1 /* Copyright (C) 2002, 2003, 2005, 2007, 2008, 2009 Free Software Foundation, Inc.
2 Contributed by Andy Vaught
3 F2003 I/O support contributed by Jerry DeLisle
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 3, or (at your option)
12 Libgfortran is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 Under Section 7 of GPL version 3, you are granted additional
18 permissions described in the GCC Runtime Library Exception, version
19 3.1, as published by the Free Software Foundation.
21 You should have received a copy of the GNU General Public License and
22 a copy of the GCC Runtime Library Exception along with this program;
23 see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
24 <http://www.gnu.org/licenses/>. */
36 typedef unsigned char uchar
;
38 /* read.c -- Deal with formatted reads */
41 /* set_integer()-- All of the integer assignments come here to
42 * actually place the value into memory. */
45 set_integer (void *dest
, GFC_INTEGER_LARGEST value
, int length
)
49 #ifdef HAVE_GFC_INTEGER_16
50 /* length=10 comes about for kind=10 real/complex BOZ, cf. PR41711. */
54 GFC_INTEGER_16 tmp
= value
;
55 memcpy (dest
, (void *) &tmp
, length
);
61 GFC_INTEGER_8 tmp
= value
;
62 memcpy (dest
, (void *) &tmp
, length
);
67 GFC_INTEGER_4 tmp
= value
;
68 memcpy (dest
, (void *) &tmp
, length
);
73 GFC_INTEGER_2 tmp
= value
;
74 memcpy (dest
, (void *) &tmp
, length
);
79 GFC_INTEGER_1 tmp
= value
;
80 memcpy (dest
, (void *) &tmp
, length
);
84 internal_error (NULL
, "Bad integer kind");
89 /* max_value()-- Given a length (kind), return the maximum signed or
93 max_value (int length
, int signed_flag
)
95 GFC_UINTEGER_LARGEST value
;
96 #if defined HAVE_GFC_REAL_16 || defined HAVE_GFC_REAL_10
102 #if defined HAVE_GFC_REAL_16 || defined HAVE_GFC_REAL_10
106 for (n
= 1; n
< 4 * length
; n
++)
107 value
= (value
<< 2) + 3;
113 value
= signed_flag
? 0x7fffffffffffffff : 0xffffffffffffffff;
116 value
= signed_flag
? 0x7fffffff : 0xffffffff;
119 value
= signed_flag
? 0x7fff : 0xffff;
122 value
= signed_flag
? 0x7f : 0xff;
125 internal_error (NULL
, "Bad integer kind");
132 /* convert_real()-- Convert a character representation of a floating
133 * point number to the machine number. Returns nonzero if there is a
134 * range problem during conversion. Note: many architectures
135 * (e.g. IA-64, HP-PA) require that the storage pointed to by the dest
136 * argument is properly aligned for the type in question. TODO:
137 * handle not-a-numbers and infinities. */
140 convert_real (st_parameter_dt
*dtp
, void *dest
, const char *buffer
, int length
)
147 *((GFC_REAL_4
*) dest
) =
148 #if defined(HAVE_STRTOF)
149 gfc_strtof (buffer
, NULL
);
151 (GFC_REAL_4
) gfc_strtod (buffer
, NULL
);
156 *((GFC_REAL_8
*) dest
) = gfc_strtod (buffer
, NULL
);
159 #if defined(HAVE_GFC_REAL_10) && defined (HAVE_STRTOLD)
161 *((GFC_REAL_10
*) dest
) = gfc_strtold (buffer
, NULL
);
165 #if defined(HAVE_GFC_REAL_16) && defined (HAVE_STRTOLD)
167 *((GFC_REAL_16
*) dest
) = gfc_strtold (buffer
, NULL
