1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-2000, 2006-2007, 2009-2012 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/identifier.h"
37 #include "data/missing-values.h"
38 #include "data/mrset.h"
39 #include "data/short-names.h"
40 #include "data/value-labels.h"
41 #include "data/value.h"
42 #include "data/variable.h"
43 #include "libpspp/array.h"
44 #include "libpspp/assertion.h"
45 #include "libpspp/compiler.h"
46 #include "libpspp/i18n.h"
47 #include "libpspp/message.h"
48 #include "libpspp/misc.h"
49 #include "libpspp/pool.h"
50 #include "libpspp/str.h"
51 #include "libpspp/stringi-set.h"
53 #include "gl/c-strtod.h"
54 #include "gl/c-ctype.h"
55 #include "gl/inttostr.h"
56 #include "gl/localcharset.h"
57 #include "gl/minmax.h"
58 #include "gl/unlocked-io.h"
59 #include "gl/xalloc.h"
63 #define _(msgid) gettext (msgid)
64 #define N_(msgid) (msgid)
68 /* subtypes 0-2 unknown */
69 EXT_INTEGER = 3, /* Machine integer info. */
70 EXT_FLOAT = 4, /* Machine floating-point info. */
71 EXT_VAR_SETS = 5, /* Variable sets. */
72 EXT_DATE = 6, /* DATE. */
73 EXT_MRSETS = 7, /* Multiple response sets. */
74 EXT_DATA_ENTRY = 8, /* SPSS Data Entry. */
75 /* subtypes 9-10 unknown */
76 EXT_DISPLAY = 11, /* Variable display parameters. */
77 /* subtype 12 unknown */
78 EXT_LONG_NAMES = 13, /* Long variable names. */
79 EXT_LONG_STRINGS = 14, /* Long strings. */
80 /* subtype 15 unknown */
81 EXT_NCASES = 16, /* Extended number of cases. */
82 EXT_FILE_ATTRS = 17, /* Data file attributes. */
83 EXT_VAR_ATTRS = 18, /* Variable attributes. */
84 EXT_MRSETS2 = 19, /* Multiple response sets (extended). */
85 EXT_ENCODING = 20, /* Character encoding. */
86 EXT_LONG_LABELS = 21 /* Value labels for long strings. */
89 /* Fields from the top-level header record. */
90 struct sfm_header_record
92 char magic[5]; /* First 4 bytes of file, then null. */
93 int weight_idx; /* 0 if unweighted, otherwise a var index. */
94 int nominal_case_size; /* Number of var positions. */
96 /* These correspond to the members of struct sfm_file_info or a dictionary
97 but in the system file's encoding rather than ASCII. */
98 char creation_date[10]; /* "dd mmm yy". */
99 char creation_time[9]; /* "hh:mm:ss". */
100 char eye_catcher[61]; /* Eye-catcher string, then product name. */
101 char file_label[65]; /* File label. */
104 struct sfm_var_record
111 int missing_value_code;
114 struct variable *var;
117 struct sfm_value_label
123 struct sfm_value_label_record
126 struct sfm_value_label *labels;
133 struct sfm_document_record
140 struct sfm_extension_record
142 off_t pos; /* Starting offset in file. */
143 size_t size; /* Size of data elements. */
144 size_t count; /* Number of data elements. */
145 void *data; /* Contents. */
148 /* System file reader. */
151 /* Resource tracking. */
152 struct pool *pool; /* All system file state. */
153 jmp_buf bail_out; /* longjmp() target for error handling. */
156 struct file_handle *fh; /* File handle. */
157 struct fh_lock *lock; /* Mutual exclusion for file handle. */
158 FILE *file; /* File stream. */
159 off_t pos; /* Position in file. */
160 bool error; /* I/O or corruption error? */
161 struct caseproto *proto; /* Format of output cases. */
164 enum integer_format integer_format; /* On-disk integer format. */
165 enum float_format float_format; /* On-disk floating point format. */
166 struct sfm_var *sfm_vars; /* Variables. */
167 size_t sfm_var_cnt; /* Number of variables. */
168 casenumber case_cnt; /* Number of cases */
169 const char *encoding; /* String encoding. */
172 bool compressed; /* File is compressed? */
173 double bias; /* Compression bias, usually 100.0. */
174 uint8_t opcodes[8]; /* Current block of opcodes. */
175 size_t opcode_idx; /* Next opcode to interpret, 8 if none left. */
176 bool corruption_warning; /* Warned about possible corruption? */
179 static const struct casereader_class sys_file_casereader_class;
181 static bool close_reader (struct sfm_reader *);
183 static struct variable *lookup_var_by_index (struct sfm_reader *, off_t,
184 const struct sfm_var_record *,
187 static void sys_msg (struct sfm_reader *r, off_t, int class,
188 const char *format, va_list args)
189 PRINTF_FORMAT (4, 0);
190 static void sys_warn (struct sfm_reader *, off_t, const char *, ...)
191 PRINTF_FORMAT (3, 4);
192 static void sys_error (struct sfm_reader *, off_t, const char *, ...)
196 static void read_bytes (struct sfm_reader *, void *, size_t);
197 static bool try_read_bytes (struct sfm_reader *, void *, size_t);
198 static int read_int (struct sfm_reader *);
199 static double read_float (struct sfm_reader *);
200 static void read_string (struct sfm_reader *, char *, size_t);
201 static void skip_bytes (struct sfm_reader *, size_t);
203 static int parse_int (struct sfm_reader *, const void *data, size_t ofs);
204 static double parse_float (struct sfm_reader *, const void *data, size_t ofs);
206 static void read_variable_record (struct sfm_reader *,
207 struct sfm_var_record *);
208 static void read_value_label_record (struct sfm_reader *,
209 struct sfm_value_label_record *,
211 static struct sfm_document_record *read_document_record (struct sfm_reader *);
212 static struct sfm_extension_record *read_extension_record (
213 struct sfm_reader *, int subtype);
214 static void skip_extension_record (struct sfm_reader *, int subtype);
216 static const char *choose_encoding (
218 const struct sfm_header_record *,
219 const struct sfm_extension_record *ext_integer,
220 const struct sfm_extension_record *ext_encoding);
222 static struct text_record *open_text_record (
223 struct sfm_reader *, const struct sfm_extension_record *,
224 bool recode_to_utf8);
225 static void close_text_record (struct sfm_reader *,
226 struct text_record *);
227 static bool read_variable_to_value_pair (struct sfm_reader *,
229 struct text_record *,
230 struct variable **var, char **value);
231 static void text_warn (struct sfm_reader *r, struct text_record *text,
232 const char *format, ...)
233 PRINTF_FORMAT (3, 4);
234 static char *text_get_token (struct text_record *,
235 struct substring delimiters, char *delimiter);
236 static bool text_match (struct text_record *, char c);
237 static bool text_read_variable_name (struct sfm_reader *, struct dictionary *,
238 struct text_record *,
239 struct substring delimiters,
241 static bool text_read_short_name (struct sfm_reader *, struct dictionary *,
242 struct text_record *,
243 struct substring delimiters,
245 static const char *text_parse_counted_string (struct sfm_reader *,
246 struct text_record *);
247 static size_t text_pos (const struct text_record *);
249 static bool close_reader (struct sfm_reader *r);
251 /* Dictionary reader. */
259 static void read_header (struct sfm_reader *, struct sfm_read_info *,
260 struct sfm_header_record *);
261 static void parse_header (struct sfm_reader *,
262 const struct sfm_header_record *,
263 struct sfm_read_info *, struct dictionary *);
264 static void parse_variable_records (struct sfm_reader *, struct dictionary *,
265 struct sfm_var_record *, size_t n);
266 static void parse_format_spec (struct sfm_reader *, off_t pos,
267 unsigned int format, enum which_format,
268 struct variable *, int *format_warning_cnt);
269 static void parse_document (struct dictionary *, struct sfm_document_record *);
270 static void parse_display_parameters (struct sfm_reader *,
271 const struct sfm_extension_record *,
272 struct dictionary *);
273 static void parse_machine_integer_info (struct sfm_reader *,
274 const struct sfm_extension_record *,
275 struct sfm_read_info *);
276 static void parse_machine_float_info (struct sfm_reader *,
277 const struct sfm_extension_record *);
278 static void parse_mrsets (struct sfm_reader *,
279 const struct sfm_extension_record *,
280 struct dictionary *);
281 static void parse_long_var_name_map (struct sfm_reader *,
282 const struct sfm_extension_record *,
283 struct dictionary *);
284 static void parse_long_string_map (struct sfm_reader *,
285 const struct sfm_extension_record *,
286 struct dictionary *);
287 static void parse_value_labels (struct sfm_reader *, struct dictionary *,
288 const struct sfm_var_record *,
290 const struct sfm_value_label_record *);
291 static void parse_data_file_attributes (struct sfm_reader *,
292 const struct sfm_extension_record *,
293 struct dictionary *);
294 static void parse_variable_attributes (struct sfm_reader *,
295 const struct sfm_extension_record *,
296 struct dictionary *);
297 static void parse_long_string_value_labels (struct sfm_reader *,
298 const struct sfm_extension_record *,
299 struct dictionary *);
301 /* Frees the strings inside INFO. */
303 sfm_read_info_destroy (struct sfm_read_info *info)
307 free (info->creation_date);
308 free (info->creation_time);
309 free (info->product);
313 /* Opens the system file designated by file handle FH for reading. Reads the
314 system file's dictionary into *DICT.
