1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010 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 "data/sys-file-reader.h"
20 #include "data/sys-file-private.h"
28 #include "data/attributes.h"
29 #include "data/case.h"
30 #include "data/casereader-provider.h"
31 #include "data/casereader.h"
32 #include "data/dictionary.h"
33 #include "data/file-handle-def.h"
34 #include "data/file-name.h"
35 #include "data/format.h"
36 #include "data/missing-values.h"
37 #include "data/mrset.h"
38 #include "data/short-names.h"
39 #include "data/value-labels.h"
40 #include "data/value.h"
41 #include "data/variable.h"
42 #include "libpspp/array.h"
43 #include "libpspp/assertion.h"
44 #include "libpspp/compiler.h"
45 #include "libpspp/i18n.h"
46 #include "libpspp/message.h"
47 #include "libpspp/misc.h"
48 #include "libpspp/pool.h"
49 #include "libpspp/str.h"
50 #include "libpspp/stringi-set.h"
52 #include "gl/c-ctype.h"
53 #include "gl/inttostr.h"
54 #include "gl/minmax.h"
55 #include "gl/unlocked-io.h"
56 #include "gl/xalloc.h"
60 #define _(msgid) gettext (msgid)
61 #define N_(msgid) (msgid)
63 /* System file reader. */
66 /* Resource tracking. */
67 struct pool *pool; /* All system file state. */
68 jmp_buf bail_out; /* longjmp() target for error handling. */
71 struct file_handle *fh; /* File handle. */
72 struct fh_lock *lock; /* Mutual exclusion for file handle. */
73 FILE *file; /* File stream. */
74 bool error; /* I/O or corruption error? */
75 struct caseproto *proto; /* Format of output cases. */
78 enum integer_format integer_format; /* On-disk integer format. */
79 enum float_format float_format; /* On-disk floating point format. */
80 int oct_cnt; /* Number of 8-byte units per case. */
81 struct sfm_var *sfm_vars; /* Variables. */
82 size_t sfm_var_cnt; /* Number of variables. */
83 casenumber case_cnt; /* Number of cases */
84 bool has_long_var_names; /* File has a long variable name map */
87 bool compressed; /* File is compressed? */
88 double bias; /* Compression bias, usually 100.0. */
89 uint8_t opcodes[8]; /* Current block of opcodes. */
90 size_t opcode_idx; /* Next opcode to interpret, 8 if none left. */
91 bool corruption_warning; /* Warned about possible corruption? */
94 static const struct casereader_class sys_file_casereader_class;
96 static bool close_reader (struct sfm_reader *);
98 static struct variable **make_var_by_value_idx (struct sfm_reader *,
100 static struct variable *lookup_var_by_value_idx (struct sfm_reader *,
103 static struct variable *lookup_var_by_short_name (struct dictionary *,
104 const char *short_name);
106 static void sys_msg (struct sfm_reader *r, int class,
107 const char *format, va_list args)
108 PRINTF_FORMAT (3, 0);
109 static void sys_warn (struct sfm_reader *, const char *, ...)
110 PRINTF_FORMAT (2, 3);
111 static void sys_error (struct sfm_reader *, const char *, ...)
115 static void read_bytes (struct sfm_reader *, void *, size_t);
116 static bool try_read_bytes (struct sfm_reader *, void *, size_t);
117 static int read_int (struct sfm_reader *);
118 static double read_float (struct sfm_reader *);
119 static void read_string (struct sfm_reader *, char *, size_t);
120 static void skip_bytes (struct sfm_reader *, size_t);
122 static struct text_record *open_text_record (struct sfm_reader *, size_t size);
123 static void close_text_record (struct sfm_reader *r,
124 struct text_record *);
125 static bool read_variable_to_value_pair (struct sfm_reader *,
127 struct text_record *,
128 struct variable **var, char **value);
129 static void text_warn (struct sfm_reader *r, struct text_record *text,
130 const char *format, ...)
131 PRINTF_FORMAT (3, 4);
132 static char *text_get_token (struct text_record *,
133 struct substring delimiters, char *delimiter);
134 static bool text_match (struct text_record *, char c);
135 static bool text_read_variable_name (struct sfm_reader *, struct dictionary *,
136 struct text_record *,
137 struct substring delimiters,
139 static bool text_read_short_name (struct sfm_reader *, struct dictionary *,
140 struct text_record *,
141 struct substring delimiters,
143 static const char *text_parse_counted_string (struct sfm_reader *,
144 struct text_record *);
145 static size_t text_pos (const struct text_record *);
147 static bool close_reader (struct sfm_reader *r);
149 /* Dictionary reader. */
157 static void read_header (struct sfm_reader *, struct dictionary *,
158 int *weight_idx, int *claimed_oct_cnt,
159 struct sfm_read_info *);
160 static void read_variable_record (struct sfm_reader *, struct dictionary *,
161 int *format_warning_cnt);
162 static void parse_format_spec (struct sfm_reader *, unsigned int,
163 enum which_format, struct variable *,
164 int *format_warning_cnt);
165 static void setup_weight (struct sfm_reader *, int weight_idx,
166 struct variable **var_by_value_idx,
167 struct dictionary *);
168 static void read_documents (struct sfm_reader *, struct dictionary *);
169 static void read_value_labels (struct sfm_reader *, struct dictionary *,
170 struct variable **var_by_value_idx);
172 static void read_extension_record (struct sfm_reader *, struct dictionary *,
173 struct sfm_read_info *);
174 static void read_machine_integer_info (struct sfm_reader *,
175 size_t size, size_t count,
176 struct sfm_read_info *,
179 static void read_machine_float_info (struct sfm_reader *,
180 size_t size, size_t count);
181 static void read_mrsets (struct sfm_reader *, size_t size, size_t count,
182 struct dictionary *);
183 static void read_display_parameters (struct sfm_reader *,
184 size_t size, size_t count,
185 struct dictionary *);
186 static void read_long_var_name_map (struct sfm_reader *,
187 size_t size, size_t count,
188 struct dictionary *);
189 static void read_long_string_map (struct sfm_reader *,
190 size_t size, size_t count,
191 struct dictionary *);
192 static void read_data_file_attributes (struct sfm_reader *,
193 size_t size, size_t count,
194 struct dictionary *);
195 static void read_variable_attributes (struct sfm_reader *,
196 size_t size, size_t count,
197 struct dictionary *);
198 static void read_long_string_value_labels (struct sfm_reader *,
199 size_t size, size_t count,
200 struct dictionary *);
202 /* Convert all the strings in DICT from the dict encoding to UTF8 */
204 recode_strings (struct dictionary *dict)
208 const char *enc = dict_get_encoding (dict);
211 enc = get_default_encoding ();
213 for (i = 0 ; i < dict_get_var_cnt (dict); ++i)
215 /* Convert the long variable name */
216 struct variable *var = dict_get_var (dict, i);
217 const char *native_name = var_get_name (var);
218 char *utf8_name = recode_string (UTF8, enc, native_name, -1);
219 if ( 0 != strcmp (utf8_name, native_name))
221 if ( NULL == dict_lookup_var (dict, utf8_name))
222 dict_rename_var (dict, var, utf8_name);
225 _("Recoded variable name duplicates an existing `%s' within system file."), utf8_name);
230 /* Convert the variable label */
231 if (var_has_label (var))
233 char *utf8_label = recode_string (UTF8, enc, var_get_label (var), -1);
234 var_set_label (var, utf8_label);
238 if (var_has_value_labels (var))
240 const struct val_lab *vl = NULL;
241 const struct val_labs *vlabs = var_get_value_labels (var);
243 for (vl = val_labs_first (vlabs); vl != NULL; vl = val_labs_next (vlabs, vl))
245 const union value *val = val_lab_get_value (vl);
246 const char *label = val_lab_get_label (vl);
247 char *new_label = NULL;
249 new_label = recode_string (UTF8, enc, label, -1);
251 var_replace_value_label (var, val, new_label);
258 /* Opens the system file designated by file handle FH for
259 reading. Reads the system file's dictionary into *DICT.