);
172 internal_error (&dtp
->common
, "Unsupported real kind during IO");
177 generate_error (&dtp
->common
, LIBERROR_READ_VALUE
,
178 "Error during floating point read");
179 next_record (dtp
, 1);
187 /* read_l()-- Read a logical value */
190 read_l (st_parameter_dt
*dtp
, const fnode
*f
, char *dest
, int length
)
197 p
= read_block_form (dtp
, &w
);
220 set_integer (dest
, (GFC_INTEGER_LARGEST
) 1, length
);
224 set_integer (dest
, (GFC_INTEGER_LARGEST
) 0, length
);
228 generate_error (&dtp
->common
, LIBERROR_READ_VALUE
,
229 "Bad value on logical read");
230 next_record (dtp
, 1);
237 read_utf8 (st_parameter_dt
*dtp
, int *nbytes
)
239 static const uchar masks
[6] = { 0x7F, 0x1F, 0x0F, 0x07, 0x02, 0x01 };
240 static const uchar patns
[6] = { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC };
247 s
= read_block_form (dtp
, nbytes
);
251 /* If this is a short read, just return. */
259 /* The number of leading 1-bits in the first byte indicates how many
261 for (nb
= 2; nb
< 7; nb
++)
262 if ((c
& ~masks
[nb
-1]) == patns
[nb
-1])
267 c
= (c
& masks
[nb
-1]);
270 s
= read_block_form (dtp
, &nread
);
273 /* Decode the bytes read. */
274 for (i
= 1; i
< nb
; i
++)
276 gfc_char4_t n
= *s
++;
278 if ((n
& 0xC0) != 0x80)
281 c
= ((c
<< 6) + (n
& 0x3F));
284 /* Make sure the shortest possible encoding was used. */
285 if (c
<= 0x7F && nb
> 1) goto invalid
;
286 if (c
<= 0x7FF && nb
> 2) goto invalid
;
287 if (c
<= 0xFFFF && nb
> 3) goto invalid
;
288 if (c
<= 0x1FFFFF && nb
> 4) goto invalid
;
289 if (c
<= 0x3FFFFFF && nb
> 5) goto invalid
;
291 /* Make sure the character is valid. */
292 if (c
> 0x7FFFFFFF || (c
>= 0xD800 && c
<= 0xDFFF))
298 generate_error (&dtp
->common
, LIBERROR_READ_VALUE
, "Invalid UTF-8 encoding");
299 return (gfc_char4_t
) '?';
304 read_utf8_char1 (st_parameter_dt
*dtp
, char *p
, int len
, int width
)
311 len
= (width
< len
) ? len
: width
;
315 /* Proceed with decoding one character at a time. */
316 for (j
= 0; j
< len
; j
++, dest
++)
318 c
= read_utf8 (dtp
, &nbytes
);
320 /* Check for a short read and if so, break out. */
324 *dest
= c
> 255 ? '?' : (uchar
) c
;
327 /* If there was a short read, pad the remaining characters. */
328 for (i
= j
; i
< len
; i
++)
334 read_default_char1 (st_parameter_dt
*dtp
, char *p
, int len
, int width
)
339 s
= read_block_form (dtp
, &width
);
346 m
= (width
> len
) ? len
: width
;
351 memset (p
+ m
, ' ', n
);
356 read_utf8_char4 (st_parameter_dt
*dtp
, void *p
, int len
, int width
)
362 len
= (width
< len
) ? len
: width
;
364 dest
= (gfc_char4_t
*) p
;
366 /* Proceed with decoding one character at a time. */
367 for (j
= 0; j
< len
; j
++, dest
++)
369 *dest
= read_utf8 (dtp
, &nbytes
);
371 /* Check for a short read and if so, break out. */
376 /* If there was a short read, pad the remaining characters. */
377 for (i
= j
; i
< len
; i
++)
378 *dest
++ = (gfc_char4_t
) ' ';
384 read_default_char4 (st_parameter_dt
*dtp
, char *p
, int len
, int width
)
390 s
= read_block_form (dtp
, &width
);
397 m
