1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-2000, 2006-2007, 2009-2013 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"
30 #include "data/attributes.h"
31 #include "data/case.h"
32 #include "data/casereader-provider.h"
33 #include "data/casereader.h"
34 #include "data/dictionary.h"
35 #include "data/file-handle-def.h"
36 #include "data/file-name.h"
37 #include "data/format.h"
38 #include "data/identifier.h"
39 #include "data/missing-values.h"
40 #include "data/mrset.h"
41 #include "data/short-names.h"
42 #include "data/value-labels.h"
43 #include "data/value.h"
44 #include "data/variable.h"
45 #include "libpspp/array.h"
46 #include "libpspp/assertion.h"
47 #include "libpspp/compiler.h"
48 #include "libpspp/i18n.h"
49 #include "libpspp/message.h"
50 #include "libpspp/misc.h"
51 #include "libpspp/pool.h"
52 #include "libpspp/str.h"
53 #include "libpspp/stringi-set.h"
55 #include "gl/c-strtod.h"
56 #include "gl/c-ctype.h"
57 #include "gl/inttostr.h"
58 #include "gl/localcharset.h"
59 #include "gl/minmax.h"
60 #include "gl/unlocked-io.h"
61 #include "gl/xalloc.h"
62 #include "gl/xalloc-oversized.h"
66 #define _(msgid) gettext (msgid)
67 #define N_(msgid) (msgid)
71 /* subtypes 0-2 unknown */
72 EXT_INTEGER = 3, /* Machine integer info. */
73 EXT_FLOAT = 4, /* Machine floating-point info. */
74 EXT_VAR_SETS = 5, /* Variable sets. */
75 EXT_DATE = 6, /* DATE. */
76 EXT_MRSETS = 7, /* Multiple response sets. */
77 EXT_DATA_ENTRY = 8, /* SPSS Data Entry. */
78 /* subtype 9 unknown */
79 EXT_PRODUCT_INFO = 10, /* Extra product info text. */
80 EXT_DISPLAY = 11, /* Variable display parameters. */
81 /* subtype 12 unknown */
82 EXT_LONG_NAMES = 13, /* Long variable names. */
83 EXT_LONG_STRINGS = 14, /* Long strings. */
84 /* subtype 15 unknown */
85 EXT_NCASES = 16, /* Extended number of cases. */
86 EXT_FILE_ATTRS = 17, /* Data file attributes. */
87 EXT_VAR_ATTRS = 18, /* Variable attributes. */
88 EXT_MRSETS2 = 19, /* Multiple response sets (extended). */
89 EXT_ENCODING = 20, /* Character encoding. */
90 EXT_LONG_LABELS = 21, /* Value labels for long strings. */
91 EXT_LONG_MISSING = 22, /* Missing values for long strings. */
92 EXT_DATAVIEW = 24 /* "Format properties in dataview table". */
95 /* Fields from the top-level header record. */
96 struct sfm_header_record
98 char magic[5]; /* First 4 bytes of file, then null. */
99 int weight_idx; /* 0 if unweighted, otherwise a var index. */
100 int nominal_case_size; /* Number of var positions. */
102 /* These correspond to the members of struct sfm_file_info or a dictionary
103 but in the system file's encoding rather than ASCII. */
104 char creation_date[10]; /* "dd mmm yy". */
105 char creation_time[9]; /* "hh:mm:ss". */
106 char eye_catcher[61]; /* Eye-catcher string, then product name. */
107 char file_label[65]; /* File label. */
110 struct sfm_var_record
117 int missing_value_code;
120 struct variable *var;
123 struct sfm_value_label
129 struct sfm_value_label_record
132 struct sfm_value_label *labels;
139 struct sfm_document_record
146 struct sfm_extension_record
148 int subtype; /* Record subtype. */
149 off_t pos; /* Starting offset in file. */
150 size_t size; /* Size of data elements. */
151 size_t count; /* Number of data elements. */
152 void *data; /* Contents. */
155 /* System file reader. */
158 /* Resource tracking. */
159 struct pool *pool; /* All system file state. */
160 jmp_buf bail_out; /* longjmp() target for error handling. */
163 struct file_handle *fh; /* File handle. */
164 struct fh_lock *lock; /* Mutual exclusion for file handle. */
165 FILE *file; /* File stream. */
166 off_t pos; /* Position in file. */
167 bool error; /* I/O or corruption error? */
168 struct caseproto *proto; /* Format of output cases. */