260 If INFO is non-null, then it receives additional info about the
263 sfm_open_reader (struct file_handle *fh, struct dictionary **dict,
264 struct sfm_read_info *volatile info)
266 struct sfm_reader *volatile r = NULL;
267 struct variable **var_by_value_idx;
268 struct sfm_read_info local_info;
269 int format_warning_cnt = 0;
274 *dict = dict_create ();
276 /* Create and initialize reader. */
277 r = pool_create_container (struct sfm_reader, pool);
283 r->has_long_var_names = false;
284 r->opcode_idx = sizeof r->opcodes;
285 r->corruption_warning = false;
287 /* TRANSLATORS: this fragment will be interpolated into
288 messages in fh_lock() that identify types of files. */
289 r->lock = fh_lock (fh, FH_REF_FILE, N_("system file"), FH_ACC_READ, false);
293 r->file = fn_open (fh_get_file_name (fh), "rb");
296 msg (ME, _("Error opening \"%s\" for reading as a system file: %s."),
297 fh_get_file_name (r->fh), strerror (errno));
301 /* Initialize info. */
304 memset (info, 0, sizeof *info);
306 if (setjmp (r->bail_out))
311 read_header (r, *dict, &weight_idx, &claimed_oct_cnt, info);
313 /* Read all the variable definition records. */
314 rec_type = read_int (r);
315 while (rec_type == 2)
317 read_variable_record (r, *dict, &format_warning_cnt);
318 rec_type = read_int (r);
321 /* Figure out the case format. */
322 var_by_value_idx = make_var_by_value_idx (r, *dict);
323 setup_weight (r, weight_idx, var_by_value_idx, *dict);
325 /* Read all the rest of the dictionary records. */
326 while (rec_type != 999)
331 read_value_labels (r, *dict, var_by_value_idx);
335 sys_error (r, _("Misplaced type 4 record."));
338 read_documents (r, *dict);
342 read_extension_record (r, *dict, info);
346 sys_error (r, _("Unrecognized record type %d."), rec_type);
348 rec_type = read_int (r);
352 if ( ! r->has_long_var_names )
355 for (i = 0; i < dict_get_var_cnt (*dict); i++)
357 struct variable *var = dict_get_var (*dict, i);
358 char short_name[SHORT_NAME_LEN + 1];
359 char long_name[SHORT_NAME_LEN + 1];
361 strcpy (short_name, var_get_name (var));
363 strcpy (long_name, short_name);
364 str_lowercase (long_name);
366 /* Set long name. Renaming a variable may clear the short
367 name, but we want to retain it, so re-set it
369 dict_rename_var (*dict, var, long_name);
370 var_set_short_name (var, 0, short_name);
373 r->has_long_var_names = true;
376 recode_strings (*dict);
378 /* Read record 999 data, which is just filler. */
381 /* Warn if the actual amount of data per case differs from the
382 amount that the header claims. SPSS version 13 gets this
383 wrong when very long strings are involved, so don't warn in
385 if (claimed_oct_cnt != -1 && claimed_oct_cnt != r->oct_cnt
386 && info->version_major != 13)
387 sys_warn (r, _("File header claims %d variable positions but "
388 "%d were read from file."),
389 claimed_oct_cnt, r->oct_cnt);
391 /* Create an index of dictionary variable widths for
392 sfm_read_case to use. We cannot use the `struct variable's
393 from the dictionary we created, because the caller owns the
394 dictionary and may destroy or modify its variables. */
395 sfm_dictionary_to_sfm_vars (*dict, &r->sfm_vars, &r->sfm_var_cnt);
396 pool_register (r->pool, free, r->sfm_vars);
397 r->proto = caseproto_ref_pool (dict_get_proto (*dict), r->pool);
399 pool_free (r->pool, var_by_value_idx);
400 return casereader_create_sequential
402 r->case_cnt == -1 ? CASENUMBER_MAX: r->case_cnt,
403 &sys_file_casereader_class, r);
407 dict_destroy (*dict);
412 /* Closes a system file after we're done with it.
413 Returns true if an I/O error has occurred on READER, false
416 close_reader (struct sfm_reader *r)
425 if (fn_close (fh_get_file_name (r->fh), r->file) == EOF)
427 msg (ME, _("Error closing system file \"%s\": %s."),
428 fh_get_file_name (r->fh), strerror (errno));
438 pool_destroy (r->pool);
443 /* Destroys READER. */
445 sys_file_casereader_destroy (struct casereader *reader UNUSED, void *r_)
447 struct sfm_reader *r = r_;
451 /* Returns true if FILE is an SPSS system file,
454 sfm_detect (FILE *file)
458 if (fread (rec_type, 4, 1, file) != 1)
462 return !strcmp ("$FL2", rec_type);
465 /* Reads the global header of the system file.
466 Sets DICT's file label to the system file's label.
467 Sets *WEIGHT_IDX to 0 if the system file is unweighted,
468 or to the value index of the weight variable otherwise.
469 Sets *CLAIMED_OCT_CNT to the number of "octs" (8-byte units)
470 per case that the file claims to have (although it is not
472 Initializes INFO with header information. */
474 read_header (struct sfm_reader *r, struct dictionary *dict,
475 int *weight_idx, int *claimed_oct_cnt,
476 struct sfm_read_info *info)
479 char eye_catcher[61];
480 uint8_t raw_layout_code[4];
482 char creation_date[10];
483 char creation_time[9];
485 struct substring file_label_ss;
486 struct substring product;
488 read_string (r, rec_type, sizeof rec_type);
489 read_string (r, eye_catcher, sizeof eye_catcher);
491 if (strcmp ("$FL2", rec_type) != 0)
492 sys_error (r, _("This is not an SPSS system file."));
494 /* Identify integer format. */
495 read_bytes (r, raw_layout_code, sizeof raw_layout_code);
496 if ((!integer_identify (2, raw_layout_code, sizeof raw_layout_code,
498 && !integer_identify (3, raw_layout_code, sizeof raw_layout_code,
500 || (r->integer_format != INTEGER_MSB_FIRST
501 && r->integer_format != INTEGER_LSB_FIRST))
502 sys_error (r, _("This is not an SPSS system file."));
504 *claimed_oct_cnt = read_int (r);
505 if (*claimed_oct_cnt < 0 || *claimed_oct_cnt > INT_MAX / 16)
506 *claimed_oct_cnt = -1;
508 r->compressed = read_int (r) != 0;
510 *weight_idx = read_int (r);
512 r->case_cnt = read_int (r);
513 if ( r->case_cnt > INT_MAX / 2)
517 /* Identify floating-point format and obtain compression bias. */
518 read_bytes (r, raw_bias, sizeof raw_bias);
519 if (float_identify (100.0, raw_bias, sizeof raw_bias, &r->float_format) == 0)
521 uint8_t zero_bias[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
523 if (memcmp (raw_bias, zero_bias, 8))
524 sys_warn (r, _("Compression bias is not the usual "
525 "value of 100, or system file uses unrecognized "
526 "floating-point format."));
529 /* Some software is known to write all-zeros to this
530 field. Such software also writes floating-point
531 numbers in the format that we expect by default
532 (it seems that all software most likely does, in
533 reality), so don't warn in this case. */
536 if (r->integer_format == INTEGER_MSB_FIRST)
537 r->float_format = FLOAT_IEEE_DOUBLE_BE;
539 r->float_format = FLOAT_IEEE_DOUBLE_LE;
541 float_convert (r->float_format, raw_bias, FLOAT_NATIVE_DOUBLE, &r->bias);
543 read_string (r, creation_date, sizeof creation_date);
544 read_string (r, creation_time, sizeof creation_time);
545 read_string (r, file_label, sizeof file_label);
548 file_label_ss = ss_cstr (file_label);
549 ss_trim (&file_label_ss, ss_cstr (" "));
550 if (!ss_is_empty (file_label_ss))
552 ss_data (file_label_ss)[ss_length (file_label_ss)] = '\0';
553 dict_set_label (dict, ss_data (file_label_ss));
556 strcpy (info->creation_date, creation_date);
557 strcpy (info->creation_time, creation_time);
558 info->integer_format = r->integer_format;
559 info->float_format = r->float_format;
560 info->compressed = r->compressed;
561 info->case_cnt = r->case_cnt;
563 product = ss_cstr (eye_catcher);
564 ss_match_string (&product, ss_cstr ("@(#) SPSS DATA FILE"));
565 ss_trim (&product, ss_cstr (" "));
566 str_copy_buf_trunc (info->product, sizeof info->product,
567 ss_data (product), ss_length (product));
570 /* Reads a variable (type 2) record from R and adds the
571 corresponding variable to DICT.