= ((int) width
> len
) ? len
: (int) width
;
399 dest
= (gfc_char4_t
*) p
;
401 for (n
= 0; n
< m
; n
++, dest
++, s
++)
402 *dest
= (unsigned char ) *s
;
404 for (n
= 0; n
< len
- (int) width
; n
++, dest
++)
405 *dest
= (unsigned char) ' ';
409 /* read_a()-- Read a character record into a KIND=1 character destination,
410 processing UTF-8 encoding if necessary. */
413 read_a (st_parameter_dt
*dtp
, const fnode
*f
, char *p
, int length
)
419 if (wi
== -1) /* '(A)' edit descriptor */
423 /* Read in w characters, treating comma as not a separator. */
424 dtp
->u
.p
.sf_read_comma
= 0;
426 if (dtp
->u
.p
.current_unit
->flags
.encoding
== ENCODING_UTF8
)
427 read_utf8_char1 (dtp
, p
, length
, w
);
429 read_default_char1 (dtp
, p
, length
, w
);
431 dtp
->u
.p
.sf_read_comma
=
432 dtp
->u
.p
.current_unit
->decimal_status
== DECIMAL_COMMA
? 0 : 1;
436 /* read_a_char4()-- Read a character record into a KIND=4 character destination,
437 processing UTF-8 encoding if necessary. */
440 read_a_char4 (st_parameter_dt
*dtp
, const fnode
*f
, char *p
, int length
)
445 if (w
== -1) /* '(A)' edit descriptor */
448 /* Read in w characters, treating comma as not a separator. */
449 dtp
->u
.p
.sf_read_comma
= 0;
451 if (dtp
->u
.p
.current_unit
->flags
.encoding
== ENCODING_UTF8
)
452 read_utf8_char4 (dtp
, p
, length
, w
);
454 read_default_char4 (dtp
, p
, length
, w
);
456 dtp
->u
.p
.sf_read_comma
=
457 dtp
->u
.p
.current_unit
->decimal_status
== DECIMAL_COMMA
? 0 : 1;
460 /* eat_leading_spaces()-- Given a character pointer and a width,
461 * ignore the leading spaces. */
464 eat_leading_spaces (int *width
, char *p
)
468 if (*width
== 0 || *p
!= ' ')
480 next_char (st_parameter_dt
*dtp
, char **p
, int *w
)
495 if (dtp
->u
.p
.blank_status
!= BLANK_UNSPECIFIED
)
496 return ' '; /* return a blank to signal a null */
498 /* At this point, the rest of the field has to be trailing blanks */
512 /* read_decimal()-- Read a decimal integer value. The values here are
516 read_decimal (st_parameter_dt
*dtp
, const fnode
*f
, char *dest
, int length
)
518 GFC_UINTEGER_LARGEST value
, maxv
, maxv_10
;
519 GFC_INTEGER_LARGEST v
;
525 p
= read_block_form (dtp
, &w
);
530 p
= eat_leading_spaces (&w
, p
);
533 set_integer (dest
, (GFC_INTEGER_LARGEST
) 0, length
);
537 maxv
= max_value (length
, 1);
559 /* At this point we have a digit-string */
564 c
= next_char (dtp
, &p
, &w
);
570 if (dtp
->u
.p
.blank_status
== BLANK_NULL
) continue;
571 if (dtp
->u
.p
.blank_status
== BLANK_ZERO
) c
= '0';
574 if (c
< '0' || c
> '9')
577 if (value
> maxv_10
&& compile_options
.range_check
== 1)
583 if (value
> maxv
- c
&& compile_options
.range_check
== 1)
592 set_integer (dest
, v
, length
);
596 generate_error (&dtp
->common
, LIBERROR_READ_VALUE
,
597 "Bad value during integer read");
598 next_record (dtp
, 1);
602 generate_error (&dtp
->common
, LIBERROR_READ_OVERFLOW
,
603 "Value overflowed during integer read");
604 next_record (dtp
, 1);
609 /* read_radix()-- This function reads values for non-decimal radixes.
