1 /* Copyright (C) 2002, 2003, 2004, 2005
2 Free Software Foundation, Inc.
3 Contributed by Andy Vaught
5 This file is part of the GNU Fortran 95 runtime library (libgfortran).
7 Libgfortran is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2, or (at your option)
12 In addition to the permissions in the GNU General Public License, the
13 Free Software Foundation gives you unlimited permission to link the
14 compiled version of this file into combinations with other programs,
15 and to distribute those combinations without any restriction coming
16 from the use of this file. (The General Public License restrictions
17 do apply in other respects; for example, they cover modification of
18 the file, and distribution when not linked into a combine
21 Libgfortran is distributed in the hope that it will be useful,
22 but WITHOUT ANY WARRANTY; without even the implied warranty of
23 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24 GNU General Public License for more details.
26 You should have received a copy of the GNU General Public License
27 along with Libgfortran; see the file COPYING. If not, write to
28 the Free Software Foundation, 51 Franklin Street, Fifth Floor,
29 Boston, MA 02110-1301, USA. */
32 /* format.c-- parse a FORMAT string into a binary format suitable for
33 * interpretation during I/O statements */
38 #include "libgfortran.h"
41 #define FARRAY_SIZE 64
43 typedef struct fnode_array
45 struct fnode_array *next;
46 fnode array[FARRAY_SIZE];
50 typedef struct format_data
52 char *format_string, *string;
54 format_token saved_token;
55 int value, format_string_len, reversion_ok;
57 const fnode *saved_format;
63 static const fnode colon_node = { FMT_COLON, 0, NULL, NULL, {{ 0, 0, 0 }}, 0,
68 static const char posint_required[] = "Positive width required in format",
69 period_required[] = "Period required in format",
70 nonneg_required[] = "Nonnegative width required in format",
71 unexpected_element[] = "Unexpected element in format",
72 unexpected_end[] = "Unexpected end of format string",
73 bad_string[] = "Unterminated character constant in format",
74 bad_hollerith[] = "Hollerith constant extends past the end of the format",
75 reversion_error[] = "Exhausted data descriptors in format";
78 /* next_char()-- Return the next character in the format string.
79 * Returns -1 when the string is done. If the literal flag is set,
80 * spaces are significant, otherwise they are not. */
83 next_char (format_data *fmt, int literal)
89 if (fmt->format_string_len == 0)
92 fmt->format_string_len--;
93 c = toupper (*fmt->format_string++);
95 while (c == ' ' && !literal);
101 /* unget_char()-- Back up one character position. */
103 #define unget_char(fmt) \
104 { fmt->format_string--; fmt->format_string_len++; }
107 /* get_fnode()-- Allocate a new format node, inserting it into the
108 * current singly linked list. These are initially allocated from the
112 get_fnode (format_data *fmt, fnode **head, fnode **tail, format_token t)
116 if (fmt->avail == &fmt->last->array[FARRAY_SIZE])
118 fmt->last->next = get_mem (sizeof (fnode_array));
119 fmt->last = fmt->last->next;
120 fmt->last->next = NULL;
121 fmt->avail = &fmt->last->array[0];
124 memset (f, '\0', sizeof (fnode));
136 f->source = fmt->format_string;
141 /* free_format_data()-- Free all allocated format data. */