572 Also skips past additional variable records for long string
575 read_variable_record (struct sfm_reader *r, struct dictionary *dict,
576 int *format_warning_cnt)
579 int has_variable_label;
580 int missing_value_code;
585 struct variable *var;
588 width = read_int (r);
589 has_variable_label = read_int (r);
590 missing_value_code = read_int (r);
591 print_format = read_int (r);
592 write_format = read_int (r);
593 read_string (r, name, sizeof name);
594 name[strcspn (name, " ")] = '\0';
596 /* Check variable name. */
597 if (name[0] == '$' || name[0] == '#')
598 sys_error (r, "Variable name begins with invalid character `%c'.",
600 if (!var_is_plausible_name (name, false))
601 sys_error (r, _("Invalid variable name `%s'."), name);
603 /* Create variable. */
604 if (width < 0 || width > 255)
605 sys_error (r, _("Bad width %d for variable %s."), width, name);
606 var = dict_create_var (dict, name, width);
609 _("Duplicate variable name `%s' within system file."),
612 /* Set the short name the same as the long name. */
613 var_set_short_name (var, 0, var_get_name (var));
615 /* Get variable label, if any. */
616 if (has_variable_label != 0 && has_variable_label != 1)
617 sys_error (r, _("Variable label indicator field is not 0 or 1."));
618 if (has_variable_label == 1)
620 size_t len, read_len;
625 /* Read up to 255 bytes of label. */
626 read_len = MIN (sizeof label - 1, len);
627 read_string (r, label, read_len + 1);
628 var_set_label (var, label);
630 /* Skip unread label bytes. */
631 skip_bytes (r, len - read_len);
633 /* Skip label padding up to multiple of 4 bytes. */
634 skip_bytes (r, ROUND_UP (len, 4) - len);
637 /* Set missing values. */
638 if (missing_value_code != 0)
640 struct missing_values mv;
643 mv_init_pool (r->pool, &mv, var_get_width (var));
644 if (var_is_numeric (var))
646 if (missing_value_code < -3 || missing_value_code > 3
647 || missing_value_code == -1)
648 sys_error (r, _("Numeric missing value indicator field is not "
649 "-3, -2, 0, 1, 2, or 3."));
650 if (missing_value_code < 0)
652 double low = read_float (r);
653 double high = read_float (r);
654 mv_add_range (&mv, low, high);
655 missing_value_code = -missing_value_code - 2;
657 for (i = 0; i < missing_value_code; i++)
658 mv_add_num (&mv, read_float (r));
662 int mv_width = MAX (width, 8);
665 if (missing_value_code < 1 || missing_value_code > 3)
666 sys_error (r, _("String missing value indicator field is not "
669 value_init (&value, mv_width);
670 value_set_missing (&value, mv_width);
671 for (i = 0; i < missing_value_code; i++)
673 uint8_t *s = value_str_rw (&value, mv_width);
674 read_bytes (r, s, 8);
677 value_destroy (&value, mv_width);
679 var_set_missing_values (var, &mv);
683 parse_format_spec (r, print_format, PRINT_FORMAT, var, format_warning_cnt);
684 parse_format_spec (r, write_format, WRITE_FORMAT, var, format_warning_cnt);
686 /* Account for values.
687 Skip long string continuation records, if any. */
688 nv = width == 0 ? 1 : DIV_RND_UP (width, 8);
694 for (i = 1; i < nv; i++)
696 /* Check for record type 2 and width -1. */
697 if (read_int (r) != 2 || read_int (r) != -1)
698 sys_error (r, _("Missing string continuation record."));
700 /* Skip and ignore remaining continuation data. */
701 has_variable_label = read_int (r);
702 missing_value_code = read_int (r);
703 print_format = read_int (r);
704 write_format = read_int (r);
705 read_string (r, name, sizeof name);
707 /* Variable label fields on continuation records have
708 been spotted in system files created by "SPSS Power
709 Macintosh Release 6.1". */
710 if (has_variable_label)
711 skip_bytes (r, ROUND_UP (read_int (r), 4));
716 /* Translates the format spec from sysfile format to internal
719 parse_format_spec (struct sfm_reader *r, unsigned int s,
720 enum which_format which, struct variable *v,
721 int *format_warning_cnt)
723 const int max_format_warnings = 8;
725 uint8_t raw_type = s >> 16;
731 if (!fmt_from_io (raw_type, &f.type))
732 sys_error (r, _("Unknown variable format %"PRIu8"."), raw_type);
737 ok = fmt_check_output (&f) && fmt_check_width_compat (&f, var_get_width (v));
742 if (which == PRINT_FORMAT)
743 var_set_print_format (v, &f);
745 var_set_write_format (v, &f);
747 else if (*++format_warning_cnt <= max_format_warnings)
749 char fmt_string[FMT_STRING_LEN_MAX + 1];
750 sys_warn (r, _("%s variable %s has invalid %s format %s."),
751 var_is_numeric (v) ? _("Numeric") : _("String"),
753 which == PRINT_FORMAT ? _("print") : _("write"),
754 fmt_to_string (&f, fmt_string));
756 if (*format_warning_cnt == max_format_warnings)
757 sys_warn (r, _("Suppressing further invalid format warnings."));
761 /* Sets the weighting variable in DICT to the variable
762 corresponding to the given 1-based VALUE_IDX, if VALUE_IDX is
765 setup_weight (struct sfm_reader *r, int weight_idx,
766 struct variable **var_by_value_idx, struct dictionary *dict)
770 struct variable *weight_var
771 = lookup_var_by_value_idx (r, var_by_value_idx, weight_idx);
772 if (var_is_numeric (weight_var))
773 dict_set_weight (dict, weight_var);
775 sys_error (r, _("Weighting variable must be numeric."));
779 /* Reads a document record, type 6, from system file R, and sets up
780 the documents and n_documents fields in the associated
783 read_documents (struct sfm_reader *r, struct dictionary *dict)
788 if (dict_get_documents (dict) != NULL)
789 sys_error (r, _("Multiple type 6 (document) records."));
791 line_cnt = read_int (r);
793 sys_error (r, _("Number of document lines (%d) "
794 "must be greater than 0."), line_cnt);
796 documents = pool_nmalloc (r->pool, line_cnt + 1, DOC_LINE_LENGTH);
797 read_string (r, documents, DOC_LINE_LENGTH * line_cnt + 1);
798 if (strlen (documents) == DOC_LINE_LENGTH * line_cnt)
799 dict_set_documents (dict, documents);
801 sys_error (r, _("Document line contains null byte."));
802 pool_free (r->pool, documents);
805 /* Read a type 7 extension record. */
807 read_extension_record (struct sfm_reader *r, struct dictionary *dict,
808 struct sfm_read_info *info)
810 int subtype = read_int (r);
811 size_t size = read_int (r);
812 size_t count = read_int (r);
813 size_t bytes = size * count;
815 /* Check that SIZE * COUNT + 1 doesn't overflow. Adding 1
816 allows an extra byte for a null terminator, used by some
817 extension processing routines. */
818 if (size != 0 && size_overflow_p (xsum (1, xtimes (count, size))))
819 sys_error (r, "Record type 7 subtype %d too large.", subtype);
824 read_machine_integer_info (r, size, count, info, dict);
828 read_machine_float_info (r, size, count);
832 /* Variable sets information. We don't use these yet.