171 enum integer_format integer_format; /* On-disk integer format. */
172 enum float_format float_format; /* On-disk floating point format. */
173 struct sfm_var *sfm_vars; /* Variables. */
174 size_t sfm_var_cnt; /* Number of variables. */
175 casenumber case_cnt; /* Number of cases */
176 const char *encoding; /* String encoding. */
179 enum sfm_compression compression;
180 double bias; /* Compression bias, usually 100.0. */
181 uint8_t opcodes[8]; /* Current block of opcodes. */
182 size_t opcode_idx; /* Next opcode to interpret, 8 if none left. */
183 bool corruption_warning; /* Warned about possible corruption? */
185 /* ZLIB decompression. */
186 long long int ztrailer_ofs; /* Offset of ZLIB trailer at end of file. */
187 #define ZIN_BUF_SIZE 4096
188 uint8_t *zin_buf; /* Inflation input buffer. */
189 #define ZOUT_BUF_SIZE 16384
190 uint8_t *zout_buf; /* Inflation output buffer. */
191 unsigned int zout_end; /* Number of bytes of data in zout_buf. */
192 unsigned int zout_pos; /* First unconsumed byte in zout_buf. */
193 z_stream zstream; /* ZLIB inflater. */
196 static const struct casereader_class sys_file_casereader_class;
198 static bool close_reader (struct sfm_reader *);
200 static struct variable *lookup_var_by_index (struct sfm_reader *, off_t,
201 const struct sfm_var_record *,
204 static void sys_msg (struct sfm_reader *r, off_t, int class,
205 const char *format, va_list args)
206 PRINTF_FORMAT (4, 0);
207 static void sys_warn (struct sfm_reader *, off_t, const char *, ...)
208 PRINTF_FORMAT (3, 4);
209 static void sys_error (struct sfm_reader *, off_t, const char *, ...)
213 static void read_bytes (struct sfm_reader *, void *, size_t);
214 static bool try_read_bytes (struct sfm_reader *, void *, size_t);
215 static int read_int (struct sfm_reader *);
216 static long long int read_int64 (struct sfm_reader *);
217 static void read_string (struct sfm_reader *, char *, size_t);
218 static void skip_bytes (struct sfm_reader *, size_t);
220 /* ZLIB compressed data handling. */
221 static void read_zheader (struct sfm_reader *);
222 static void open_zstream (struct sfm_reader *);
223 static void close_zstream (struct sfm_reader *);
224 static bool read_bytes_zlib (struct sfm_reader *, void *, size_t);
225 static void read_compressed_bytes (struct sfm_reader *, void *, size_t);
226 static bool try_read_compressed_bytes (struct sfm_reader *, void *, size_t);
227 static double read_compressed_float (struct sfm_reader *);
229 static char *fix_line_ends (const char *);
231 static int parse_int (struct sfm_reader *, const void *data, size_t ofs);
232 static double parse_float (struct sfm_reader *, const void *data, size_t ofs);
234 static void read_variable_record (struct sfm_reader *,
235 struct sfm_var_record *);
236 static void read_value_label_record (struct sfm_reader *,
237 struct sfm_value_label_record *,
239 static struct sfm_document_record *read_document_record (struct sfm_reader *);
240 static struct sfm_extension_record *read_extension_record (
241 struct sfm_reader *, int subtype);
242 static void skip_extension_record (struct sfm_reader *, int subtype);
244 static const char *choose_encoding (
246 const struct sfm_header_record *,
247 const struct sfm_extension_record *ext_integer,
248 const struct sfm_extension_record *ext_encoding);
250 static struct text_record *open_text_record (
251 struct sfm_reader *, const struct sfm_extension_record *,
252 bool recode_to_utf8);
253 static void close_text_record (struct sfm_reader *,
254 struct text_record *);
255 static bool read_variable_to_value_pair (struct sfm_reader *,
257 struct text_record *,
258 struct variable **var, char **value);
259 static void text_warn (struct sfm_reader *r, struct text_record *text,
260 const char *format, ...)