610 * The difference here is that we treat the values here as unsigned
611 * values for the purposes of overflow. If minus sign is present and
612 * the top bit is set, the value will be incorrect. */
615 read_radix (st_parameter_dt
*dtp
, const fnode
*f
, char *dest
, int length
,
618 GFC_UINTEGER_LARGEST value
, maxv
, maxv_r
;
619 GFC_INTEGER_LARGEST v
;
625 p
= read_block_form (dtp
, &w
);
630 p
= eat_leading_spaces (&w
, p
);
633 set_integer (dest
, (GFC_INTEGER_LARGEST
) 0, length
);
637 maxv
= max_value (length
, 0);
638 maxv_r
= maxv
/ radix
;
659 /* At this point we have a digit-string */
664 c
= next_char (dtp
, &p
, &w
);
669 if (dtp
->u
.p
.blank_status
== BLANK_NULL
) continue;
670 if (dtp
->u
.p
.blank_status
== BLANK_ZERO
) c
= '0';
676 if (c
< '0' || c
> '1')
681 if (c
< '0' || c
> '7')
706 c
= c
- 'a' + '9' + 1;
715 c
= c
- 'A' + '9' + 1;
729 value
= radix
* value
;
731 if (maxv
- c
< value
)
740 set_integer (dest
, v
, length
);
744 generate_error (&dtp
->common
, LIBERROR_READ_VALUE
,
745 "Bad value during integer read");
746 next_record (dtp
, 1);
750 generate_error (&dtp
->common
, LIBERROR_READ_OVERFLOW
,
751 "Value overflowed during integer read");
752 next_record (dtp
, 1);
757 /* read_f()-- Read a floating point number with F-style editing, which
758 is what all of the other floating point descriptors behave as. The
759 tricky part is that optional spaces are allowed after an E or D,
760 and the implicit decimal point if a decimal point is not present in
764 read_f (st_parameter_dt
*dtp
, const fnode
*f
, char *dest
, int length
)
766 int w
, seen_dp
, exponent
;
771 int seen_int_digit
; /* Seen a digit before the decimal point? */
772 int seen_dec_digit
; /* Seen a digit after the decimal point? */
781 /* Read in the next block. */
782 p
= read_block_form (dtp
, &w
);
785 p
= eat_leading_spaces (&w
, (char*) p
);
789 /* In this buffer we're going to re-format the number cleanly to be parsed
790 by convert_real in the end; this assures we're using strtod from the
791 C library for parsing and thus probably get the best accuracy possible.
792 This process may add a '+0.0' in front of the number as well as change the
793 exponent because of an implicit decimal point or the like. Thus allocating
794 strlen ("+0.0e-1000") == 10 characters plus one for NUL more than the
795 original buffer had should be enough. */
796 buffer
= gfc_alloca (w
+ 11);
800 if (*p
== '-' || *p
== '+')
808 p
= eat_leading_spaces (&w
, (char*) p
);
812 /* Process the mantissa string. */
818 if (dtp
->u
.p
.current_unit
->decimal_status
!= DECIMAL_COMMA
)
831 if (dtp
->u
.p
.blank_status
== BLANK_ZERO
)
836 else if (dtp
->u
.p
.blank_status
== BLANK_NULL
)
839 /* TODO: Should we check instead that there are only trailing
840 blanks here, as is done below for exponents? */
881 /* No exponent has been seen, so we use the current scale factor. */
882 exponent
= - dtp
->u
.p
.scale_factor
;
885 /* At this point the start of an exponent has been found. */
887 p
= eat_leading_spaces (&w
, (char*) p
);
888 if (*p
== '-' || *p
== '+')