316 Ordinarily the reader attempts to automatically detect the character
317 encoding based on the file's contents. This isn't always possible,
318 especially for files written by old versions of SPSS or PSPP, so specifying
319 a nonnull ENCODING overrides the choice of character encoding.
321 If INFO is non-null, then it receives additional info about the system file,
322 which the caller must eventually free with sfm_read_info_destroy() when it
323 is no longer needed. */
325 sfm_open_reader (struct file_handle *fh, const char *volatile encoding,
326 struct dictionary **dictp, struct sfm_read_info *infop)
328 struct sfm_reader *volatile r = NULL;
329 struct sfm_read_info *volatile info;
331 struct sfm_header_record header;
333 struct sfm_var_record *vars;
334 size_t n_vars, allocated_vars;
336 struct sfm_value_label_record *labels;
337 size_t n_labels, allocated_labels;
339 struct sfm_document_record *document;
341 struct sfm_extension_record *extensions[32];
343 struct dictionary *volatile dict = NULL;
346 /* Create and initialize reader. */
347 r = pool_create_container (struct sfm_reader, pool);
353 r->opcode_idx = sizeof r->opcodes;
354 r->corruption_warning = false;
356 info = infop ? infop : xmalloc (sizeof *info);
357 memset (info, 0, sizeof *info);
359 /* TRANSLATORS: this fragment will be interpolated into
360 messages in fh_lock() that identify types of files. */
361 r->lock = fh_lock (fh, FH_REF_FILE, N_("system file"), FH_ACC_READ, false);
365 r->file = fn_open (fh_get_file_name (fh), "rb");
368 msg (ME, _("Error opening `%s' for reading as a system file: %s."),
369 fh_get_file_name (r->fh), strerror (errno));
373 if (setjmp (r->bail_out))
377 read_header (r, info, &header);
380 n_vars = allocated_vars = 0;
383 n_labels = allocated_labels = 0;
387 memset (extensions, 0, sizeof extensions);
397 read_int (r); /* Skip filler. */
404 if (n_vars >= allocated_vars)
405 vars = pool_2nrealloc (r->pool, vars, &allocated_vars,
407 read_variable_record (r, &vars[n_vars++]);
411 if (n_labels >= allocated_labels)
412 labels = pool_2nrealloc (r->pool, labels, &allocated_labels,
414 read_value_label_record (r, &labels[n_labels++], n_vars);
418 /* A Type 4 record is always immediately after a type 3 record,
419 so the code for type 3 records reads the type 4 record too. */
420 sys_error (r, r->pos, _("Misplaced type 4 record."));
423 if (document != NULL)
424 sys_error (r, r->pos, _("Duplicate type 6 (document) record."));
425 document = read_document_record (r);
429 subtype = read_int (r);
430 if (subtype < 0 || subtype >= sizeof extensions / sizeof *extensions)
433 _("Unrecognized record type 7, subtype %d. Please "
434 "send a copy of this file, and the syntax which "
435 "created it to %s."),
436 subtype, PACKAGE_BUGREPORT);
437 skip_extension_record (r, subtype);
439 else if (extensions[subtype] != NULL)
442 _("Record type 7, subtype %d found here has the same "
443 "type as the record found near offset 0x%llx. "
444 "Please send a copy of this file, and the syntax "
445 "which created it to %s."),
446 subtype, (long long int) extensions[subtype]->pos,
448 skip_extension_record (r, subtype);
451 extensions[subtype] = read_extension_record (r, subtype);
455 sys_error (r, r->pos, _("Unrecognized record type %d."), type);
460 /* Now actually parse what we read.
462 First, figure out the correct character encoding, because this determines
463 how the rest of the header data is to be interpreted. */
464 dict = dict_create (encoding
466 : choose_encoding (r, &header, extensions[EXT_INTEGER],
467 extensions[EXT_ENCODING]));
468 r->encoding = dict_get_encoding (dict);
470 /* These records don't use variables at all. */
471 if (document != NULL)
472 parse_document (dict, document);
474 if (extensions[EXT_INTEGER] != NULL)
475 parse_machine_integer_info (r, extensions[EXT_INTEGER], info);
477 if (extensions[EXT_FLOAT] != NULL)
478 parse_machine_float_info (r, extensions[EXT_FLOAT]);
480 if (extensions[EXT_FILE_ATTRS] != NULL)
481 parse_data_file_attributes (r, extensions[EXT_FILE_ATTRS], dict);
483 parse_header (r, &header, info, dict);
485 /* Parse the variable records, the basis of almost everything else. */
486 parse_variable_records (r, dict, vars, n_vars);
488 /* Parse value labels and the weight variable immediately after the variable
489 records. These records use indexes into var_recs[], so we must parse them
490 before those indexes become invalidated by very long string variables. */
491 for (i = 0; i < n_labels; i++)
492 parse_value_labels (r, dict, vars, n_vars, &labels[i]);
493 if (header.weight_idx != 0)
495 struct variable *weight_var;
497 weight_var = lookup_var_by_index (r, 76, vars, n_vars,
499 if (var_is_numeric (weight_var))
500 dict_set_weight (dict, weight_var);
502 sys_error (r, -1, _("Weighting variable must be numeric "
503 "(not string variable `%s')."),
504 var_get_name (weight_var));
507 if (extensions[EXT_DISPLAY] != NULL)
508 parse_display_parameters (r, extensions[EXT_DISPLAY], dict);
510 /* The following records use short names, so they need to be parsed before
511 parse_long_var_name_map() changes short names to long names. */
512 if (extensions[EXT_MRSETS] != NULL)
513 parse_mrsets (r, extensions[EXT_MRSETS], dict);
515 if (extensions[EXT_MRSETS2] != NULL)
516 parse_mrsets (r, extensions[EXT_MRSETS2], dict);
518 if (extensions[EXT_LONG_STRINGS] != NULL)
519 parse_long_string_map (r, extensions[EXT_LONG_STRINGS], dict);
521 /* Now rename variables to their long names. */
522 parse_long_var_name_map (r, extensions[EXT_LONG_NAMES], dict);
524 /* The following records use long names, so they need to follow renaming. */
525 if (extensions[EXT_VAR_ATTRS] != NULL)
526 parse_variable_attributes (r, extensions[EXT_VAR_ATTRS], dict);
528 if (extensions[EXT_LONG_LABELS] != NULL)
529 parse_long_string_value_labels (r, extensions[EXT_LONG_LABELS], dict);
531 /* Warn if the actual amount of data per case differs from the
532 amount that the header claims. SPSS version 13 gets this
533 wrong when very long strings are involved, so don't warn in
535 if (header.nominal_case_size != -1 && header.nominal_case_size != n_vars
536 && info->version_major != 13)
537 sys_warn (r, -1, _("File header claims %d variable positions but "
538 "%zu were read from file."),
539 header.nominal_case_size, n_vars);
541 /* Create an index of dictionary variable widths for
542 sfm_read_case to use. We cannot use the `struct variable's
543 from the dictionary we created, because the caller owns the
544 dictionary and may destroy or modify its variables. */
545 sfm_dictionary_to_sfm_vars (dict, &r->sfm_vars, &r->sfm_var_cnt);
546 pool_register (r->pool, free, r->sfm_vars);
547 r->proto = caseproto_ref_pool (dict_get_proto (dict), r->pool);
552 sfm_read_info_destroy (info);
556 return casereader_create_sequential
558 r->case_cnt == -1 ? CASENUMBER_MAX: r->case_cnt,
559 &sys_file_casereader_class, r);
564 sfm_read_info_destroy (info);
574 /* Closes a system file after we're done with it.