833 They only apply to GUIs; see VARSETS on the APPLY
834 DICTIONARY command in SPSS documentation. */
838 /* DATE variable information. We don't use it yet, but we
844 read_mrsets (r, size, count, dict);
848 /* Used by the SPSS Data Entry software. */
852 read_display_parameters (r, size, count, dict);
856 read_long_var_name_map (r, size, count, dict);
860 read_long_string_map (r, size, count, dict);
864 /* Extended number of cases. Not important. */
868 read_data_file_attributes (r, size, count, dict);
872 read_variable_attributes (r, size, count, dict);
876 /* New in SPSS 16. Contains a single string that describes
877 the character encoding, e.g. "windows-1252". */
879 char *encoding = pool_calloc (r->pool, size, count + 1);
880 read_string (r, encoding, count + 1);
881 dict_set_encoding (dict, encoding);
886 /* New in SPSS 16. Encodes value labels for long string
888 read_long_string_value_labels (r, size, count, dict);
892 sys_warn (r, _("Unrecognized record type 7, subtype %d. Please send a copy of this file, and the syntax which created it to %s"),
893 subtype, PACKAGE_BUGREPORT);
897 skip_bytes (r, bytes);
900 /* Read record type 7, subtype 3. */
902 read_machine_integer_info (struct sfm_reader *r, size_t size, size_t count,
903 struct sfm_read_info *info,
904 struct dictionary *dict)
906 int version_major = read_int (r);
907 int version_minor = read_int (r);
908 int version_revision = read_int (r);
909 int machine_code UNUSED = read_int (r);
910 int float_representation = read_int (r);
911 int compression_code UNUSED = read_int (r);
912 int integer_representation = read_int (r);
913 int character_code = read_int (r);
915 int expected_float_format;
916 int expected_integer_format;
918 if (size != 4 || count != 8)
919 sys_error (r, _("Bad size (%zu) or count (%zu) field on record type 7, "
923 /* Save version info. */
924 info->version_major = version_major;
925 info->version_minor = version_minor;
926 info->version_revision = version_revision;
928 /* Check floating point format. */
929 if (r->float_format == FLOAT_IEEE_DOUBLE_BE
930 || r->float_format == FLOAT_IEEE_DOUBLE_LE)
931 expected_float_format = 1;
932 else if (r->float_format == FLOAT_Z_LONG)
933 expected_float_format = 2;
934 else if (r->float_format == FLOAT_VAX_G || r->float_format == FLOAT_VAX_D)
935 expected_float_format = 3;
938 if (float_representation != expected_float_format)
939 sys_error (r, _("Floating-point representation indicated by "
940 "system file (%d) differs from expected (%d)."),
941 r->float_format, expected_float_format);
943 /* Check integer format. */
944 if (r->integer_format == INTEGER_MSB_FIRST)
945 expected_integer_format = 1;
946 else if (r->integer_format == INTEGER_LSB_FIRST)
947 expected_integer_format = 2;
950 if (integer_representation != expected_integer_format)
952 static const char *const endian[] = {N_("Little Endian"), N_("Big Endian")};
953 sys_warn (r, _("Integer format indicated by system file (%s) "
954 "differs from expected (%s)."),
955 gettext (endian[integer_representation == 1]),
956 gettext (endian[expected_integer_format == 1]));
961 Record 7 (20) provides a much more reliable way of
962 setting the encoding.
963 The character_code is used as a fallback only.
965 if ( NULL == dict_get_encoding (dict))
967 switch (character_code)
970 dict_set_encoding (dict, "EBCDIC-US");
974 /* These ostensibly mean "7-bit ASCII" and "8-bit ASCII"[sic]
975 respectively. However, there are known to be many files
976 in the wild with character code 2, yet have data which are
978 Therefore we ignore these values.
982 dict_set_encoding (dict, "MS_KANJI");
985 dict_set_encoding (dict, "UTF-7");
988 dict_set_encoding (dict, "UTF-8");
993 snprintf (enc, 100, "CP%d", character_code);
994 dict_set_encoding (dict, enc);
1001 /* Read record type 7, subtype 4. */
1003 read_machine_float_info (struct sfm_reader *r, size_t size, size_t count)
1005 double sysmis = read_float (r);
1006 double highest = read_float (r);
1007 double lowest = read_float (r);
1009 if (size != 8 || count != 3)
1010 sys_error (r, _("Bad size (%zu) or count (%zu) on extension 4."),
1013 if (sysmis != SYSMIS)
1014 sys_warn (r, _("File specifies unexpected value %g as %s."),
1017 if (highest != HIGHEST)
1018 sys_warn (r, _("File specifies unexpected value %g as %s."),
1019 highest, "HIGHEST");
1021 if (lowest != LOWEST)
1022 sys_warn (r, _("File specifies unexpected value %g as %s."),
1026 /* Read record type 7, subtype 7 or 19. */
1028 read_mrsets (struct sfm_reader *r, size_t size, size_t count,
1029 struct dictionary *dict)
1031 struct text_record *text;
1032 struct mrset *mrset;
1034 text = open_text_record (r, size * count);
1037 const char *name, *label, *counted;
1038 struct stringi_set var_names;
1039 size_t allocated_vars;
1043 mrset = xzalloc (sizeof *mrset);
1045 name = text_get_token (text, ss_cstr ("="), NULL);
1048 mrset->name = xstrdup (name);
1050 if (text_match (text, 'C'))
1052 mrset->type = MRSET_MC;
1053 if (!text_match (text, ' '))
1055 sys_warn (r, _("Missing space following 'C' at offset %zu "
1056 "in MRSETS record"), text_pos (text));
1060 else if (text_match (text, 'D'))
1062 mrset->type = MRSET_MD;
1063 mrset->cat_source = MRSET_VARLABELS;
1065 else if (text_match (text, 'E'))
1069 mrset->type = MRSET_MD;
1070 mrset->cat_source = MRSET_COUNTEDVALUES;
1071 if (!text_match (text, ' '))
1073 sys_warn (r, _("Missing space following 'E' at offset %zu "
1074 "in MRSETS record"), text_pos (text));
1078 number = text_get_token (text, ss_cstr (" "), NULL);
1079 if (!strcmp (number, "11"))
1080 mrset->label_from_var_label = true;
1081 else if (strcmp (number, "1"))
1082 sys_warn (r, _("Unexpected label source value \"%s\" "
1083 "following 'E' at offset %zu in MRSETS record"),
1084 number, text_pos (text));
1088 sys_warn (r, _("Missing 'C', 'D', or 'E' at offset %zu "
1089 "in MRSETS record."),
1094 if (mrset->type == MRSET_MD)
1096 counted = text_parse_counted_string (r, text);
1097 if (counted == NULL)
1101 label = text_parse_counted_string (r, text);
1104 mrset->label = label[0] != '\0' ? xstrdup (label) : NULL;
1106 stringi_set_init (&var_names);
1111 struct variable *var;
1112 const char *var_name;
1114 var_name = text_get_token (text, ss_cstr (" \n"), &delimiter);
1115 if (var_name == NULL)
1117 sys_warn (r, _("Missing new-line parsing variable names "
1118 "at offset %zu in MRSETS record."),
1123 var = lookup_var_by_short_name (dict, var_name);
1126 if (!stringi_set_insert (&var_names, var_name))
1128 sys_warn (r, _("Duplicate variable name %s "
1129 "at offset %zu in MRSETS record."),
1130 var_name, text_pos (text));
1134 if (mrset->label == NULL && mrset->label_from_var_label
1135 && var_has_label (var))
1136 mrset->label = xstrdup (var_get_label (var));
1139 && var_get_type (var) != var_get_type (mrset->vars[0]))
1141 sys_warn (r, _("MRSET %s contains both string and "
1142 "numeric variables."), name);
1145 width = MIN (width, var_get_width (var));
1147 if (mrset->n_vars >= allocated_vars)
1148 mrset->vars = x2nrealloc (mrset->vars, &allocated_vars,
1149 sizeof *mrset->vars);
1150 mrset->vars[mrset->n_vars++] = var;
1152 while (delimiter != '\n');
1154 if (mrset->n_vars < 2)
1156 sys_warn (r, _("MRSET %s has only %zu variables."), mrset->name,
1158 mrset_destroy (mrset);
1162 if (mrset->type == MRSET_MD)
1164 mrset->width = width;
1165 value_init (&mrset->counted, width);
1167 mrset->counted.f = strtod (counted, NULL);
1169 value_copy_str_rpad (&mrset->counted, width,
1170 (const uint8_t *) counted, ' ');
1173 dict_add_mrset (dict, mrset);
1176 mrset_destroy (mrset);
1177 close_text_record (r, text);
1180 /* Read record type 7, subtype 11, which specifies how variables
