 if (lexer->put_token)
659 return lexer->put_token;
668 while (isspace ((unsigned char) *lexer->prog))
675 else if (!lex_get_line (lexer))
678 if (lexer->put_token)
679 return lexer->put_token;
682 if ((toupper ((unsigned char) *lexer->prog) == 'X'
683 || toupper ((unsigned char) *lexer->prog) == 'B'
684 || toupper ((unsigned char) *lexer->prog) == 'O')
685 && (lexer->prog[1] == '\'' || lexer->prog[1] == '"'))
692 /* Makes the current token become the next token to be read; the
693 current token is set to T. */
695 lex_put_back (struct lexer *lexer, int t)
701 /* Makes the current token become the next token to be read; the
702 current token is set to the identifier ID. */
704 lex_put_back_id (struct lexer *lexer, const char *id)
706 assert (lex_id_to_token (ss_cstr (id)) == T_ID);
709 ds_assign_cstr (&lexer->tokstr, id);
710 str_copy_trunc (lexer->tokid, sizeof lexer->tokid, ds_cstr (&lexer->tokstr));
713 /* Weird line processing functions. */
715 /* Returns the entire contents of the current line. */
717 lex_entire_line (struct lexer *lexer)
719 return ds_cstr (&lexer->line_buffer);
722 const struct string *
723 lex_entire_line_ds (struct lexer *lexer)
725 return &lexer->line_buffer;
728 /* As lex_entire_line(), but only returns the part of the current line
729 that hasn't already been tokenized.
730 If END_DOT is non-null, stores nonzero into *END_DOT if the line
731 ends with a terminal dot, or zero if it doesn't. */
733 lex_rest_of_line (struct lexer *lexer, int *end_dot)
736 *end_dot = lexer->dot;
740 /* Causes the rest of the current input line to be ignored for
741 tokenization purposes. */
743 lex_discard_line (struct lexer *lexer)
745 ds_cstr (&lexer->line_buffer); /* Ensures ds_end points to something valid */
746 lexer->prog = ds_end (&lexer->line_buffer);
748 lexer->put_token = 0;
752 /* Discards the rest of the current command.
753 When we're reading commands from a file, we skip tokens until
754 a terminal dot or EOF.
755 When we're reading commands interactively from the user,
756 that's just discarding the current line, because presumably
757 the user doesn't want to finish typing a command that will be
760 lex_discard_rest_of_command (struct lexer *lexer)
762 if (!getl_is_interactive (lexer->ss))
764 while (lexer->token != T_STOP && lexer->token != '.')
768 lex_discard_line (lexer);
771 /* Weird line reading functions. */
773 /* Remove C-style comments in STRING, begun by slash-star and
774 terminated by star-slash or newline. */
776 strip_comments (struct string *string)
784 for (cp = ds_cstr (string); *cp; )
786 /* If we're not in a comment, check for quote marks. */
791 else if (*cp == '\'' || *cp == '"')
795 /* If we're not inside a quotation, check for comment. */
798 if (cp[0] == '/' && cp[1] == '*')
805 else if (in_comment && cp[0] == '*' && cp[1] == '/')
814 /* Check commenting. */
821 /* Prepares LINE, which is subject to the given SYNTAX rules, for
822 tokenization by stripping comments and determining whether it
823 is the beginning or end of a command and storing into
824 *LINE_STARTS_COMMAND and *LINE_ENDS_COMMAND appropriately. */
826 lex_preprocess_line (struct string *line,
827 enum getl_syntax syntax,
828 bool *line_starts_command,
829 bool *line_ends_command)
831 strip_comments (line);
832 ds_rtrim (line, ss_cstr (CC_SPACES));
833 *line_ends_command = (ds_chomp (line, get_endcmd ())
834 || (ds_is_empty (line) && get_nulline ()));
835 *line_starts_command = false;
836 if (syntax == GETL_BATCH)
838 int first = ds_first (line);
839 *line_starts_command = !isspace (first);
840 if (first == '+' || first == '-')
841 *ds_data (line) = ' ';
845 /* Reads a line, without performing any preprocessing.