575 Returns true if an I/O error has occurred on READER, false
578 close_reader (struct sfm_reader *r)
587 if (fn_close (fh_get_file_name (r->fh), r->file) == EOF)
589 msg (ME, _("Error closing system file `%s': %s."),
590 fh_get_file_name (r->fh), strerror (errno));
600 pool_destroy (r->pool);
605 /* Destroys READER. */
607 sys_file_casereader_destroy (struct casereader *reader UNUSED, void *r_)
609 struct sfm_reader *r = r_;
613 /* Returns true if FILE is an SPSS system file,
616 sfm_detect (FILE *file)
620 if (fread (magic, 4, 1, file) != 1)
624 return !strcmp (ASCII_MAGIC, magic) || !strcmp (EBCDIC_MAGIC, magic);
627 /* Reads the global header of the system file. Initializes *HEADER and *INFO,
628 except for the string fields in *INFO, which parse_header() will initialize
629 later once the file's encoding is known. */
631 read_header (struct sfm_reader *r, struct sfm_read_info *info,
632 struct sfm_header_record *header)
634 uint8_t raw_layout_code[4];
637 read_string (r, header->magic, sizeof header->magic);
638 read_string (r, header->eye_catcher, sizeof header->eye_catcher);
640 if (strcmp (ASCII_MAGIC, header->magic)
641 && strcmp (EBCDIC_MAGIC, header->magic))
642 sys_error (r, 0, _("This is not an SPSS system file."));
644 /* Identify integer format. */
645 read_bytes (r, raw_layout_code, sizeof raw_layout_code);
646 if ((!integer_identify (2, raw_layout_code, sizeof raw_layout_code,
648 && !integer_identify (3, raw_layout_code, sizeof raw_layout_code,
650 || (r->integer_format != INTEGER_MSB_FIRST
651 && r->integer_format != INTEGER_LSB_FIRST))
652 sys_error (r, 64, _("This is not an SPSS system file."));
654 header->nominal_case_size = read_int (r);
655 if (header->nominal_case_size < 0
656 || header->nominal_case_size > INT_MAX / 16)
657 header->nominal_case_size = -1;
659 r->compressed = read_int (r) != 0;
661 header->weight_idx = read_int (r);
663 r->case_cnt = read_int (r);
664 if ( r->case_cnt > INT_MAX / 2)
667 /* Identify floating-point format and obtain compression bias. */
668 read_bytes (r, raw_bias, sizeof raw_bias);
669 if (float_identify (100.0, raw_bias, sizeof raw_bias, &r->float_format) == 0)
671 uint8_t zero_bias[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
673 if (memcmp (raw_bias, zero_bias, 8))
674 sys_warn (r, r->pos - 8,
675 _("Compression bias is not the usual "
676 "value of 100, or system file uses unrecognized "
677 "floating-point format."));
680 /* Some software is known to write all-zeros to this
681 field. Such software also writes floating-point
682 numbers in the format that we expect by default
683 (it seems that all software most likely does, in
684 reality), so don't warn in this case. */
687 if (r->integer_format == INTEGER_MSB_FIRST)
688 r->float_format = FLOAT_IEEE_DOUBLE_BE;
690 r->float_format = FLOAT_IEEE_DOUBLE_LE;
692 float_convert (r->float_format, raw_bias, FLOAT_NATIVE_DOUBLE, &r->bias);
694 read_string (r, header->creation_date, sizeof header->creation_date);
695 read_string (r, header->creation_time, sizeof header->creation_time);
696 read_string (r, header->file_label, sizeof header->file_label);
699 info->integer_format = r->integer_format;
700 info->float_format = r->float_format;
701 info->compressed = r->compressed;
702 info->case_cnt = r->case_cnt;
705 /* Reads a variable (type 2) record from R into RECORD. */
707 read_variable_record (struct sfm_reader *r, struct sfm_var_record *record)
709 int has_variable_label;
711 memset (record, 0, sizeof *record);
713 record->pos = r->pos;
714 record->width = read_int (r);
715 has_variable_label = read_int (r);
716 record->missing_value_code = read_int (r);
717 record->print_format = read_int (r);
718 record->write_format = read_int (r);
719 read_bytes (r, record->name, sizeof record->name);
721 if (has_variable_label == 1)
723 enum { MAX_LABEL_LEN = 255 };
724 size_t len, read_len;
728 /* Read up to MAX_LABEL_LEN bytes of label. */
729 read_len = MIN (MAX_LABEL_LEN, len);
730 record->label = pool_malloc (r->pool, read_len + 1);
731 read_string (r, record->label, read_len + 1);
733 /* Skip unread label bytes. */
734 skip_bytes (r, len - read_len);
736 /* Skip label padding up to multiple of 4 bytes. */
737 skip_bytes (r, ROUND_UP (len, 4) - len);
739 else if (has_variable_label != 0)
740 sys_error (r, record->pos,
741 _("Variable label indicator field is not 0 or 1."));
743 /* Set missing values. */
744 if (record->missing_value_code != 0)
746 int code = record->missing_value_code;
747 if (record->width == 0)
749 if (code < -3 || code > 3 || code == -1)
750 sys_error (r, record->pos,
751 _("Numeric missing value indicator field is not "
752 "-3, -2, 0, 1, 2, or 3."));
756 if (code < 1 || code > 3)
757 sys_error (r, record->pos,
758 _("String missing value indicator field is not "
762 read_bytes (r, record->missing, 8 * abs (code));
766 /* Reads value labels from R into RECORD. */
768 read_value_label_record (struct sfm_reader *r,
769 struct sfm_value_label_record *record,
774 /* Read type 3 record. */
775 record->pos = r->pos;
776 record->n_labels = read_int (r);
777 if (record->n_labels > SIZE_MAX / sizeof *record->labels)
778 sys_error (r, r->pos - 4, _("Invalid number of labels %zu."),
780 record->labels = pool_nmalloc (r->pool, record->n_labels,
781 sizeof *record->labels);
782 for (i = 0; i < record->n_labels; i++)
784 struct sfm_value_label *label = &record->labels[i];
785 unsigned char label_len;
788 read_bytes (r, label->value, sizeof label->value);
790 /* Read label length. */
791 read_bytes (r, &label_len, sizeof label_len);
792 padded_len = ROUND_UP (label_len + 1, 8);
794 /* Read label, padding. */
795 label->label = pool_malloc (r->pool, padded_len + 1);
796 read_bytes (r, label->label, padded_len - 1);
797 label->label[label_len] = '\0';
800 /* Read record type of type 4 record. */
801 if (read_int (r) != 4)
802 sys_error (r, r->pos - 4,
803 _("Variable index record (type 4) does not immediately "
804 "follow value label record (type 3) as it should."));
806 /* Read number of variables associated with value label from type 4
808 record->n_vars = read_int (r);
809 if (record->n_vars < 1 || record->n_vars > n_vars)
810 sys_error (r, r->pos - 4,
811 _("Number of variables associated with a value label (%zu) "
812 "is not between 1 and the number of variables (%zu)."),
813 record->n_vars, n_vars);
814 record->vars = pool_nmalloc (r->pool, record->n_vars, sizeof *record->vars);
815 for (i = 0; i < record->n_vars; i++)
816 record->vars[i] = read_int (r);
819 /* Reads a document record from R and returns it. */
820 static struct sfm_document_record *
821 read_document_record (struct sfm_reader *r)
823 struct sfm_document_record *record;
826 record = pool_malloc (r->pool, sizeof *record);
827 record->pos = r->pos;
829 n_lines = read_int (r);
830 if (n_lines <= 0 || n_lines >= INT_MAX / DOC_LINE_LENGTH)
831 sys_error (r, record->pos,
832 _("Number of document lines (%d) "
833 "must be greater than 0 and less than %d."),
834 n_lines, INT_MAX / DOC_LINE_LENGTH);
836 record->n_lines = n_lines;
837 record->documents = pool_malloc (r->pool, DOC_LINE_LENGTH * n_lines);
838 read_bytes (r, record->documents, DOC_LINE_LENGTH * n_lines);
844 read_extension_record_header (struct sfm_reader *r, int subtype,
845 struct sfm_extension_record *record)
847 record->pos = r->pos;
848 record->size = read_int (r);
849 record->count = read_int (r);
851 /* Check that SIZE * COUNT + 1 doesn't overflow. Adding 1
852 allows an extra byte for a null terminator, used by some
853 extension processing routines. */
854 if (record->size != 0
855 && size_overflow_p (xsum (1, xtimes (record->count, record->size))))
856 sys_error (r, record->pos, "Record type 7 subtype %d too large.", subtype);
859 /* Reads an extension record from R into RECORD. */
860 static struct sfm_extension_record *
861 read_extension_record (struct sfm_reader *r, int subtype)
863 struct extension_record_type
870 static const struct extension_record_type types[] =
872 /* Implemented record types. */
873 { EXT_INTEGER, 4, 8 },
875 { EXT_MRSETS, 1, 0 },
876 { EXT_DISPLAY, 4, 0 },
877 { EXT_LONG_NAMES, 1, 0 },
878 { EXT_LONG_STRINGS, 1, 0 },
879 { EXT_NCASES, 8, 2 },
880 { EXT_FILE_ATTRS, 1, 0 },
881 { EXT_VAR_ATTRS, 1, 0 },
882 { EXT_MRSETS2, 1, 0 },
883 { EXT_ENCODING, 1, 0 },
884 { EXT_LONG_LABELS, 1, 0 },
886 /* Ignored record types. */
887 { EXT_VAR_SETS, 0, 0 },
889 { EXT_DATA_ENTRY, 0, 0 },
892 const struct extension_record_type *type;
893 struct sfm_extension_record *record;
896 record = pool_malloc (r->pool, sizeof *record);
897 read_extension_record_header (r, subtype, record);
898 n_bytes = record->count * record->size;
900 for (type = types; type < &types[sizeof types / sizeof *types]; type++)
901 if (subtype == type->subtype)
903 if (type->size > 0 && record->size != type->size)
904 sys_warn (r, record->pos,
905 _("Record type 7, subtype %d has bad size %zu "
906 "(expected %d)."), subtype, record->size, type->size);
907 else if (type->count > 0 && record->count != type->count)
908 sys_warn (r, record->pos,
909 _("Record type 7, subtype %d has bad count %zu "
910 "(expected %d)."), subtype, record->count, type->count);
911 else if (type->count == 0 && type->size == 0)
913 /* Ignore this record. */
917 char *data = pool_malloc (r->pool, n_bytes + 1);
918 data[n_bytes] = '\0';
921 read_bytes (r, record->data, n_bytes);
928 sys_warn (r, record->pos,
929 _("Unrecognized record type 7, subtype %d. Please send a "
930 "copy of this file, and the syntax which created it to %s."),
931 subtype, PACKAGE_BUGREPORT);
934 skip_bytes (r, n_bytes);
939 skip_extension_record (struct sfm_reader *r, int subtype)
941 struct sfm_extension_record record;
943 read_extension_record_header (r, subtype, &record);
944 skip_bytes (r, record.count * record.size);
948 parse_header (struct sfm_reader *r, const struct sfm_header_record *header,
949 struct sfm_read_info *info, struct dictionary *dict)
951 const char *dict_encoding = dict_get_encoding (dict);
952 struct substring product;
953 struct substring label;
955 /* Convert file label to UTF-8 and put it into DICT. */
956 label = recode_substring_pool ("UTF-8", dict_encoding,
957 ss_cstr (header->file_label), r->pool);
958 ss_trim (&label, ss_cstr (" "));
959 label.string[label.length] = '\0';
960 dict_set_label (dict, label.string);
962 /* Put creation date and time in UTF-8 into INFO. */
963 info->creation_date = recode_string ("UTF-8", dict_encoding,
964 header->creation_date, -1);
965 info->creation_time = recode_string ("UTF-8", dict_encoding,
966 header->creation_time, -1);
968 /* Put product name into INFO, dropping eye-catcher string if present. */
969 product = recode_substring_pool ("UTF-8", dict_encoding,
970 ss_cstr (header->eye_catcher), r->pool);
971 ss_match_string (&product, ss_cstr ("@(#) SPSS DATA FILE"));
972 ss_trim (&product, ss_cstr (" "));
973 info->product = ss_xstrdup (product);
976 /* Reads a variable (type 2) record from R and adds the
977 corresponding variable to DICT.