1181 should be displayed in GUI environments. */
1183 read_display_parameters (struct sfm_reader *r, size_t size, size_t count,
1184 struct dictionary *dict)
1187 bool includes_width;
1188 bool warned = false;
1193 sys_warn (r, _("Bad size %zu on extension 11."), size);
1194 skip_bytes (r, size * count);
1198 n_vars = dict_get_var_cnt (dict);
1199 if (count == 3 * n_vars)
1200 includes_width = true;
1201 else if (count == 2 * n_vars)
1202 includes_width = false;
1205 sys_warn (r, _("Extension 11 has bad count %zu (for %zu variables)."),
1207 skip_bytes (r, size * count);
1211 for (i = 0; i < n_vars; ++i)
1213 struct variable *v = dict_get_var (dict, i);
1214 int measure = read_int (r);
1215 int width = includes_width ? read_int (r) : 0;
1216 int align = read_int (r);
1218 /* SPSS 14 sometimes seems to set string variables' measure
1220 if (0 == measure && var_is_alpha (v))
1223 if (measure < 1 || measure > 3 || align < 0 || align > 2)
1226 sys_warn (r, _("Invalid variable display parameters "
1227 "for variable %zu (%s). "
1228 "Default parameters substituted."),
1229 i, var_get_name (v));
1234 var_set_measure (v, (measure == 1 ? MEASURE_NOMINAL
1235 : measure == 2 ? MEASURE_ORDINAL
1237 var_set_alignment (v, (align == 0 ? ALIGN_LEFT
1238 : align == 1 ? ALIGN_RIGHT
1241 /* Older versions (SPSS 9.0) sometimes set the display
1242 width to zero. This causes confusion in the GUI, so
1243 only set the width if it is nonzero. */
1245 var_set_display_width (v, width);
1249 /* Reads record type 7, subtype 13, which gives the long name
1250 that corresponds to each short name. Modifies variable names
1251 in DICT accordingly. */
1253 read_long_var_name_map (struct sfm_reader *r, size_t size, size_t count,
1254 struct dictionary *dict)
1256 struct text_record *text;
1257 struct variable *var;
1260 text = open_text_record (r, size * count);
1261 while (read_variable_to_value_pair (r, dict, text, &var, &long_name))
1264 size_t short_name_cnt;
1267 /* Validate long name. */
1268 if (!var_is_valid_name (long_name, false))
1270 sys_warn (r, _("Long variable mapping from %s to invalid "
1271 "variable name `%s'."),
1272 var_get_name (var), long_name);
1276 /* Identify any duplicates. */
1277 if (strcasecmp (var_get_short_name (var, 0), long_name)
1278 && dict_lookup_var (dict, long_name) != NULL)
1280 sys_warn (r, _("Duplicate long variable name `%s' "
1281 "within system file."), long_name);
1285 /* Renaming a variable may clear its short names, but we
1286 want to retain them, so we save them and re-set them
1288 short_name_cnt = var_get_short_name_cnt (var);
1289 short_names = xnmalloc (short_name_cnt, sizeof *short_names);
1290 for (i = 0; i < short_name_cnt; i++)
1292 const char *s = var_get_short_name (var, i);
1293 short_names[i] = s != NULL ? xstrdup (s) : NULL;
1296 /* Set long name. */
1297 dict_rename_var (dict, var, long_name);
1299 /* Restore short names. */
1300 for (i = 0; i < short_name_cnt; i++)
1302 var_set_short_name (var, i, short_names[i]);
1303 free (short_names[i]);
1307 close_text_record (r, text);
1308 r->has_long_var_names = true;
1311 /* Reads record type 7, subtype 14, which gives the real length
1312 of each very long string. Rearranges DICT accordingly. */
1314 read_long_string_map (struct sfm_reader *r, size_t size, size_t count,
1315 struct dictionary *dict)
1317 struct text_record *text;
1318 struct variable *var;
1321 text = open_text_record (r, size * count);
1322 while (read_variable_to_value_pair (r, dict, text, &var, &length_s))
1324 size_t idx = var_get_dict_index (var);
1330 length = strtol (length_s, NULL, 10);
1331 if (length < 1 || length > MAX_STRING)
1333 sys_warn (r, _("%s listed as string of invalid length %s "
1334 "in very length string record."),
1335 var_get_name (var), length_s);
1339 /* Check segments. */
1340 segment_cnt = sfm_width_to_segments (length);
1341 if (segment_cnt == 1)
1343 sys_warn (r, _("%s listed in very long string record with width %s, "
1344 "which requires only one segment."),
1345 var_get_name (var), length_s);
1348 if (idx + segment_cnt > dict_get_var_cnt (dict))
1349 sys_error (r, _("Very long string %s overflows dictionary."),
1350 var_get_name (var));
1352 /* Get the short names from the segments and check their
1354 for (i = 0; i < segment_cnt; i++)
1356 struct variable *seg = dict_get_var (dict, idx + i);
1357 int alloc_width = sfm_segment_alloc_width (length, i);
1358 int width = var_get_width (seg);
1361 var_set_short_name (var, i, var_get_short_name (seg, 0));
1362 if (ROUND_UP (width, 8) != ROUND_UP (alloc_width, 8))
1363 sys_error (r, _("Very long string with width %ld has segment %d "
1364 "of width %d (expected %d)"),
1365 length, i, width, alloc_width);
1367 dict_delete_consecutive_vars (dict, idx + 1, segment_cnt - 1);
1368 var_set_width (var, length);
1370 close_text_record (r, text);
1371 dict_compact_values (dict);
1374 /* Reads value labels from sysfile H and inserts them into the
1375 associated dictionary. */
1377 read_value_labels (struct sfm_reader *r,
1378 struct dictionary *dict, struct variable **var_by_value_idx)
1380 struct pool *subpool;
1384 uint8_t raw_value[8]; /* Value as uninterpreted bytes. */
1385 union value value; /* Value. */
1386 char *label; /* Null-terminated label string. */
1389 struct label *labels = NULL;
1390 int label_cnt; /* Number of labels. */
1392 struct variable **var = NULL; /* Associated variables. */
1393 int var_cnt; /* Number of associated variables. */
1394 int max_width; /* Maximum width of string variables. */
1398 subpool = pool_create_subpool (r->pool);
1400 /* Read the type 3 record and record its contents. We can't do
1401 much with the data yet because we don't know whether it is
1402 of numeric or string type. */
1404 /* Read number of labels. */
1405 label_cnt = read_int (r);
1407 if (size_overflow_p (xtimes (label_cnt, sizeof *labels)))
1409 sys_warn (r, _("Invalid number of labels: %d. Ignoring labels."),
1414 /* Read each value/label tuple into labels[]. */
1415 labels = pool_nalloc (subpool, label_cnt, sizeof *labels);
1416 for (i = 0; i < label_cnt; i++)
1418 struct label *label = labels + i;
1419 unsigned char label_len;
1423 read_bytes (r, label->raw_value, sizeof label->raw_value);
1425 /* Read label length. */
1426 read_bytes (r, &label_len, sizeof label_len);
1427 padded_len = ROUND_UP (label_len + 1, 8);
1429 /* Read label, padding. */
1430 label->label = pool_alloc (subpool, padded_len + 1);
1431 read_bytes (r, label->label, padded_len - 1);
1432 label->label[label_len] = 0;
1435 /* Now, read the type 4 record that has the list of variables
1436 to which the value labels are to be applied. */
1438 /* Read record type of type 4 record. */
1439 if (read_int (r) != 4)
1440 sys_error (r, _("Variable index record (type 4) does not immediately "
1441 "follow value label record (type 3) as it should."));
1443 /* Read number of variables associated with value label from type 4
1445 var_cnt = read_int (r);
1446 if (var_cnt < 1 || var_cnt > dict_get_var_cnt (dict))
1447 sys_error (r, _("Number of variables associated with a value label (%d) "
1448 "is not between 1 and the number of variables (%zu)."),
1449 var_cnt, dict_get_var_cnt (dict));
1451 /* Read the list of variables. */
1452 var = pool_nalloc (subpool, var_cnt, sizeof *var);
1454 for (i = 0; i < var_cnt; i++)
1456 var[i] = lookup_var_by_value_idx (r, var_by_value_idx, read_int (r));
1457 if (var_get_width (var[i]) > 8)
1458 sys_error (r, _("Value labels may not be added to long string "
1459 "variables (e.g. %s) using records types 3 and 4."),
1460 var_get_name (var[i]));
1461 max_width = MAX (max_width, var_get_width (var[i]));
1464 /* Type check the variables. */
1465 for (i = 1; i < var_cnt; i++)