144 free_format_data (st_parameter_dt *dtp)
146 fnode_array *fa, *fa_next;
147 format_data *fmt = dtp->u.p.fmt;
152 for (fa = fmt->array.next; fa; fa = fa_next)
163 /* format_lex()-- Simple lexical analyzer for getting the next token
164 * in a FORMAT string. We support a one-level token pushback in the
165 * fmt->saved_token variable. */
168 format_lex (format_data *fmt)
175 if (fmt->saved_token != FMT_NONE)
177 token = fmt->saved_token;
178 fmt->saved_token = FMT_NONE;
183 c = next_char (fmt, 0);
192 c = next_char (fmt, 0);
199 fmt->value = c - '0';
203 c = next_char (fmt, 0);
207 fmt->value = 10 * fmt->value + c - '0';
213 fmt->value = -fmt->value;
214 token = FMT_SIGNED_INT;
227 fmt->value = c - '0';
231 c = next_char (fmt, 0);
235 fmt->value = 10 * fmt->value + c - '0';
239 token = (fmt->value == 0) ? FMT_ZERO : FMT_POSINT;
263 switch (next_char (fmt, 0))
292 switch (next_char (fmt, 0))
309 switch (next_char (fmt, 0))
329 fmt->string = fmt->format_string;
330 fmt->value = 0; /* This is the length of the string */
334 c = next_char (fmt, 1);
337 token = FMT_BADSTRING;
338 fmt->error = bad_string;
344 c = next_char (fmt, 1);
348 token = FMT_BADSTRING;
349 fmt->error = bad_string;
387 switch (next_char (fmt, 0))
436 /* parse_format_list()-- Parse a format list. Assumes that a left
437 * paren has already been seen. Returns a list representing the
438 * parenthesis node which contains the rest of the list. */
441 parse_format_list (st_parameter_dt *dtp)
444 format_token t, u, t2;
446 format_data *fmt = dtp->u.p.fmt;
450 /* Get the next format item */
452 t = format_lex (fmt);
459 t = format_lex (fmt);
463 get_fnode (fmt, &head, &tail, FMT_LPAREN);
464 tail->repeat = repeat;
465 tail->u.child = parse_format_list (dtp);
466 if (fmt->error != NULL)
472 get_fnode (fmt, &head, &tail, FMT_SLASH);
473 tail->repeat = repeat;
477 get_fnode (fmt, &head, &tail, FMT_X);
479 tail->u.k = fmt->value;
490 get_fnode (fmt, &head, &tail, FMT_LPAREN);
492 tail->u.child = parse_format_list (dtp);
493 if (fmt->error != NULL)
498 case FMT_SIGNED_INT: /* Signed integer can only precede a P format. */
499 case FMT_ZERO: /* Same for zero. */
500 t = format_lex (fmt);
503 fmt->error = "Expected P edit descriptor in format";
508 get_fnode (fmt, &head, &tail, FMT_P);
509 tail->u.k = fmt->value;
512 t = format_lex (fmt);
513 if (t == FMT_F || t == FMT_EN || t == FMT_ES || t == FMT_D
514 || t == FMT_G || t == FMT_E)
520 fmt->saved_token = t;
523 case FMT_P: /* P and X require a prior number */
524 fmt->error = "P descriptor requires leading scale factor";
531 If we would be pedantic in the library, we would have to reject
532 an X descriptor without an integer prefix:
534 fmt->error = "X descriptor requires leading space count";
537 However, this is an extension supported by many Fortran compilers,
538 including Cray, HP, AIX, and IRIX. Therefore, we allow it in the
539 runtime library, and make the front end reject it if the compiler
540 is in pedantic mode. The interpretation of 'X' is '1X'.
542 get_fnode (fmt, &head, &tail, FMT_X);
548 get_fnode (fmt, &head, &tail, FMT_STRING);
550 tail->u.string.p = fmt->string;
551 tail->u.string.length = fmt->value;
560 get_fnode (fmt, &head, &tail, t);
565 get_fnode (fmt, &head, &tail, FMT_COLON);
570 get_fnode (fmt, &head, &tail, FMT_SLASH);
576 get_fnode (fmt, &head, &tail, FMT_DOLLAR);