978 Also skips past additional variable records for long string
981 parse_variable_records (struct sfm_reader *r, struct dictionary *dict,
982 struct sfm_var_record *var_recs, size_t n_var_recs)
984 const char *dict_encoding = dict_get_encoding (dict);
985 struct sfm_var_record *rec;
988 for (rec = var_recs; rec < &var_recs[n_var_recs]; )
990 struct variable *var;
995 name = recode_string_pool ("UTF-8", dict_encoding,
996 rec->name, 8, r->pool);
997 name[strcspn (name, " ")] = '\0';
999 if (!dict_id_is_valid (dict, name, false)
1000 || name[0] == '$' || name[0] == '#')
1001 sys_error (r, rec->pos, _("Invalid variable name `%s'."), name);
1003 if (rec->width < 0 || rec->width > 255)
1004 sys_error (r, rec->pos,
1005 _("Bad width %d for variable %s."), rec->width, name);
1007 var = rec->var = dict_create_var (dict, name, rec->width);
1009 sys_error (r, rec->pos, _("Duplicate variable name `%s'."), name);
1011 /* Set the short name the same as the long name. */
1012 var_set_short_name (var, 0, name);
1014 /* Get variable label, if any. */
1019 utf8_label = recode_string_pool ("UTF-8", dict_encoding,
1020 rec->label, -1, r->pool);
1021 var_set_label (var, utf8_label, false);
1024 /* Set missing values. */
1025 if (rec->missing_value_code != 0)
1027 int width = var_get_width (var);
1028 struct missing_values mv;
1030 mv_init_pool (r->pool, &mv, width);
1031 if (var_is_numeric (var))
1033 bool has_range = rec->missing_value_code < 0;
1034 int n_discrete = (has_range
1035 ? rec->missing_value_code == -3
1036 : rec->missing_value_code);
1041 double low = parse_float (r, rec->missing, 0);
1042 double high = parse_float (r, rec->missing, 8);
1043 mv_add_range (&mv, low, high);
1047 for (i = 0; i < n_discrete; i++)
1049 mv_add_num (&mv, parse_float (r, rec->missing, ofs));
1057 value_init_pool (r->pool, &value, width);
1058 value_set_missing (&value, width);
1059 for (i = 0; i < rec->missing_value_code; i++)
1061 uint8_t *s = value_str_rw (&value, width);
1062 memcpy (s, rec->missing + 8 * i, MIN (width, 8));
1063 mv_add_str (&mv, s);
1066 var_set_missing_values (var, &mv);
1070 parse_format_spec (r, rec->pos + 12, rec->print_format,
1071 PRINT_FORMAT, var, &n_warnings);
1072 parse_format_spec (r, rec->pos + 16, rec->write_format,
1073 WRITE_FORMAT, var, &n_warnings);
1075 /* Account for values.
1076 Skip long string continuation records, if any. */
1077 n_values = rec->width == 0 ? 1 : DIV_RND_UP (rec->width, 8);
1078 for (i = 1; i < n_values; i++)
1079 if (i + (rec - var_recs) >= n_var_recs || rec[i].width != -1)
1080 sys_error (r, rec->pos, _("Missing string continuation record."));
1085 /* Translates the format spec from sysfile format to internal
1088 parse_format_spec (struct sfm_reader *r, off_t pos, unsigned int format,
1089 enum which_format which, struct variable *v,
1092 const int max_warnings = 8;
1093 uint8_t raw_type = format >> 16;
1094 uint8_t w = format >> 8;
1103 ok = (fmt_from_io (raw_type, &f.type)
1104 && fmt_check_output (&f)
1105 && fmt_check_width_compat (&f, var_get_width (v)));
1110 if (which == PRINT_FORMAT)
1111 var_set_print_format (v, &f);
1113 var_set_write_format (v, &f);
1115 else if (format == 0)
1117 /* Actually observed in the wild. No point in warning about it. */
1119 else if (++*n_warnings <= max_warnings)
1121 if (which == PRINT_FORMAT)
1122 sys_warn (r, pos, _("Variable %s with width %d has invalid print "
1124 var_get_name (v), var_get_width (v), format);
1126 sys_warn (r, pos, _("Variable %s with width %d has invalid write "
1128 var_get_name (v), var_get_width (v), format);
1130 if (*n_warnings == max_warnings)
1131 sys_warn (r, -1, _("Suppressing further invalid format warnings."));
1136 parse_document (struct dictionary *dict, struct sfm_document_record *record)
1140 for (p = record->documents;
1141 p < record->documents + DOC_LINE_LENGTH * record->n_lines;
1142 p += DOC_LINE_LENGTH)
1144 struct substring line;
1146 line = recode_substring_pool ("UTF-8", dict_get_encoding (dict),
1147 ss_buffer (p, DOC_LINE_LENGTH), NULL);
1148 ss_rtrim (&line, ss_cstr (" "));
1149 line.string[line.length] = '\0';
1151 dict_add_document_line (dict, line.string, false);
1157 /* Parses record type 7, subtype 3. */
1159 parse_machine_integer_info (struct sfm_reader *r,
1160 const struct sfm_extension_record *record,
1161 struct sfm_read_info *info)
1163 int float_representation, expected_float_format;
1164 int integer_representation, expected_integer_format;
1166 /* Save version info. */
1167 info->version_major = parse_int (r, record->data, 0);
1168 info->version_minor = parse_int (r, record->data, 4);
1169 info->version_revision = parse_int (r, record->data, 8);
1171 /* Check floating point format. */
1172 float_representation = parse_int (r, record->data, 16);
1173 if (r->float_format == FLOAT_IEEE_DOUBLE_BE
1174 || r->float_format == FLOAT_IEEE_DOUBLE_LE)
1175 expected_float_format = 1;
1176 else if (r->float_format == FLOAT_Z_LONG)
1177 expected_float_format = 2;
1178 else if (r->float_format == FLOAT_VAX_G || r->float_format == FLOAT_VAX_D)
1179 expected_float_format = 3;
1182 if (float_representation != expected_float_format)
1183 sys_error (r, record->pos, _("Floating-point representation indicated by "
1184 "system file (%d) differs from expected (%d)."),
1185 float_representation, expected_float_format);
1187 /* Check integer format. */
1188 integer_representation = parse_int (r, record->data, 24);
1189 if (r->integer_format == INTEGER_MSB_FIRST)
1190 expected_integer_format = 1;
1191 else if (r->integer_format == INTEGER_LSB_FIRST)
1192 expected_integer_format = 2;
1195 if (integer_representation != expected_integer_format)
1196 sys_warn (r, record->pos,
1197 _("Integer format indicated by system file (%d) "
1198 "differs from expected (%d)."),
1199 integer_representation, expected_integer_format);
1204 choose_encoding (struct sfm_reader *r,
1205 const struct sfm_header_record *header,
1206 const struct sfm_extension_record *ext_integer,
1207 const struct sfm_extension_record *ext_encoding)
1209 /* The EXT_ENCODING record is a more reliable way to determine dictionary
1212 return ext_encoding->data;
1214 /* But EXT_INTEGER is better than nothing as a fallback. */
1217 int codepage = parse_int (r, ext_integer->data, 7 * 4);
1218 const char *encoding;
1227 /* These ostensibly mean "7-bit ASCII" and "8-bit ASCII"[sic]
1228 respectively. However, there are known to be many files in the wild
1229 with character code 2, yet have data which are clearly not ASCII.