1466 if (var_get_type (var[i]) != var_get_type (var[0]))
1467 sys_error (r, _("Variables associated with value label are not all of "
1468 "identical type. Variable %s is %s, but variable "
1470 var_get_name (var[0]),
1471 var_is_numeric (var[0]) ? _("numeric") : _("string"),
1472 var_get_name (var[i]),
1473 var_is_numeric (var[i]) ? _("numeric") : _("string"));
1475 /* Fill in labels[].value, now that we know the desired type. */
1476 for (i = 0; i < label_cnt; i++)
1478 struct label *label = labels + i;
1480 value_init_pool (subpool, &label->value, max_width);
1481 if (var_is_alpha (var[0]))
1482 u8_buf_copy_rpad (value_str_rw (&label->value, max_width), max_width,
1483 label->raw_value, sizeof label->raw_value, ' ');
1485 label->value.f = float_get_double (r->float_format, label->raw_value);
1488 /* Assign the `value_label's to each variable. */
1489 for (i = 0; i < var_cnt; i++)
1491 struct variable *v = var[i];
1494 /* Add each label to the variable. */
1495 for (j = 0; j < label_cnt; j++)
1497 struct label *label = &labels[j];
1498 if (!var_add_value_label (v, &label->value, label->label))
1500 if (var_is_numeric (var[0]))
1501 sys_warn (r, _("Duplicate value label for %g on %s."),
1502 label->value.f, var_get_name (v));
1504 sys_warn (r, _("Duplicate value label for \"%.*s\" on %s."),
1505 max_width, value_str (&label->value, max_width),
1511 pool_destroy (subpool);
1514 /* Reads a set of custom attributes from TEXT into ATTRS.
1515 ATTRS may be a null pointer, in which case the attributes are
1516 read but discarded. */
1518 read_attributes (struct sfm_reader *r, struct text_record *text,
1519 struct attrset *attrs)
1523 struct attribute *attr;
1527 /* Parse the key. */
1528 key = text_get_token (text, ss_cstr ("("), NULL);
1532 attr = attribute_create (key);
1533 for (index = 1; ; index++)
1535 /* Parse the value. */
1539 value = text_get_token (text, ss_cstr ("\n"), NULL);
1542 text_warn (r, text, _("Error parsing attribute value %s[%d]"),
1547 length = strlen (value);
1548 if (length >= 2 && value[0] == '\'' && value[length - 1] == '\'')
1550 value[length - 1] = '\0';
1551 attribute_add_value (attr, value + 1);
1556 _("Attribute value %s[%d] is not quoted: %s"),
1558 attribute_add_value (attr, value);
1561 /* Was this the last value for this attribute? */
1562 if (text_match (text, ')'))
1566 attrset_add (attrs, attr);
1568 attribute_destroy (attr);
1570 while (!text_match (text, '/'));
1573 /* Reads record type 7, subtype 17, which lists custom
1574 attributes on the data file. */
1576 read_data_file_attributes (struct sfm_reader *r,
1577 size_t size, size_t count,
1578 struct dictionary *dict)
1580 struct text_record *text = open_text_record (r, size * count);
1581 read_attributes (r, text, dict_get_attributes (dict));
1582 close_text_record (r, text);
1586 skip_long_string_value_labels (struct sfm_reader *r, size_t n_labels)
1590 for (i = 0; i < n_labels; i++)
1592 size_t value_length, label_length;
1594 value_length = read_int (r);
1595 skip_bytes (r, value_length);
1596 label_length = read_int (r);
1597 skip_bytes (r, label_length);
1602 read_long_string_value_labels (struct sfm_reader *r,
1603 size_t size, size_t count,
1604 struct dictionary *d)
1606 const off_t start = ftello (r->file);
1607 while (ftello (r->file) - start < size * count)
1609 char var_name[VAR_NAME_LEN + 1];
1617 var_name_len = read_int (r);
1618 if (var_name_len > VAR_NAME_LEN)
1619 sys_error (r, _("Variable name length in long string value label "
1620 "record (%d) exceeds %d-byte limit."),
1621 var_name_len, VAR_NAME_LEN);
1622 read_string (r, var_name, var_name_len + 1);
1623 width = read_int (r);
1624 n_labels = read_int (r);
1626 v = dict_lookup_var (d, var_name);
1629 sys_warn (r, _("Ignoring long string value record for "
1630 "unknown variable %s."), var_name);
1631 skip_long_string_value_labels (r, n_labels);
1634 if (var_is_numeric (v))
1636 sys_warn (r, _("Ignoring long string value record for "
1637 "numeric variable %s."), var_name);
1638 skip_long_string_value_labels (r, n_labels);
1641 if (width != var_get_width (v))
1643 sys_warn (r, _("Ignoring long string value record for variable %s "
1644 "because the record's width (%d) does not match the "
1645 "variable's width (%d)"),
1646 var_name, width, var_get_width (v));
1647 skip_long_string_value_labels (r, n_labels);
1652 value_init_pool (r->pool, &value, width);
1653 for (i = 0; i < n_labels; i++)
1655 size_t value_length, label_length;
1660 value_length = read_int (r);
1661 if (value_length == width)
1662 read_bytes (r, value_str_rw (&value, width), width);
1665 sys_warn (r, _("Ignoring long string value %zu for variable %s, "
1666 "with width %d, that has bad value width %zu."),
1667 i, var_get_name (v), width, value_length);
1668 skip_bytes (r, value_length);
1673 label_length = read_int (r);
1674 read_string (r, label, MIN (sizeof label, label_length + 1));
1675 if (label_length >= sizeof label)
1677 /* Skip and silently ignore label text after the
1678 first 255 bytes. The maximum documented length
1679 of a label is 120 bytes so this is more than
1681 skip_bytes (r, (label_length + 1) - sizeof label);
1684 if (!skip && !var_add_value_label (v, &value, label))
1685 sys_warn (r, _("Duplicate value label for \"%.*s\" on %s."),
1686 width, value_str (&value, width), var_get_name (v));
1692 /* Reads record type 7, subtype 18, which lists custom
1693 attributes on individual variables. */
1695 read_variable_attributes (struct sfm_reader *r,
1696 size_t size, size_t count,
1697 struct dictionary *dict)
1699 struct text_record *text = open_text_record (r, size * count);
1702 struct variable *var;
1703 if (!text_read_variable_name (r, dict, text, ss_cstr (":"), &var))
1705 read_attributes (r, text, var != NULL ? var_get_attributes (var) : NULL);
1707 close_text_record (r, text);
1713 static void partial_record (struct sfm_reader *r)
1716 static void read_error (struct casereader *, const struct sfm_reader *);
1718 static bool read_case_number (struct sfm_reader *, double *);
1719 static bool read_case_string (struct sfm_reader *, uint8_t *, size_t);
1720 static int read_opcode (struct sfm_reader *);
1721 static bool read_compressed_number (struct sfm_reader *, double *);
1722 static bool read_compressed_string (struct sfm_reader *, uint8_t *);
1723 static bool read_whole_strings (struct sfm_reader *, uint8_t *, size_t);
1724 static bool skip_whole_strings (struct sfm_reader *, size_t);
1726 /* Reads and returns one case from READER's file. Returns a null
1727 pointer if not successful. */
1728 static struct ccase *
1729 sys_file_casereader_read (struct casereader *reader, void *r_)
1731 struct sfm_reader *r = r_;
1732 struct ccase *volatile c;
1738 c = case_create (r->proto);
1739 if (setjmp (r->bail_out))
1741 casereader_force_error (reader);
1746 for (i = 0; i < r->sfm_var_cnt; i++)
1748 struct sfm_var *sv = &r->sfm_vars[i];
1749 union value *v = case_data_rw_idx (c, sv->case_index);
1751 if (sv->var_width == 0)
1753 if (!read_case_number (r, &v->f))
1758 uint8_t *s = value_str_rw (v, sv->var_width);
1759 if (!read_case_string (r, s + sv->offset, sv->segment_width))
1761 if (!skip_whole_strings (r, ROUND_DOWN (sv->padding, 8)))
1771 if (r->case_cnt != -1)
1772 read_error (reader, r);
1776 /* Issues an error that R ends in a partial record. */
1778 partial_record (struct sfm_reader *r)
1780 sys_error (r, _("File ends in partial case."));
1783 /* Issues an error that an unspecified error occurred SFM, and
1786 read_error (struct casereader *r, const struct sfm_reader *sfm)
1788 msg (ME, _("Error reading case from file %s."), fh_get_name (sfm->fh));
1789 casereader_force_error (r);
1792 /* Reads a number from R and stores its value in *D.