261 PRINTF_FORMAT (3, 4);
262 static char *text_get_token (struct text_record *,
263 struct substring delimiters, char *delimiter);
264 static bool text_match (struct text_record *, char c);
265 static bool text_read_variable_name (struct sfm_reader *, struct dictionary *,
266 struct text_record *,
267 struct substring delimiters,
269 static bool text_read_short_name (struct sfm_reader *, struct dictionary *,
270 struct text_record *,
271 struct substring delimiters,
273 static const char *text_parse_counted_string (struct sfm_reader *,
274 struct text_record *);
275 static size_t text_pos (const struct text_record *);
276 static const char *text_get_all (const struct text_record *);
278 static bool close_reader (struct sfm_reader *r);
280 /* Dictionary reader. */
288 static void read_header (struct sfm_reader *, struct sfm_read_info *,
289 struct sfm_header_record *);
290 static void parse_header (struct sfm_reader *,
291 const struct sfm_header_record *,
292 struct sfm_read_info *, struct dictionary *);
293 static void parse_variable_records (struct sfm_reader *, struct dictionary *,
294 struct sfm_var_record *, size_t n);
295 static void parse_format_spec (struct sfm_reader *, off_t pos,
296 unsigned int format, enum which_format,
297 struct variable *, int *format_warning_cnt);
298 static void parse_document (struct dictionary *, struct sfm_document_record *);
299 static void parse_display_parameters (struct sfm_reader *,
300 const struct sfm_extension_record *,
301 struct dictionary *);
302 static void parse_machine_integer_info (struct sfm_reader *,
303 const struct sfm_extension_record *,
304 struct sfm_read_info *);
305 static void parse_machine_float_info (struct sfm_reader *,
306 const struct sfm_extension_record *);
307 static void parse_extra_product_info (struct sfm_reader *,
308 const struct sfm_extension_record *,
309 struct sfm_read_info *);
310 static void parse_mrsets (struct sfm_reader *,
311 const struct sfm_extension_record *,
312 struct dictionary *);
313 static void parse_long_var_name_map (struct sfm_reader *,
314 const struct sfm_extension_record *,
315 struct dictionary *);
316 static void parse_long_string_map (struct sfm_reader *,
317 const struct sfm_extension_record *,
318 struct dictionary *);
319 static void parse_value_labels (struct sfm_reader *, struct dictionary *,
320 const struct sfm_var_record *,
322 const struct sfm_value_label_record *);
323 static void parse_data_file_attributes (struct sfm_reader *,
324 const struct sfm_extension_record *,
325 struct dictionary *);
326 static void parse_variable_attributes (struct sfm_reader *,
327 const struct sfm_extension_record *,
328 struct dictionary *);
329 static void assign_variable_roles (struct sfm_reader *, struct dictionary *);
330 static void parse_long_string_value_labels (struct sfm_reader *,
331 const struct sfm_extension_record *,
332 struct dictionary *);
333 static void parse_long_string_missing_values (struct sfm_reader *,
334 const struct sfm_extension_record *,
335 struct dictionary *);
337 /* Frees the strings inside INFO. */
339 sfm_read_info_destroy (struct sfm_read_info *info)
343 free (info->creation_date);
344 free (info->creation_time);
345 free (info->product);
346 free (info->product_ext);
350 /* Opens the system file designated by file handle FH for reading. Reads the
351 system file's dictionary into *DICT.
353 Ordinarily the reader attempts to automatically detect the character
354 encoding based on the file's contents. This isn't always possible,
355 especially for files written by old versions of SPSS or PSPP, so specifying
356 a nonnull ENCODING overrides the choice of character encoding.
358 If INFO is non-null, then it receives additional info about the system file,
359 which the caller must eventually free with sfm_read_info_destroy() when it
360 is no longer needed. */
362 sfm_open_reader (struct file_handle *fh, const char *volatile encoding,
363 struct dictionary **dictp, struct sfm_read_info *infop)
365 struct sfm_reader *volatile r = NULL;
366 struct sfm_read_info *volatile info;
368 struct sfm_header_record header;
370 struct sfm_var_record *vars;
371 size_t n_vars, allocated_vars;
373 struct sfm_value_label_record *labels;
374 size_t n_labels, allocated_labels;
376 struct sfm_document_record *document;
378 struct sfm_extension_record *extensions[32];
380 struct dictionary *volatile dict = NULL;
383 /* Create and initialize reader. */
384 r = pool_create_container (struct sfm_reader, pool);
390 r->opcode_idx = sizeof r->opcodes;
391 r->corruption_warning = false;
392 r->zin_buf = r->zout_buf = NULL;
394 info = infop ? infop : xmalloc (sizeof *info);
395 memset (info, 0, sizeof *info);