1230 Therefore we ignore these values. */
1237 encoding = sys_get_encoding_from_codepage (codepage);
1238 if (encoding != NULL)
1244 /* If the file magic number is EBCDIC then its character data is too. */
1245 if (!strcmp (header->magic, EBCDIC_MAGIC))
1248 return locale_charset ();
1251 /* Parses record type 7, subtype 4. */
1253 parse_machine_float_info (struct sfm_reader *r,
1254 const struct sfm_extension_record *record)
1256 double sysmis = parse_float (r, record->data, 0);
1257 double highest = parse_float (r, record->data, 8);
1258 double lowest = parse_float (r, record->data, 16);
1260 if (sysmis != SYSMIS)
1261 sys_warn (r, record->pos, _("File specifies unexpected value %g as %s."),
1264 if (highest != HIGHEST)
1265 sys_warn (r, record->pos, _("File specifies unexpected value %g as %s."),
1266 highest, "HIGHEST");
1268 if (lowest != LOWEST)
1269 sys_warn (r, record->pos, _("File specifies unexpected value %g as %s."),
1273 /* Parses record type 7, subtype 7 or 19. */
1275 parse_mrsets (struct sfm_reader *r, const struct sfm_extension_record *record,
1276 struct dictionary *dict)
1278 struct text_record *text;
1279 struct mrset *mrset;
1281 text = open_text_record (r, record, false);
1284 const char *counted = NULL;
1287 struct stringi_set var_names;
1288 size_t allocated_vars;
1292 mrset = xzalloc (sizeof *mrset);
1294 name = text_get_token (text, ss_cstr ("="), NULL);
1297 mrset->name = recode_string ("UTF-8", r->encoding, name, -1);
1299 if (mrset->name[0] != '$')
1301 sys_warn (r, record->pos,
1302 _("`%s' does not begin with `$' at offset %zu "
1303 "in MRSETS record."), mrset->name, text_pos (text));
1307 if (text_match (text, 'C'))
1309 mrset->type = MRSET_MC;
1310 if (!text_match (text, ' '))
1312 sys_warn (r, record->pos,
1313 _("Missing space following `%c' at offset %zu "
1314 "in MRSETS record."), 'C', text_pos (text));
1318 else if (text_match (text, 'D'))
1320 mrset->type = MRSET_MD;
1321 mrset->cat_source = MRSET_VARLABELS;
1323 else if (text_match (text, 'E'))
1327 mrset->type = MRSET_MD;
1328 mrset->cat_source = MRSET_COUNTEDVALUES;
1329 if (!text_match (text, ' '))
1331 sys_warn (r, record->pos,
1332 _("Missing space following `%c' at offset %zu "
1333 "in MRSETS record."), 'E', text_pos (text));
1337 number = text_get_token (text, ss_cstr (" "), NULL);
1338 if (!strcmp (number, "11"))
1339 mrset->label_from_var_label = true;
1340 else if (strcmp (number, "1"))
1341 sys_warn (r, record->pos,
1342 _("Unexpected label source value `%s' following `E' "
1343 "at offset %zu in MRSETS record."),
1344 number, text_pos (text));
1348 sys_warn (r, record->pos,
1349 _("Missing `C', `D', or `E' at offset %zu "
1350 "in MRSETS record."),
1355 if (mrset->type == MRSET_MD)
1357 counted = text_parse_counted_string (r, text);
1358 if (counted == NULL)
1362 label = text_parse_counted_string (r, text);
1365 if (label[0] != '\0')
1366 mrset->label = recode_string ("UTF-8", r->encoding, label, -1);
1368 stringi_set_init (&var_names);
1373 const char *raw_var_name;
1374 struct variable *var;
1377 raw_var_name = text_get_token (text, ss_cstr (" \n"), &delimiter);
1378 if (raw_var_name == NULL)
1380 sys_warn (r, record->pos,
1381 _("Missing new-line parsing variable names "
1382 "at offset %zu in MRSETS record."),
1386 var_name = recode_string ("UTF-8", r->encoding, raw_var_name, -1);
1388 var = dict_lookup_var (dict, var_name);
1394 if (!stringi_set_insert (&var_names, var_name))
1396 sys_warn (r, record->pos,
1397 _("Duplicate variable name %s "
1398 "at offset %zu in MRSETS record."),
1399 var_name, text_pos (text));
1405 if (mrset->label == NULL && mrset->label_from_var_label
1406 && var_has_label (var))
1407 mrset->label = xstrdup (var_get_label (var));
1410 && var_get_type (var) != var_get_type (mrset->vars[0]))
1412 sys_warn (r, record->pos,
1413 _("MRSET %s contains both string and "
1414 "numeric variables."), name);
1417 width = MIN (width, var_get_width (var));
1419 if (mrset->n_vars >= allocated_vars)
1420 mrset->vars = x2nrealloc (mrset->vars, &allocated_vars,
1421 sizeof *mrset->vars);
1422 mrset->vars[mrset->n_vars++] = var;
1424 while (delimiter != '\n');
1426 if (mrset->n_vars < 2)
1428 sys_warn (r, record->pos,
1429 _("MRSET %s has only %zu variables."), mrset->name,
1431 mrset_destroy (mrset);
1432 stringi_set_destroy (&var_names);
1436 if (mrset->type == MRSET_MD)
1438 mrset->width = width;
1439 value_init (&mrset->counted, width);
1441 mrset->counted.f = c_strtod (counted, NULL);
1443 value_copy_str_rpad (&mrset->counted, width,
1444 (const uint8_t *) counted, ' ');
1447 dict_add_mrset (dict, mrset);
1449 stringi_set_destroy (&var_names);
1451 mrset_destroy (mrset);
1452 close_text_record (r, text);
1455 /* Read record type 7, subtype 11, which specifies how variables
1456 should be displayed in GUI environments. */
1458 parse_display_parameters (struct sfm_reader *r,
1459 const struct sfm_extension_record *record,
1460 struct dictionary *dict)
1462 bool includes_width;
1463 bool warned = false;
1468 n_vars = dict_get_var_cnt (dict);
1469 if (record->count == 3 * n_vars)
1470 includes_width = true;
1471 else if (record->count == 2 * n_vars)
1472 includes_width = false;
1475 sys_warn (r, record->pos,
1476 _("Extension 11 has bad count %zu (for %zu variables)."),
1477 record->count, n_vars);
1482 for (i = 0; i < n_vars; ++i)
1484 struct variable *v = dict_get_var (dict, i);
1485 int measure, width, align;
1487 measure = parse_int (r, record->data, ofs);
1492 width = parse_int (r, record->data, ofs);
1498 align = parse_int (r, record->data, ofs);
1501 /* SPSS 14 sometimes seems to set string variables' measure
1503 if (0 == measure && var_is_alpha (v))
1506 if (measure < 1 || measure > 3 || align < 0 || align > 2)
1509 sys_warn (r, record->pos,
1510 _("Invalid variable display parameters for variable "
1511 "%zu (%s). Default parameters substituted."),
1512 i, var_get_name (v));
1517 var_set_measure (v, (measure == 1 ? MEASURE_NOMINAL
1518 : measure == 2 ? MEASURE_ORDINAL
1520 var_set_alignment (v, (align == 0 ? ALIGN_LEFT
1521 : align == 1 ? ALIGN_RIGHT
1524 /* Older versions (SPSS 9.0) sometimes set the display
1525 width to zero. This causes confusion in the GUI, so
1526 only set the width if it is nonzero. */
1528 var_set_display_width (v, width);
1533 rename_var_and_save_short_names (struct dictionary *dict, struct variable *var,
1534 const char *new_name)
1536 size_t n_short_names;
1540 /* Renaming a variable may clear its short names, but we
1541 want to retain them, so we save them and re-set them
1543 n_short_names = var_get_short_name_cnt (var);
1544 short_names = xnmalloc (n_short_names, sizeof *short_names);