1793 If R is compressed, reads a compressed number;
1794 otherwise, reads a number in the regular way.
1795 Returns true if successful, false if end of file is
1796 reached immediately. */
1798 read_case_number (struct sfm_reader *r, double *d)
1803 if (!try_read_bytes (r, number, sizeof number))
1805 float_convert (r->float_format, number, FLOAT_NATIVE_DOUBLE, d);
1809 return read_compressed_number (r, d);
1812 /* Reads LENGTH string bytes from R into S.
1813 Always reads a multiple of 8 bytes; if LENGTH is not a
1814 multiple of 8, then extra bytes are read and discarded without
1816 Reads compressed strings if S is compressed.
1817 Returns true if successful, false if end of file is
1818 reached immediately. */
1820 read_case_string (struct sfm_reader *r, uint8_t *s, size_t length)
1822 size_t whole = ROUND_DOWN (length, 8);
1823 size_t partial = length % 8;
1827 if (!read_whole_strings (r, s, whole))
1834 if (!read_whole_strings (r, bounce, sizeof bounce))
1840 memcpy (s + whole, bounce, partial);
1846 /* Reads and returns the next compression opcode from R. */
1848 read_opcode (struct sfm_reader *r)
1850 assert (r->compressed);
1854 if (r->opcode_idx >= sizeof r->opcodes)
1856 if (!try_read_bytes (r, r->opcodes, sizeof r->opcodes))
1860 opcode = r->opcodes[r->opcode_idx++];
1867 /* Reads a compressed number from R and stores its value in D.
1868 Returns true if successful, false if end of file is
1869 reached immediately. */
1871 read_compressed_number (struct sfm_reader *r, double *d)
1873 int opcode = read_opcode (r);
1881 *d = read_float (r);
1885 float_convert (r->float_format, " ", FLOAT_NATIVE_DOUBLE, d);
1886 if (!r->corruption_warning)
1888 r->corruption_warning = true;
1889 sys_warn (r, _("Possible compressed data corruption: "
1890 "compressed spaces appear in numeric field."));
1899 *d = opcode - r->bias;
1906 /* Reads a compressed 8-byte string segment from R and stores it
1908 Returns true if successful, false if end of file is
1909 reached immediately. */
1911 read_compressed_string (struct sfm_reader *r, uint8_t *dst)
1913 int opcode = read_opcode (r);
1921 read_bytes (r, dst, 8);
1925 memset (dst, ' ', 8);
1930 double value = opcode - r->bias;
1931 float_convert (FLOAT_NATIVE_DOUBLE, &value, r->float_format, dst);
1934 /* This has actually been seen "in the wild". The submitter of the
1935 file that showed that the contents decoded as spaces, but they
1936 were at the end of the field so it's possible that the null
1937 bytes just acted as null terminators. */
1939 else if (!r->corruption_warning)
1941 r->corruption_warning = true;
1942 sys_warn (r, _("Possible compressed data corruption: "
1943 "string contains compressed integer (opcode %d)"),
1953 /* Reads LENGTH string bytes from R into S.
1954 LENGTH must be a multiple of 8.
1955 Reads compressed strings if S is compressed.
1956 Returns true if successful, false if end of file is
1957 reached immediately. */
1959 read_whole_strings (struct sfm_reader *r, uint8_t *s, size_t length)
1961 assert (length % 8 == 0);
1963 return try_read_bytes (r, s, length);
1967 for (ofs = 0; ofs < length; ofs += 8)
1968 if (!read_compressed_string (r, s + ofs))
1978 /* Skips LENGTH string bytes from R.
1979 LENGTH must be a multiple of 8.
1980 (LENGTH is also limited to 1024, but that's only because the
1981 current caller never needs more than that many bytes.)
1982 Returns true if successful, false if end of file is
1983 reached immediately. */
1985 skip_whole_strings (struct sfm_reader *r, size_t length)
1987 uint8_t buffer[1024];
1988 assert (length < sizeof buffer);
1989 return read_whole_strings (r, buffer, length);
1992 /* Creates and returns a table that can be used for translating a value
1993 index into a case to a "struct variable *" for DICT. Multiple
1994 system file fields reference variables this way.
1996 This table must be created before processing the very long
1997 string extension record, because that record causes some
1998 values to be deleted from the case and the dictionary to be
2000 static struct variable **
2001 make_var_by_value_idx (struct sfm_reader *r, struct dictionary *dict)
2003 struct variable **var_by_value_idx;
2007 var_by_value_idx = pool_nmalloc (r->pool,
2008 r->oct_cnt, sizeof *var_by_value_idx);
2009 for (i = 0; i < dict_get_var_cnt (dict); i++)
2011 struct variable *v = dict_get_var (dict, i);
2012 int nv = var_is_numeric (v) ? 1 : DIV_RND_UP (var_get_width (v), 8);
2015 var_by_value_idx[value_idx++] = v;
2016 for (j = 1; j < nv; j++)
2017 var_by_value_idx[value_idx++] = NULL;
2019 assert (value_idx == r->oct_cnt);
2021 return var_by_value_idx;
2024 /* Returns the "struct variable" corresponding to the given
2025 1-basd VALUE_IDX in VAR_BY_VALUE_IDX. Verifies that the index
2027 static struct variable *
2028 lookup_var_by_value_idx (struct sfm_reader *r,
2029 struct variable **var_by_value_idx, int value_idx)
2031 struct variable *var;
2033 if (value_idx < 1 || value_idx > r->oct_cnt)
2034 sys_error (r, _("Variable index %d not in valid range 1...%d."),
2035 value_idx, r->oct_cnt);
2037 var = var_by_value_idx[value_idx - 1];
2039 sys_error (r, _("Variable index %d refers to long string "
2046 /* Returns the variable in D with the given SHORT_NAME,
2047 or a null pointer if there is none. */
2048 static struct variable *
2049 lookup_var_by_short_name (struct dictionary *d, const char *short_name)
2051 struct variable *var;
2055 /* First try looking up by full name. This often succeeds. */
2056 var = dict_lookup_var (d, short_name);
2057 if (var != NULL && !strcasecmp (var_get_short_name (var, 0), short_name))
2060 /* Iterate through the whole dictionary as a fallback. */
2061 var_cnt = dict_get_var_cnt (d);
2062 for (i = 0; i < var_cnt; i++)
2064 var = dict_get_var (d, i);
2065 if (!strcasecmp (var_get_short_name (var, 0), short_name))
2072 /* Helpers for reading records that contain structured text
2075 /* Maximum number of warnings to issue for a single text
2077 #define MAX_TEXT_WARNINGS 5
2082 struct substring buffer; /* Record contents. */
2083 size_t pos; /* Current position in buffer. */
2084 int n_warnings; /* Number of warnings issued or suppressed. */
2087 /* Reads SIZE bytes into a text record for R,
2088 and returns the new text record. */
2089 static struct text_record *
2090 open_text_record (struct sfm_reader *r, size_t size)
2092 struct text_record *text = pool_alloc (r->pool, sizeof *text);
2093 char *buffer = pool_malloc (r->pool, size + 1);
2094 read_bytes (r, buffer, size);
2095 text->buffer = ss_buffer (buffer, size);
2097 text->n_warnings = 0;
2101 /* Closes TEXT, frees its storage, and issues a final warning
2102 about suppressed warnings if necesary. */
2104 close_text_record (struct sfm_reader *r, struct text_record *text)
2106 if (text->n_warnings > MAX_TEXT_WARNINGS)
2107 sys_warn (r, _("Suppressed %d additional related warnings."),
2108 text->n_warnings - MAX_TEXT_WARNINGS);
2109 pool_free (r->pool, ss_data (text->buffer));
2112 /* Reads a variable=value pair from TEXT.