846 Sets *SYNTAX, if SYNTAX is non-null, to the line's syntax
849 lex_get_line_raw (struct lexer *lexer, enum getl_syntax *syntax)
851 enum getl_syntax dummy;
852 bool ok = getl_read_line (lexer->ss, &lexer->line_buffer,
853 syntax != NULL ? syntax : &dummy);
857 /* Reads a line for use by the tokenizer, and preprocesses it by
858 removing comments, stripping trailing whitespace and the
859 terminal dot, and removing leading indentors. */
861 lex_get_line (struct lexer *lexer)
863 bool line_starts_command;
864 enum getl_syntax syntax;
866 if (!lex_get_line_raw (lexer, &syntax))
869 lex_preprocess_line (&lexer->line_buffer, syntax,
870 &line_starts_command, &lexer->dot);
871 if (line_starts_command)
872 lexer->put_token = '.';
874 lexer->prog = ds_cstr (&lexer->line_buffer);
880 /* Returns the name of a token. */
882 lex_token_name (int token)
884 if (lex_is_keyword (token))
885 return lex_id_name (token);
886 else if (token < 256)
888 static char t[256][2];
898 /* Returns an ASCII representation of the current token as a
899 malloc()'d string. */
901 lex_token_representation (struct lexer *lexer)
905 switch (lexer->token)
910 return ds_xstrdup (&lexer->tokstr);
918 for (sp = ds_cstr (&lexer->tokstr); sp < ds_end (&lexer->tokstr); sp++)
919 if (!isprint ((unsigned char) *sp))
925 token_rep = xmalloc (2 + ds_length (&lexer->tokstr) * 2 + 1 + 1);
933 for (sp = ds_cstr (&lexer->tokstr); *sp; )
937 *dp++ = (unsigned char) *sp++;
940 for (sp = ds_cstr (&lexer->tokstr); sp < ds_end (&lexer->tokstr); sp++)
942 *dp++ = (((unsigned char) *sp) >> 4)["0123456789ABCDEF"];
943 *dp++ = (((unsigned char) *sp) & 15)["0123456789ABCDEF"];
953 token_rep = xmalloc (1);
958 return xstrdup ("**");
961 return xstrdup (lex_token_name (lexer->token));
967 /* Really weird functions. */
969 /* Most of the time, a `-' is a lead-in to a negative number. But
970 sometimes it's actually part of the syntax. If a dash can be part
971 of syntax then this function is called to rip it off of a
974 lex_negative_to_dash (struct lexer *lexer)
976 if (lexer->token == T_NEG_NUM)
978 lexer->token = T_POS_NUM;
979 lexer->tokval = -lexer->tokval;
980 ds_assign_substring (&lexer->tokstr, ds_substr (&lexer->tokstr, 1, SIZE_MAX));
986 /* Skip a COMMENT command. */
988 lex_skip_comment (struct lexer *lexer)
992 if (!lex_get_line (lexer))
994 lexer->put_token = T_STOP;
999 if (lexer->put_token == '.')
1002 ds_cstr (&lexer->line_buffer); /* Ensures ds_end will point to a valid char */
1003 lexer->prog = ds_end (&lexer->line_buffer);
1009 /* Private functions. */
1011 /* When invoked, tokstr contains a string of binary, octal, or
1012 hex digits, according to TYPE. The string is converted to
1013 characters having the specified values. */
1015 convert_numeric_string_to_char_string (struct lexer *lexer,
1016 enum string_type type)
1018 const char *base_name;
1028 base_name = _("binary");
1033 base_name = _("octal");
1038 base_name = _("hex");
1046 byte_cnt = ds_length (&lexer->tokstr) / chars_per_byte;
1047 if (ds_length (&lexer->tokstr) % chars_per_byte)
1048 msg (SE, _("String of %s digits has %d characters, which is not a "
1050 base_name, ds_length (&lexer->tokstr), chars_per_byte);
1052 p = ds_cstr (&lexer->tokstr);
1053 for (i = 0; i < byte_cnt; i++)
1059 for (j = 0; j < chars_per_byte; j++, p++)
1063 if (*p >= '0' && *p <= '9')
1067 static const char alpha[] = "abcdef";
1068 const char *q = strchr (alpha, tolower ((unsigned char) *p));
1077 msg (SE, _("`%c' is not a valid %s digit."), *p, base_name);
1079 value = value * base + v;
1082 ds_cstr (&lexer->tokstr)[i] = (unsigned char) value;
1085 ds_truncate (&lexer->tokstr, byte_cnt);
1088 /* Parses a string from the input buffer into tokstr. The input
1089 buffer pointer lexer->prog must point to the initial single or double
1090 quote. TYPE indicates the type of string to be parsed.