1545 for (i = 0; i < n_short_names; i++)
1547 const char *s = var_get_short_name (var, i);
1548 short_names[i] = s != NULL ? xstrdup (s) : NULL;
1551 /* Set long name. */
1552 dict_rename_var (dict, var, new_name);
1554 /* Restore short names. */
1555 for (i = 0; i < n_short_names; i++)
1557 var_set_short_name (var, i, short_names[i]);
1558 free (short_names[i]);
1563 /* Parses record type 7, subtype 13, which gives the long name that corresponds
1564 to each short name. Modifies variable names in DICT accordingly. */
1566 parse_long_var_name_map (struct sfm_reader *r,
1567 const struct sfm_extension_record *record,
1568 struct dictionary *dict)
1570 struct text_record *text;
1571 struct variable *var;
1576 /* There are no long variable names. Use the short variable names,
1577 converted to lowercase, as the long variable names. */
1580 for (i = 0; i < dict_get_var_cnt (dict); i++)
1582 struct variable *var = dict_get_var (dict, i);
1585 new_name = utf8_to_lower (var_get_name (var));
1586 rename_var_and_save_short_names (dict, var, new_name);
1593 /* Rename each of the variables, one by one. (In a correctly constructed
1594 system file, this cannot create any intermediate duplicate variable names,
1595 because all of the new variable names are longer than any of the old
1596 variable names and thus there cannot be any overlaps.) */
1597 text = open_text_record (r, record, true);
1598 while (read_variable_to_value_pair (r, dict, text, &var, &long_name))
1600 /* Validate long name. */
1601 if (!dict_id_is_valid (dict, long_name, false))
1603 sys_warn (r, record->pos,
1604 _("Long variable mapping from %s to invalid "
1605 "variable name `%s'."),
1606 var_get_name (var), long_name);
1610 /* Identify any duplicates. */
1611 if (utf8_strcasecmp (var_get_short_name (var, 0), long_name)
1612 && dict_lookup_var (dict, long_name) != NULL)
1614 sys_warn (r, record->pos,
1615 _("Duplicate long variable name `%s'."), long_name);
1619 rename_var_and_save_short_names (dict, var, long_name);
1621 close_text_record (r, text);
1624 /* Reads record type 7, subtype 14, which gives the real length
1625 of each very long string. Rearranges DICT accordingly. */
1627 parse_long_string_map (struct sfm_reader *r,
1628 const struct sfm_extension_record *record,
1629 struct dictionary *dict)
1631 struct text_record *text;
1632 struct variable *var;
1635 text = open_text_record (r, record, true);
1636 while (read_variable_to_value_pair (r, dict, text, &var, &length_s))
1638 size_t idx = var_get_dict_index (var);
1644 length = strtol (length_s, NULL, 10);
1645 if (length < 1 || length > MAX_STRING)
1647 sys_warn (r, record->pos,
1648 _("%s listed as string of invalid length %s "
1649 "in very long string record."),
1650 var_get_name (var), length_s);
1654 /* Check segments. */
1655 segment_cnt = sfm_width_to_segments (length);
1656 if (segment_cnt == 1)
1658 sys_warn (r, record->pos,
1659 _("%s listed in very long string record with width %s, "
1660 "which requires only one segment."),
1661 var_get_name (var), length_s);
1664 if (idx + segment_cnt > dict_get_var_cnt (dict))
1665 sys_error (r, record->pos,
1666 _("Very long string %s overflows dictionary."),
1667 var_get_name (var));
1669 /* Get the short names from the segments and check their
1671 for (i = 0; i < segment_cnt; i++)
1673 struct variable *seg = dict_get_var (dict, idx + i);
1674 int alloc_width = sfm_segment_alloc_width (length, i);
1675 int width = var_get_width (seg);
1678 var_set_short_name (var, i, var_get_short_name (seg, 0));
1679 if (ROUND_UP (width, 8) != ROUND_UP (alloc_width, 8))
1680 sys_error (r, record->pos,
1681 _("Very long string with width %ld has segment %d "
1682 "of width %d (expected %d)."),
1683 length, i, width, alloc_width);
1685 dict_delete_consecutive_vars (dict, idx + 1, segment_cnt - 1);
1686 var_set_width (var, length);
1688 close_text_record (r, text);
1689 dict_compact_values (dict);
1693 parse_value_labels (struct sfm_reader *r, struct dictionary *dict,
1694 const struct sfm_var_record *var_recs, size_t n_var_recs,
1695 const struct sfm_value_label_record *record)
1697 struct variable **vars;
1701 utf8_labels = pool_nmalloc (r->pool, sizeof *utf8_labels, record->n_labels);
1702 for (i = 0; i < record->n_labels; i++)
1703 utf8_labels[i] = recode_string_pool ("UTF-8", dict_get_encoding (dict),
1704 record->labels[i].label, -1,
1707 vars = pool_nmalloc (r->pool, record->n_vars, sizeof *vars);
1708 for (i = 0; i < record->n_vars; i++)
1709 vars[i] = lookup_var_by_index (r, record->pos,
1710 var_recs, n_var_recs, record->vars[i]);
1712 for (i = 1; i < record->n_vars; i++)
1713 if (var_get_type (vars[i]) != var_get_type (vars[0]))
1714 sys_error (r, record->pos,
1715 _("Variables associated with value label are not all of "
1716 "identical type. Variable %s is %s, but variable "
1718 var_get_name (vars[0]),
1719 var_is_numeric (vars[0]) ? _("numeric") : _("string"),
1720 var_get_name (vars[i]),
1721 var_is_numeric (vars[i]) ? _("numeric") : _("string"));
1723 for (i = 0; i < record->n_vars; i++)
1725 struct variable *var = vars[i];
1729 width = var_get_width (var);
1731 sys_error (r, record->pos,
1732 _("Value labels may not be added to long string "
1733 "variables (e.g. %s) using records types 3 and 4."),
1734 var_get_name (var));
1736 for (j = 0; j < record->n_labels; j++)
1738 struct sfm_value_label *label = &record->labels[j];
1741 value_init (&value, width);
1743 value.f = parse_float (r, label->value, 0);
1745 memcpy (value_str_rw (&value, width), label->value, width);
1747 if (!var_add_value_label (var, &value, utf8_labels[j]))
1749 if (var_is_numeric (var))
1750 sys_warn (r, record->pos,
1751 _("Duplicate value label for %g on %s."),
1752 value.f, var_get_name (var));
1754 sys_warn (r, record->pos,
1755 _("Duplicate value label for `%.*s' on %s."),
1756 width, value_str (&value, width),
1757 var_get_name (var));
1760 value_destroy (&value, width);
1764 pool_free (r->pool, vars);
1765 for (i = 0; i < record->n_labels; i++)
1766 pool_free (r->pool, utf8_labels[i]);
1767 pool_free (r->pool, utf8_labels);
1770 static struct variable *
1771 lookup_var_by_index (struct sfm_reader *r, off_t offset,
1772 const struct sfm_var_record *var_recs, size_t n_var_recs,
1775 const struct sfm_var_record *rec;
1777 if (idx < 1 || idx > n_var_recs)
1779 sys_error (r, offset,
1780 _("Variable index %d not in valid range 1...%zu."),
1785 rec = &var_recs[idx - 1];
1786 if (rec->var == NULL)
1788 sys_error (r, offset,
1789 _("Variable index %d refers to long string continuation."),
1797 /* Parses a set of custom attributes from TEXT into ATTRS.