578 notify_std (GFC_STD_GNU, "Extension: $ descriptor");
584 t2 = format_lex (fmt);
585 if (t2 != FMT_POSINT)
587 fmt->error = posint_required;
590 get_fnode (fmt, &head, &tail, t);
591 tail->u.n = fmt->value;
611 get_fnode (fmt, &head, &tail, FMT_STRING);
613 if (fmt->format_string_len < 1)
615 fmt->error = bad_hollerith;
619 tail->u.string.p = fmt->format_string;
620 tail->u.string.length = 1;
623 fmt->format_string++;
624 fmt->format_string_len--;
629 fmt->error = unexpected_end;
639 fmt->error = unexpected_element;
643 /* In this state, t must currently be a data descriptor. Deal with
644 things that can/must follow the descriptor */
649 t = format_lex (fmt);
652 fmt->error = "Repeat count cannot follow P descriptor";
656 fmt->saved_token = t;
657 get_fnode (fmt, &head, &tail, FMT_P);
662 t = format_lex (fmt);
665 fmt->error = posint_required;
669 get_fnode (fmt, &head, &tail, FMT_L);
670 tail->u.n = fmt->value;
671 tail->repeat = repeat;
675 t = format_lex (fmt);
678 fmt->saved_token = t;
679 fmt->value = -1; /* Width not present */
682 get_fnode (fmt, &head, &tail, FMT_A);
683 tail->repeat = repeat;
684 tail->u.n = fmt->value;
693 get_fnode (fmt, &head, &tail, t);
694 tail->repeat = repeat;
696 u = format_lex (fmt);
697 if (t == FMT_F || dtp->u.p.mode == WRITING)
699 if (u != FMT_POSINT && u != FMT_ZERO)
701 fmt->error = nonneg_required;
709 fmt->error = posint_required;
714 tail->u.real.w = fmt->value;
716 t = format_lex (fmt);
719 fmt->error = period_required;
723 t = format_lex (fmt);
724 if (t != FMT_ZERO && t != FMT_POSINT)
726 fmt->error = nonneg_required;
730 tail->u.real.d = fmt->value;
732 if (t == FMT_D || t == FMT_F)
737 /* Look for optional exponent */
738 t = format_lex (fmt);
740 fmt->saved_token = t;
743 t = format_lex (fmt);
746 fmt->error = "Positive exponent width required in format";
750 tail->u.real.e = fmt->value;
756 if (repeat > fmt->format_string_len)
758 fmt->error = bad_hollerith;
762 get_fnode (fmt, &head, &tail, FMT_STRING);
764 tail->u.string.p = fmt->format_string;
765 tail->u.string.length = repeat;
768 fmt->format_string += fmt->value;
769 fmt->format_string_len -= repeat;
777 get_fnode (fmt, &head, &tail, t);
778 tail->repeat = repeat;
780 t = format_lex (fmt);
782 if (dtp->u.p.mode == READING)
786 fmt->error = posint_required;
792 if (t != FMT_ZERO && t != FMT_POSINT)
794 fmt->error = nonneg_required;
799 tail->u.integer.w = fmt->value;
800 tail->u.integer.m = -1;
802 t = format_lex (fmt);
805 fmt->saved_token = t;
809 t = format_lex (fmt);
810 if (t != FMT_ZERO && t != FMT_POSINT)
812 fmt->error = nonneg_required;
816 tail->u.integer.m = fmt->value;
819 if (tail->u.integer.w != 0 && tail->u.integer.m > tail->u.integer.w)
821 fmt->error = "Minimum digits exceeds field width";
828 fmt->error = unexpected_element;
832 /* Between a descriptor and what comes next */
834 t = format_lex (fmt);
844 get_fnode (fmt, &head, &tail, FMT_SLASH);
853 fmt->error = unexpected_end;
857 /* Assume a missing comma, this is a GNU extension */
861 /* Optional comma is a weird between state where we've just finished
862 reading a colon, slash or P descriptor. */
864 t = format_lex (fmt);
873 default: /* Assume that we have another format item */
874 fmt->saved_token = t;
885 /* format_error()-- Generate an error message for a format statement.