2113 Looks up the variable in DICT and stores it into *VAR.
2114 Stores a null-terminated value into *VALUE. */
2116 read_variable_to_value_pair (struct sfm_reader *r, struct dictionary *dict,
2117 struct text_record *text,
2118 struct variable **var, char **value)
2122 if (!text_read_short_name (r, dict, text, ss_cstr ("="), var))
2125 *value = text_get_token (text, ss_buffer ("\t\0", 2), NULL);
2129 text->pos += ss_span (ss_substr (text->buffer, text->pos, SIZE_MAX),
2130 ss_buffer ("\t\0", 2));
2138 text_read_variable_name (struct sfm_reader *r, struct dictionary *dict,
2139 struct text_record *text, struct substring delimiters,
2140 struct variable **var)
2144 name = text_get_token (text, delimiters, NULL);
2148 *var = dict_lookup_var (dict, name);
2152 text_warn (r, text, _("Dictionary record refers to unknown variable %s."),
2159 text_read_short_name (struct sfm_reader *r, struct dictionary *dict,
2160 struct text_record *text, struct substring delimiters,
2161 struct variable **var)
2163 char *short_name = text_get_token (text, delimiters, NULL);
2164 if (short_name == NULL)
2167 *var = lookup_var_by_short_name (dict, short_name);
2169 text_warn (r, text, _("Dictionary record refers to unknown variable %s."),
2174 /* Displays a warning for the current file position, limiting the
2175 number to MAX_TEXT_WARNINGS for TEXT. */
2177 text_warn (struct sfm_reader *r, struct text_record *text,
2178 const char *format, ...)
2180 if (text->n_warnings++ < MAX_TEXT_WARNINGS)
2184 va_start (args, format);
2185 sys_msg (r, MW, format, args);
2191 text_get_token (struct text_record *text, struct substring delimiters,
2194 struct substring token;
2197 if (!ss_tokenize (text->buffer, delimiters, &text->pos, &token))
2200 end = &ss_data (token)[ss_length (token)];
2201 if (delimiter != NULL)
2204 return ss_data (token);
2207 /* Reads a integer value expressed in decimal, then a space, then a string that
2208 consists of exactly as many bytes as specified by the integer, then a space,
2209 from TEXT. Returns the string, null-terminated, as a subset of TEXT's
2210 buffer (so the caller should not free the string). */
2212 text_parse_counted_string (struct sfm_reader *r, struct text_record *text)
2222 int c = text->buffer.string[text->pos];
2223 if (c < '0' || c > '9')
2225 n = (n * 10) + (c - '0');
2228 if (start == text->pos)
2230 sys_warn (r, _("Expecting digit at offset %zu in MRSETS record."),
2235 if (!text_match (text, ' '))
2237 sys_warn (r, _("Expecting space at offset %zu in MRSETS record."),
2242 if (text->pos + n > text->buffer.length)
2244 sys_warn (r, _("%zu-byte string starting at offset %zu "
2245 "exceeds record length %zu."),
2246 n, text->pos, text->buffer.length);
2250 s = &text->buffer.string[text->pos];
2254 _("Expecting space at offset %zu following %zu-byte string."),
2264 text_match (struct text_record *text, char c)
2266 if (text->buffer.string[text->pos] == c)
2275 /* Returns the current byte offset inside the TEXT's string. */
2277 text_pos (const struct text_record *text)
2284 /* Displays a corruption message. */
2286 sys_msg (struct sfm_reader *r, int class, const char *format, va_list args)
2291 ds_init_empty (&text);
2292 ds_put_format (&text, "\"%s\" near offset 0x%llx: ",
2293 fh_get_file_name (r->fh), (long long int) ftello (r->file));
2294 ds_put_vformat (&text, format, args);
2296 m.category = msg_class_to_category (class);
2297 m.severity = msg_class_to_severity (class);
2298 m.where.file_name = NULL;
2299 m.where.line_number = 0;
2300 m.text = ds_cstr (&text);
2305 /* Displays a warning for the current file position. */
2307 sys_warn (struct sfm_reader *r, const char *format, ...)
2311 va_start (args, format);
2312 sys_msg (r, MW, format, args);
2316 /* Displays an error for the current file position,
2317 marks it as in an error state,
2318 and aborts reading it using longjmp. */
2320 sys_error (struct sfm_reader *r, const char *format, ...)
2324 va_start (args, format);
2325 sys_msg (r, ME, format, args);
2329 longjmp (r->bail_out, 1);
2332 /* Reads BYTE_CNT bytes into BUF.
2333 Returns true if exactly BYTE_CNT bytes are successfully read.
2334 Aborts if an I/O error or a partial read occurs.
2335 If EOF_IS_OK, then an immediate end-of-file causes false to be
2336 returned; otherwise, immediate end-of-file causes an abort
2339 read_bytes_internal (struct sfm_reader *r, bool eof_is_ok,
2340 void *buf, size_t byte_cnt)
2342 size_t bytes_read = fread (buf, 1, byte_cnt, r->file);
2343 if (bytes_read == byte_cnt)
2345 else if (ferror (r->file))
2346 sys_error (r, _("System error: %s."), strerror (errno));
2347 else if (!eof_is_ok || bytes_read != 0)
2348 sys_error (r, _("Unexpected end of file."));
2353 /* Reads BYTE_CNT into BUF.
2354 Aborts upon I/O error or if end-of-file is encountered. */
2356 read_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
2358 read_bytes_internal (r, false, buf, byte_cnt);
2361 /* Reads BYTE_CNT bytes into BUF.
2362 Returns true if exactly BYTE_CNT bytes are successfully read.
2363 Returns false if an immediate end-of-file is encountered.
2364 Aborts if an I/O error or a partial read occurs. */
2366 try_read_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
2368 return read_bytes_internal (r, true, buf, byte_cnt);
2371 /* Reads a 32-bit signed integer from R and returns its value in
2374 read_int (struct sfm_reader *r)
2377 read_bytes (r, integer, sizeof integer);
2378 return integer_get (r->integer_format, integer, sizeof integer);
2381 /* Reads a 64-bit floating-point number from R and returns its
2382 value in host format. */
2384 read_float (struct sfm_reader *r)
2387 read_bytes (r, number, sizeof number);
2388 return float_get_double (r->float_format, number);
2391 /* Reads exactly SIZE - 1 bytes into BUFFER
2392 and stores a null byte into BUFFER[SIZE - 1]. */
2394 read_string (struct sfm_reader *r, char *buffer, size_t size)
2397 read_bytes (r, buffer, size - 1);
2398 buffer[size - 1] = '\0';
2401 /* Skips BYTES bytes forward in R. */
2403 skip_bytes (struct sfm_reader *r, size_t bytes)
2408 size_t chunk = MIN (sizeof buffer, bytes);
2409 read_bytes (r, buffer, chunk);
2414 static const struct casereader_class sys_file_casereader_class =
2416 sys_file_casereader_read,
2417 sys_file_casereader_destroy,