1798 ATTRS may be a null pointer, in which case the attributes are
1799 read but discarded. */
1801 parse_attributes (struct sfm_reader *r, struct text_record *text,
1802 struct attrset *attrs)
1806 struct attribute *attr;
1810 /* Parse the key. */
1811 key = text_get_token (text, ss_cstr ("("), NULL);
1815 attr = attribute_create (key);
1816 for (index = 1; ; index++)
1818 /* Parse the value. */
1822 value = text_get_token (text, ss_cstr ("\n"), NULL);
1825 text_warn (r, text, _("Error parsing attribute value %s[%d]."),
1830 length = strlen (value);
1831 if (length >= 2 && value[0] == '\'' && value[length - 1] == '\'')
1833 value[length - 1] = '\0';
1834 attribute_add_value (attr, value + 1);
1839 _("Attribute value %s[%d] is not quoted: %s."),
1841 attribute_add_value (attr, value);
1844 /* Was this the last value for this attribute? */
1845 if (text_match (text, ')'))
1849 attrset_add (attrs, attr);
1851 attribute_destroy (attr);
1853 while (!text_match (text, '/'));
1856 /* Reads record type 7, subtype 17, which lists custom
1857 attributes on the data file. */
1859 parse_data_file_attributes (struct sfm_reader *r,
1860 const struct sfm_extension_record *record,
1861 struct dictionary *dict)
1863 struct text_record *text = open_text_record (r, record, true);
1864 parse_attributes (r, text, dict_get_attributes (dict));
1865 close_text_record (r, text);
1868 /* Parses record type 7, subtype 18, which lists custom
1869 attributes on individual variables. */
1871 parse_variable_attributes (struct sfm_reader *r,
1872 const struct sfm_extension_record *record,
1873 struct dictionary *dict)
1875 struct text_record *text;
1876 struct variable *var;
1878 text = open_text_record (r, record, true);
1879 while (text_read_variable_name (r, dict, text, ss_cstr (":"), &var))
1880 parse_attributes (r, text, var != NULL ? var_get_attributes (var) : NULL);
1881 close_text_record (r, text);
1885 check_overflow (struct sfm_reader *r,
1886 const struct sfm_extension_record *record,
1887 size_t ofs, size_t length)
1889 size_t end = record->size * record->count;
1890 if (length >= end || ofs + length > end)
1891 sys_error (r, record->pos + end,
1892 _("Long string value label record ends unexpectedly."));
1896 parse_long_string_value_labels (struct sfm_reader *r,
1897 const struct sfm_extension_record *record,
1898 struct dictionary *dict)
1900 const char *dict_encoding = dict_get_encoding (dict);
1901 size_t end = record->size * record->count;
1908 struct variable *var;
1913 /* Parse variable name length. */
1914 check_overflow (r, record, ofs, 4);
1915 var_name_len = parse_int (r, record->data, ofs);
1918 /* Parse variable name, width, and number of labels. */
1919 check_overflow (r, record, ofs, var_name_len + 8);
1920 var_name = recode_string_pool ("UTF-8", dict_encoding,
1921 (const char *) record->data + ofs,
1922 var_name_len, r->pool);
1923 width = parse_int (r, record->data, ofs + var_name_len);
1924 n_labels = parse_int (r, record->data, ofs + var_name_len + 4);
1925 ofs += var_name_len + 8;
1927 /* Look up 'var' and validate. */
1928 var = dict_lookup_var (dict, var_name);
1930 sys_warn (r, record->pos + ofs,
1931 _("Ignoring long string value record for "
1932 "unknown variable %s."), var_name);
1933 else if (var_is_numeric (var))
1935 sys_warn (r, record->pos + ofs,
1936 _("Ignoring long string value record for "
1937 "numeric variable %s."), var_name);
1940 else if (width != var_get_width (var))
1942 sys_warn (r, record->pos + ofs,
1943 _("Ignoring long string value record for variable %s "
1944 "because the record's width (%d) does not match the "
1945 "variable's width (%d)."),
1946 var_name, width, var_get_width (var));
1951 value_init_pool (r->pool, &value, width);
1952 for (i = 0; i < n_labels; i++)
1954 size_t value_length, label_length;
1955 bool skip = var == NULL;
1957 /* Parse value length. */
1958 check_overflow (r, record, ofs, 4);
1959 value_length = parse_int (r, record->data, ofs);
1963 check_overflow (r, record, ofs, value_length);
1966 if (value_length == width)
1967 memcpy (value_str_rw (&value, width),
1968 (const uint8_t *) record->data + ofs, width);
1971 sys_warn (r, record->pos + ofs,
1972 _("Ignoring long string value %zu for variable "
1973 "%s, with width %d, that has bad value "
1975 i, var_get_name (var), width, value_length);
1979 ofs += value_length;
1981 /* Parse label length. */
1982 check_overflow (r, record, ofs, 4);
1983 label_length = parse_int (r, record->data, ofs);
1987 check_overflow (r, record, ofs, label_length);
1992 label = recode_string_pool ("UTF-8", dict_encoding,
1993 (const char *) record->data + ofs,
1994 label_length, r->pool);
1995 if (!var_add_value_label (var, &value, label))
1996 sys_warn (r, record->pos + ofs,
1997 _("Duplicate value label for `%.*s' on %s."),
1998 width, value_str (&value, width),
1999 var_get_name (var));
2000 pool_free (r->pool, label);
2002 ofs += label_length;
2009 static void partial_record (struct sfm_reader *r)
2012 static void read_error (struct casereader *, const struct sfm_reader *);
2014 static bool read_case_number (struct sfm_reader *, double *);
2015 static bool read_case_string (struct sfm_reader *, uint8_t *, size_t);
2016 static int read_opcode (struct sfm_reader *);
2017 static bool read_compressed_number (struct sfm_reader *, double *);
2018 static bool read_compressed_string (struct sfm_reader *, uint8_t *);
2019 static bool read_whole_strings (struct sfm_reader *, uint8_t *, size_t);
2020 static bool skip_whole_strings (struct sfm_reader *, size_t);
2022 /* Reads and returns one case from READER's file. Returns a null
2023 pointer if not successful. */
2024 static struct ccase *
2025 sys_file_casereader_read (struct casereader *reader, void *r_)
2027 struct sfm_reader *r = r_;
2028 struct ccase *volatile c;
2034 c = case_create (r->proto);
2035 if (setjmp (r->bail_out))
2037 casereader_force_error (reader);
2042 for (i = 0; i < r->sfm_var_cnt; i++)
2044 struct sfm_var *sv = &r->sfm_vars[i];
2045 union value *v = case_data_rw_idx (c, sv->case_index);
2047 if (sv->var_width == 0)
2049 if (!read_case_number (r, &v->f))
2054 uint8_t *s = value_str_rw (v, sv->var_width);
2055 if (!read_case_string (r, s + sv->offset, sv->segment_width))
2057 if (!skip_whole_strings (r, ROUND_DOWN (sv->padding, 8)))
2066 if (r->case_cnt != -1)
2067 read_error (reader, r);
2072 /* Issues an error that R ends in a partial record. */
2074 partial_record (struct sfm_reader *r)
2076 sys_error (r, r->pos, _("File ends in partial case."));
2079 /* Issues an error that an unspecified error occurred SFM, and
2082 read_error (struct casereader *r, const struct sfm_reader *sfm)
2084 msg (ME, _("Error reading case from file %s."), fh_get_name (sfm->fh));
2085 casereader_force_error (r);
2088 /* Reads a number from R and stores its value in *D.
2089 If R is compressed, reads a compressed number;
2090 otherwise, reads a number in the regular way.
2091 Returns true if successful, false if end of file is
2092 reached immediately. */
2094 read_case_number (struct sfm_reader *r, double *d)
2099 if (!try_read_bytes (r, number, sizeof number))
2101 float_convert (r->float_format, number, FLOAT_NATIVE_DOUBLE, d);
2105 return read_compressed_number (r, d);
2108 /* Reads LENGTH string bytes from R into S.
2109 Always reads a multiple of 8 bytes; if LENGTH is not a
2110 multiple of 8, then extra bytes are read and discarded without
2112 Reads compressed strings if S is compressed.
2113 Returns true if successful, false if end of file is
2114 reached immediately. */
2116 read_case_string (struct sfm_reader *r, uint8_t *s, size_t length)
2118 size_t whole = ROUND_DOWN (length, 8);
2119 size_t partial = length % 8;
2123 if (!read_whole_strings (r, s, whole))
2130 if (!read_whole_strings (r, bounce, sizeof bounce))
2136 memcpy (s + whole, bounce, partial);
2142 /* Reads and returns the next compression opcode from R. */
2144 read_opcode (struct sfm_reader *r)
2146 assert (r->compressed);
2150 if (r->opcode_idx >= sizeof r->opcodes)
2152 if (!try_read_bytes (r, r->opcodes, sizeof r->opcodes))
2156 opcode = r->opcodes[r->opcode_idx++];
2163 /* Reads a compressed number from R and stores its value in D.
2164 Returns true if successful, false if end of file is
2165 reached immediately. */
2167 read_compressed_number (struct sfm_reader *r, double *d)
2169 int opcode = read_opcode (r);
2177 *d = read_float (r);
2181 float_convert (r->float_format, " ", FLOAT_NATIVE_DOUBLE, d);
2182 if (!r->corruption_warning)
2184 r->corruption_warning = true;
2185 sys_warn (r, r->pos,
2186 _("Possible compressed data corruption: "
2187 "compressed spaces appear in numeric field."));
2196 *d = opcode - r->bias;
2203 /* Reads a compressed 8-byte string segment from R and stores it
2205 Returns true if successful, false if end of file is
2206 reached immediately. */
2208 read_compressed_string (struct sfm_reader *r, uint8_t *dst)
2210 int opcode = read_opcode (r);
2218 read_bytes (r, dst, 8);
2222 memset (dst, ' ', 8);
2227 double value = opcode - r->bias;
2228 float_convert (FLOAT_NATIVE_DOUBLE, &value, r->float_format, dst);
2231 /* This has actually been seen "in the wild". The submitter of the
2232 file that showed that the contents decoded as spaces, but they
2233 were at the end of the field so it's possible that the null
2234 bytes just acted as null terminators. */
2236 else if (!r->corruption_warning)
2238 r->corruption_warning = true;
2239 sys_warn (r, r->pos,
2240 _("Possible compressed data corruption: "
2241 "string contains compressed integer (opcode %d)."),
2251 /* Reads LENGTH string bytes from R into S.
2252 LENGTH must be a multiple of 8.
2253 Reads compressed strings if S is compressed.
2254 Returns true if successful, false if end of file is
2255 reached immediately. */
2257 read_whole_strings (struct sfm_reader *r, uint8_t *s, size_t length)
2259 assert (length % 8 == 0);
2261 return try_read_bytes (r, s, length);
2265 for (ofs = 0; ofs < length; ofs += 8)
2266 if (!read_compressed_string (r, s + ofs))
2276 /* Skips LENGTH string bytes from R.
2277 LENGTH must be a multiple of 8.
2278 (LENGTH is also limited to 1024, but that's only because the
2279 current caller never needs more than that many bytes.)