886 * If the node that gives the location of the error is NULL, the error
887 * is assumed to happen at parse time, and the current location of the
890 * We generate a message showing where the problem is. We take extra
891 * care to print only the relevant part of the format if it is longer
892 * than a standard 80 column display. */
895 format_error (st_parameter_dt *dtp, const fnode *f, const char *message)
897 int width, i, j, offset;
898 char *p, buffer[300];
899 format_data *fmt = dtp->u.p.fmt;
902 fmt->format_string = f->source;
904 st_sprintf (buffer, "%s\n", message);
906 j = fmt->format_string - dtp->format;
908 offset = (j > 60) ? j - 40 : 0;
911 width = dtp->format_len - offset;
916 /* Show the format */
918 p = strchr (buffer, '\0');
920 memcpy (p, dtp->format + offset, width);
925 /* Show where the problem is */
927 for (i = 1; i < j; i++)
933 generate_error (&dtp->common, ERROR_FORMAT, buffer);
937 /* parse_format()-- Parse a format string. */
940 parse_format (st_parameter_dt *dtp)
944 dtp->u.p.fmt = fmt = get_mem (sizeof (format_data));
945 fmt->format_string = dtp->format;
946 fmt->format_string_len = dtp->format_len;
949 fmt->saved_token = FMT_NONE;
953 /* Initialize variables used during traversal of the tree */
955 fmt->reversion_ok = 0;
956 fmt->saved_format = NULL;
958 /* Allocate the first format node as the root of the tree */
960 fmt->last = &fmt->array;
961 fmt->last->next = NULL;
962 fmt->avail = &fmt->array.array[0];
964 memset (fmt->avail, 0, sizeof (*fmt->avail));
965 fmt->avail->format = FMT_LPAREN;
966 fmt->avail->repeat = 1;
969 if (format_lex (fmt) == FMT_LPAREN)
970 fmt->array.array[0].u.child = parse_format_list (dtp);
972 fmt->error = "Missing initial left parenthesis in format";
975 format_error (dtp, NULL, fmt->error);
979 /* revert()-- Do reversion of the format. Control reverts to the left
980 * parenthesis that matches the rightmost right parenthesis. From our
981 * tree structure, we are looking for the rightmost parenthesis node
982 * at the second level, the first level always being a single
983 * parenthesis node. If this node doesn't exit, we use the top
987 revert (st_parameter_dt *dtp)
990 format_data *fmt = dtp->u.p.fmt;
992 dtp->u.p.reversion_flag = 1;
996 for (f = fmt->array.array[0].u.child; f; f = f->next)
997 if (f->format == FMT_LPAREN)
1000 /* If r is NULL because no node was found, the whole tree will be used */
1002 fmt->array.array[0].current = r;
1003 fmt->array.array[0].count = 0;
1007 /* next_format0()-- Get the next format node without worrying about
1008 * reversion. Returns NULL when we hit the end of the list.
1009 * Parenthesis nodes are incremented after the list has been
1010 * exhausted, other nodes are incremented before they are returned. */
1012 static const fnode *
1013 next_format0 (fnode * f)
1020 if (f->format != FMT_LPAREN)
1023 if (f->count <= f->repeat)
1030 /* Deal with a parenthesis node */
1032 for (; f->count < f->repeat; f->count++)
1034 if (f->current == NULL)
1035 f->current = f->u.child;
1037 for (; f->current != NULL; f->current = f->current->next)
1039 r = next_format0 (f->current);
1050 /* next_format()-- Return the next format node. If the format list
1051 * ends up being exhausted, we do reversion. Reversion is only
1052 * allowed if the we've seen a data descriptor since the
1053 * initialization or the last reversion. We return NULL if the there
1054 * are no more data descriptors to return (which is an error
1058 next_format (st_parameter_dt *dtp)
1062 format_data *fmt = dtp->u.p.fmt;
1064 if (fmt->saved_format != NULL)
1065 { /* Deal with a pushed-back format node */
1066 f = fmt->saved_format;
1067 fmt->saved_format = NULL;
1071 f = next_format0 (&fmt->array.array[0]);
1074 if (!fmt->reversion_ok)
1077 fmt->reversion_ok = 0;
1080 f = next_format0 (&fmt->array.array[0]);
1083 format_error (dtp, NULL, reversion_error);
1087 /* Push the first reverted token and return a colon node in case
1088 * there are no more data items. */
1090 fmt->saved_format = f;
1094 /* If this is a data edit descriptor, then reversion has become OK. */
1098 if (!fmt->reversion_ok &&
1099 (t == FMT_I || t == FMT_B || t == FMT_O || t == FMT_Z || t == FMT_F ||
1100 t == FMT_E || t == FMT_EN || t == FMT_ES || t == FMT_G || t == FMT_L ||
1101 t == FMT_A || t == FMT_D))
1102 fmt->reversion_ok = 1;
1107 /* unget_format()-- Push the given format back so that it will be
1108 * returned on the next call to next_format() without affecting
1109 * counts. This is necessary when we've encountered a data
1110 * descriptor, but don't know what the data item is yet. The format
1111 * node is pushed back, and we return control to the main program,
1112 * which calls the library back with the data item (or not). */
1115 unget_format (st_parameter_dt *dtp, const fnode *f)
1117 dtp->u.p.fmt->saved_format = f;