1091 Returns token type. */
1093 parse_string (struct lexer *lexer, enum string_type type)
1095 if (type != CHARACTER_STRING)
1098 /* Accumulate the entire string, joining sections indicated by +
1102 /* Single or double quote. */
1103 int c = *lexer->prog++;
1105 /* Accumulate section. */
1108 /* Check end of line. */
1109 if (*lexer->prog == '\0')
1111 msg (SE, _("Unterminated string constant."));
1115 /* Double quote characters to embed them in strings. */
1116 if (*lexer->prog == c)
1118 if (lexer->prog[1] == c)
1124 ds_put_char (&lexer->tokstr, *lexer->prog++);
1128 /* Skip whitespace after final quote mark. */
1133 while (isspace ((unsigned char) *lexer->prog))
1141 if (!lex_get_line (lexer))
1145 /* Skip plus sign. */
1146 if (*lexer->prog != '+')
1150 /* Skip whitespace after plus sign. */
1155 while (isspace ((unsigned char) *lexer->prog))
1163 if (!lex_get_line (lexer))
1165 msg (SE, _("Unexpected end of file in string concatenation."));
1170 /* Ensure that a valid string follows. */
1171 if (*lexer->prog != '\'' && *lexer->prog != '"')
1173 msg (SE, _("String expected following `+'."));
1178 /* We come here when we've finished concatenating all the string sections
1179 into one large string. */
1181 if (type != CHARACTER_STRING)
1182 convert_numeric_string_to_char_string (lexer, type);
1184 if (ds_length (&lexer->tokstr) > 255)
1186 msg (SE, _("String exceeds 255 characters in length (%d characters)."),
1187 ds_length (&lexer->tokstr));
1188 ds_truncate (&lexer->tokstr, 255);
1195 /* Reads one token from the lexer and writes a textual representation
1196 on stdout for debugging purposes. */
1198 dump_token (struct lexer *lexer)
1204 getl_location (&curfn, &curln);
1206 fprintf (stderr, "%s:%d\t", curfn, curln);
1209 switch (lexer->token)
1212 fprintf (stderr, "ID\t%s\n", lexer->tokid);
1217 fprintf (stderr, "NUM\t%f\n", lexer->tokval);
1221 fprintf (stderr, "STRING\t\"%s\"\n", ds_cstr (&lexer->tokstr));
1225 fprintf (stderr, "STOP\n");
1229 fprintf (stderr, "MISC\tEXP\"");
1233 fprintf (stderr, "MISC\tEOF\n");
1237 if (lex_is_keyword (token))
1238 fprintf (stderr, "KEYWORD\t%s\n", lex_token_name (token));
1240 fprintf (stderr, "PUNCT\t%c\n", lexer->token);
1244 #endif /* DUMP_TOKENS */
1247 /* Token Accessor Functions */
1250 lex_token (const struct lexer *lexer)
1252 return lexer->token;
1256 lex_tokval (const struct lexer *lexer)
1258 return lexer->tokval;
1262 lex_tokid (const struct lexer *lexer)
1264 return lexer->tokid;
1267 const struct string *
1268 lex_tokstr (const struct lexer *lexer)
1270 return &lexer->tokstr;