2280 Returns true if successful, false if end of file is
2281 reached immediately. */
2283 skip_whole_strings (struct sfm_reader *r, size_t length)
2285 uint8_t buffer[1024];
2286 assert (length < sizeof buffer);
2287 return read_whole_strings (r, buffer, length);
2290 /* Helpers for reading records that contain structured text
2293 /* Maximum number of warnings to issue for a single text
2295 #define MAX_TEXT_WARNINGS 5
2300 struct substring buffer; /* Record contents. */
2301 off_t start; /* Starting offset in file. */
2302 size_t pos; /* Current position in buffer. */
2303 int n_warnings; /* Number of warnings issued or suppressed. */
2304 bool recoded; /* Recoded into UTF-8? */
2307 static struct text_record *
2308 open_text_record (struct sfm_reader *r,
2309 const struct sfm_extension_record *record,
2310 bool recode_to_utf8)
2312 struct text_record *text;
2313 struct substring raw;
2315 text = pool_alloc (r->pool, sizeof *text);
2316 raw = ss_buffer (record->data, record->size * record->count);
2317 text->start = record->pos;
2318 text->buffer = (recode_to_utf8
2319 ? recode_substring_pool ("UTF-8", r->encoding, raw, r->pool)
2322 text->n_warnings = 0;
2323 text->recoded = recode_to_utf8;
2328 /* Closes TEXT, frees its storage, and issues a final warning
2329 about suppressed warnings if necesary. */
2331 close_text_record (struct sfm_reader *r, struct text_record *text)
2333 if (text->n_warnings > MAX_TEXT_WARNINGS)
2334 sys_warn (r, -1, _("Suppressed %d additional related warnings."),
2335 text->n_warnings - MAX_TEXT_WARNINGS);
2337 pool_free (r->pool, ss_data (text->buffer));
2340 /* Reads a variable=value pair from TEXT.
2341 Looks up the variable in DICT and stores it into *VAR.
2342 Stores a null-terminated value into *VALUE. */
2344 read_variable_to_value_pair (struct sfm_reader *r, struct dictionary *dict,
2345 struct text_record *text,
2346 struct variable **var, char **value)
2350 if (!text_read_short_name (r, dict, text, ss_cstr ("="), var))
2353 *value = text_get_token (text, ss_buffer ("\t\0", 2), NULL);
2357 text->pos += ss_span (ss_substr (text->buffer, text->pos, SIZE_MAX),
2358 ss_buffer ("\t\0", 2));
2366 text_read_variable_name (struct sfm_reader *r, struct dictionary *dict,
2367 struct text_record *text, struct substring delimiters,
2368 struct variable **var)
2372 name = text_get_token (text, delimiters, NULL);
2376 *var = dict_lookup_var (dict, name);
2380 text_warn (r, text, _("Dictionary record refers to unknown variable %s."),
2387 text_read_short_name (struct sfm_reader *r, struct dictionary *dict,
2388 struct text_record *text, struct substring delimiters,
2389 struct variable **var)
2391 char *short_name = text_get_token (text, delimiters, NULL);
2392 if (short_name == NULL)
2395 *var = dict_lookup_var (dict, short_name);
2397 text_warn (r, text, _("Dictionary record refers to unknown variable %s."),
2402 /* Displays a warning for the current file position, limiting the
2403 number to MAX_TEXT_WARNINGS for TEXT. */
2405 text_warn (struct sfm_reader *r, struct text_record *text,
2406 const char *format, ...)
2408 if (text->n_warnings++ < MAX_TEXT_WARNINGS)
2412 va_start (args, format);
2413 sys_msg (r, text->start + text->pos, MW, format, args);
2419 text_get_token (struct text_record *text, struct substring delimiters,
2422 struct substring token;
2425 if (!ss_tokenize (text->buffer, delimiters, &text->pos, &token))
2428 end = &ss_data (token)[ss_length (token)];
2429 if (delimiter != NULL)
2432 return ss_data (token);
2435 /* Reads a integer value expressed in decimal, then a space, then a string that
2436 consists of exactly as many bytes as specified by the integer, then a space,
2437 from TEXT. Returns the string, null-terminated, as a subset of TEXT's
2438 buffer (so the caller should not free the string). */
2440 text_parse_counted_string (struct sfm_reader *r, struct text_record *text)
2448 while (text->pos < text->buffer.length)
2450 int c = text->buffer.string[text->pos];
2451 if (c < '0' || c > '9')
2453 n = (n * 10) + (c - '0');
2456 if (text->pos >= text->buffer.length || start == text->pos)
2458 sys_warn (r, text->start,
2459 _("Expecting digit at offset %zu in MRSETS record."),
2464 if (!text_match (text, ' '))
2466 sys_warn (r, text->start,
2467 _("Expecting space at offset %zu in MRSETS record."),
2472 if (text->pos + n > text->buffer.length)
2474 sys_warn (r, text->start,
2475 _("%zu-byte string starting at offset %zu "
2476 "exceeds record length %zu."),
2477 n, text->pos, text->buffer.length);
2481 s = &text->buffer.string[text->pos];
2484 sys_warn (r, text->start,
2485 _("Expecting space at offset %zu following %zu-byte string."),
2495 text_match (struct text_record *text, char c)
2497 if (text->buffer.string[text->pos] == c)
2506 /* Returns the current byte offset (as converted to UTF-8, if it was converted)
2507 inside the TEXT's string. */
2509 text_pos (const struct text_record *text)
2516 /* Displays a corruption message. */
2518 sys_msg (struct sfm_reader *r, off_t offset,
2519 int class, const char *format, va_list args)
2524 ds_init_empty (&text);
2526 ds_put_format (&text, _("`%s' near offset 0x%llx: "),
2527 fh_get_file_name (r->fh), (long long int) offset);
2529 ds_put_format (&text, _("`%s': "), fh_get_file_name (r->fh));
2530 ds_put_vformat (&text, format, args);
2532 m.category = msg_class_to_category (class);
2533 m.severity = msg_class_to_severity (class);
2539 m.text = ds_cstr (&text);
2544 /* Displays a warning for offset OFFSET in the file. */
2546 sys_warn (struct sfm_reader *r, off_t offset, const char *format, ...)
2550 va_start (args, format);
2551 sys_msg (r, offset, MW, format, args);
2555 /* Displays an error for the current file position,
2556 marks it as in an error state,
2557 and aborts reading it using longjmp. */
2559 sys_error (struct sfm_reader *r, off_t offset, const char *format, ...)
2563 va_start (args, format);
2564 sys_msg (r, offset, ME, format, args);
2568 longjmp (r->bail_out, 1);
2571 /* Reads BYTE_CNT bytes into BUF.
2572 Returns true if exactly BYTE_CNT bytes are successfully read.
2573 Aborts if an I/O error or a partial read occurs.
2574 If EOF_IS_OK, then an immediate end-of-file causes false to be
2575 returned; otherwise, immediate end-of-file causes an abort
2578 read_bytes_internal (struct sfm_reader *r, bool eof_is_ok,
2579 void *buf, size_t byte_cnt)
2581 size_t bytes_read = fread (buf, 1, byte_cnt, r->file);
2582 r->pos += bytes_read;
2583 if (bytes_read == byte_cnt)
2585 else if (ferror (r->file))
2586 sys_error (r, r->pos, _("System error: %s."), strerror (errno));
2587 else if (!eof_is_ok || bytes_read != 0)
2588 sys_error (r, r->pos, _("Unexpected end of file."));
2593 /* Reads BYTE_CNT into BUF.
2594 Aborts upon I/O error or if end-of-file is encountered. */
2596 read_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
2598 read_bytes_internal (r, false, buf, byte_cnt);
2601 /* Reads BYTE_CNT bytes into BUF.
2602 Returns true if exactly BYTE_CNT bytes are successfully read.
2603 Returns false if an immediate end-of-file is encountered.
2604 Aborts if an I/O error or a partial read occurs. */
2606 try_read_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
2608 return read_bytes_internal (r, true, buf, byte_cnt);
2611 /* Reads a 32-bit signed integer from R and returns its value in
2614 read_int (struct sfm_reader *r)
2617 read_bytes (r, integer, sizeof integer);
2618 return integer_get (r->integer_format, integer, sizeof integer);
2621 /* Reads a 64-bit floating-point number from R and returns its
2622 value in host format. */
2624 read_float (struct sfm_reader *r)
2627 read_bytes (r, number, sizeof number);
2628 return float_get_double (r->float_format, number);
2632 parse_int (struct sfm_reader *r, const void *data, size_t ofs)
2634 return integer_get (r->integer_format, (const uint8_t *) data + ofs, 4);
2638 parse_float (struct sfm_reader *r, const void *data, size_t ofs)
2640 return float_get_double (r->float_format, (const uint8_t *) data + ofs);
2643 /* Reads exactly SIZE - 1 bytes into BUFFER
2644 and stores a null byte into BUFFER[SIZE - 1]. */
2646 read_string (struct sfm_reader *r, char *buffer, size_t size)
2649 read_bytes (r, buffer, size - 1);
2650 buffer[size - 1] = '\0';
2653 /* Skips BYTES bytes forward in R. */
2655 skip_bytes (struct sfm_reader *r, size_t bytes)
2660 size_t chunk = MIN (sizeof buffer, bytes);
2661 read_bytes (r, buffer, chunk);
2666 static const struct casereader_class sys_file_casereader_class =
2668 sys_file_casereader_read,
2669 sys_file_casereader_destroy,