896 /* At this point a digit string is required. We calculate the value
897 of the exponent in order to take account of the scale factor and
898 the d parameter before explict conversion takes place. */
903 if (dtp
->u
.p
.blank_status
== BLANK_UNSPECIFIED
)
905 while (w
> 0 && isdigit (*p
))
908 exponent
+= *p
- '0';
913 /* Only allow trailing blanks. */
922 else /* BZ or BN status is enabled. */
928 if (dtp
->u
.p
.blank_status
== BLANK_ZERO
)
931 assert (dtp
->u
.p
.blank_status
== BLANK_NULL
);
933 else if (!isdigit (*p
))
938 exponent
+= *p
- '0';
946 exponent
*= exponent_sign
;
949 /* Use the precision specified in the format if no decimal point has been
952 exponent
-= f
->u
.real
.d
;
954 /* Output a trailing '0' after decimal point if not yet found. */
955 if (seen_dp
&& !seen_dec_digit
)
958 /* Print out the exponent to finish the reformatted number. Maximum 4
959 digits for the exponent. */
968 exponent
= - exponent
;
971 assert (exponent
< 10000);
972 for (dig
= 3; dig
>= 0; --dig
)
974 out
[dig
] = (char) ('0' + exponent
% 10);
981 /* Do the actual conversion. */
982 convert_real (dtp
, dest
, buffer
, length
);
986 /* The value read is zero. */
991 *((GFC_REAL_4
*) dest
) = 0.0;
995 *((GFC_REAL_8
*) dest
) = 0.0;
998 #ifdef HAVE_GFC_REAL_10
1000 *((GFC_REAL_10
*) dest
) = 0.0;
1004 #ifdef HAVE_GFC_REAL_16
1006 *((GFC_REAL_16
*) dest
) = 0.0;
1011 internal_error (&dtp
->common
, "Unsupported real kind during IO");
1016 generate_error (&dtp
->common
, LIBERROR_READ_VALUE
,
1017 "Bad value during floating point read");
1018 next_record (dtp
, 1);
1023 /* read_x()-- Deal with the X/TR descriptor. We just read some data
1024 * and never look at it. */
1027 read_x (st_parameter_dt
*dtp
, int n
)
1032 if ((dtp
->u
.p
.current_unit
->pad_status
== PAD_NO
|| is_internal_unit (dtp
))
1033 && dtp
->u
.p
.current_unit
->bytes_left
< n
)
1034 n
= dtp
->u
.p
.current_unit
->bytes_left
;
1041 if (is_internal_unit (dtp
))
1043 p
= mem_alloc_r (dtp
->u
.p
.current_unit
->s
, &length
);
1044 if (unlikely (length
< n
))
1049 p
= fbuf_read (dtp
->u
.p
.current_unit
, &length
);
1056 if (length
== 0 && dtp
->u
.p
.item_count
== 1)
1058 if (dtp
->u
.p
.current_unit
->pad_status
== PAD_NO
)
1071 if (q
== '\n' || q
== '\r')
1073 /* Unexpected end of line. Set the position. */
1074 fbuf_seek (dtp
->u
.p
.current_unit
, n
+ 1 ,SEEK_CUR
);
1075 dtp
->u
.p
.sf_seen_eor
= 1;
1077 /* If we encounter a CR, it might be a CRLF. */
1078 if (q
== '\r') /* Probably a CRLF */
1080 /* See if there is an LF. Use fbuf_read rather then fbuf_getc so
1081 the position is not advanced unless it really is an LF. */
1083 p
= fbuf_read (dtp
->u
.p
.current_unit
, &readlen
);
1084 if (*p
== '\n' && readlen
== 1)
1086 dtp
->u
.p
.sf_seen_eor
= 2;
1087 fbuf_seek (dtp
->u
.p
.current_unit
, 1 ,SEEK_CUR
);
1096 fbuf_seek (dtp
->u
.p
.current_unit
, n
, SEEK_CUR
);
1099 if ((dtp
->common
.flags
& IOPARM_DT_HAS_SIZE
) != 0)
1100 dtp
->u
.p
.size_used
+= (GFC_IO_INT
) n
;
1101 dtp
->u
.p
.current_unit
->bytes_left
-= n
;
1102 dtp
->u
.p
.current_unit
->strm_pos
+= (gfc_offset
) n
;