397 /* TRANSLATORS: this fragment will be interpolated into
398 messages in fh_lock() that identify types of files. */
399 r->lock = fh_lock (fh, FH_REF_FILE, N_("system file"), FH_ACC_READ, false);
403 r->file = fn_open (fh_get_file_name (fh), "rb");
406 msg (ME, _("Error opening `%s' for reading as a system file: %s."),
407 fh_get_file_name (r->fh), strerror (errno));
411 if (setjmp (r->bail_out))
415 read_header (r, info, &header);
418 n_vars = allocated_vars = 0;
421 n_labels = allocated_labels = 0;
425 memset (extensions, 0, sizeof extensions);
435 read_int (r); /* Skip filler. */
442 if (n_vars >= allocated_vars)
443 vars = pool_2nrealloc (r->pool, vars, &allocated_vars,
445 read_variable_record (r, &vars[n_vars++]);
449 if (n_labels >= allocated_labels)
450 labels = pool_2nrealloc (r->pool, labels, &allocated_labels,
452 read_value_label_record (r, &labels[n_labels++], n_vars);
456 /* A Type 4 record is always immediately after a type 3 record,
457 so the code for type 3 records reads the type 4 record too. */
458 sys_error (r, r->pos, _("Misplaced type 4 record."));
461 if (document != NULL)
462 sys_error (r, r->pos, _("Duplicate type 6 (document) record."));
463 document = read_document_record (r);
467 subtype = read_int (r);
468 if (subtype < 0 || subtype >= sizeof extensions / sizeof *extensions)
471 _("Unrecognized record type 7, subtype %d. Please "
472 "send a copy of this file, and the syntax which "
473 "created it to %s."),
474 subtype, PACKAGE_BUGREPORT);
475 skip_extension_record (r, subtype);
477 else if (extensions[subtype] != NULL)
480 _("Record type 7, subtype %d found here has the same "
481 "type as the record found near offset 0x%llx. "
482 "Please send a copy of this file, and the syntax "
483 "which created it to %s."),
484 subtype, (long long int) extensions[subtype]->pos,
486 skip_extension_record (r, subtype);
489 extensions[subtype] = read_extension_record (r, subtype);
493 sys_error (r, r->pos, _("Unrecognized record type %d."), type);
498 if (r->compression == SFM_COMP_ZLIB)
501 /* Now actually parse what we read.
503 First, figure out the correct character encoding, because this determines
504 how the rest of the header data is to be interpreted. */
505 dict = dict_create (encoding
507 : choose_encoding (r, &header, extensions[EXT_INTEGER],
508 extensions[EXT_ENCODING]));
509 r->encoding = dict_get_encoding (dict);
511 /* These records don't use variables at all. */
512 if (document != NULL)
513 parse_document (dict, document);
515 if (extensions[EXT_INTEGER] != NULL)
516 parse_machine_integer_info (r, extensions[EXT_INTEGER], info);
518 if (extensions[EXT_FLOAT] != NULL)
519 parse_machine_float_info (r, extensions[EXT_FLOAT]);
521 if (extensions[EXT_PRODUCT_INFO] != NULL)
522 parse_extra_product_info (r, extensions[EXT_PRODUCT_INFO], info);
524 if (extensions[EXT_FILE_ATTRS] != NULL)
525 parse_data_file_attributes (r, extensions[EXT_FILE_ATTRS], dict);
527 parse_header (r, &header, info, dict);
529 /* Parse the variable records, the basis of almost everything else. */
530 parse_variable_records (r, dict, vars, n_vars);
532 /* Parse value labels and the weight variable immediately after the variable
533 records. These records use indexes into var_recs[], so we must parse them
534 before those indexes become invalidated by very long string variables. */
535 for (i = 0; i < n_labels; i++)
536 parse_value_labels (r, dict, vars, n_vars, &labels[i]);
537 if (header.weight_idx != 0)
539 struct variable *weight_var;
541 weight_var = lookup_var_by_index (r, 76, vars, n_vars,
543 if (var_is_numeric (weight_var))
544 dict_set_weight (dict, weight_var);
546 sys_error (r, -1, _("Weighting variable must be numeric "
547 "(not string variable `%s')."),
548 var_get_name (weight_var));
551 if (extensions[EXT_DISPLAY] != NULL)
552 parse_display_parameters (r, extensions[EXT_DISPLAY], dict);
554 /* The following records use short names, so they need to be parsed before
555 parse_long_var_name_map() changes short names to long names. */
556 if (extensions[EXT_MRSETS] != NULL)
557 parse_mrsets (r, extensions[EXT_MRSETS], dict);
559 if (extensions[EXT_MRSETS2] != NULL)
560 parse_mrsets (r, extensions[EXT_MRSETS2], dict);
562 if (extensions[EXT_LONG_STRINGS] != NULL)
563 parse_long_string_map (r, extensions[EXT_LONG_STRINGS], dict);
565 /* Now rename variables to their long names. */
566 parse_long_var_name_map (r, extensions[EXT_LONG_NAMES], dict);
568 /* The following records use long names, so they need to follow renaming. */
569 if (extensions[EXT_VAR_ATTRS] != NULL)
571 parse_variable_attributes (r, extensions[EXT_VAR_ATTRS], dict);
573 /* Roles use the $@Role attribute. */
574 assign_variable_roles (r, dict);
577 if (extensions[EXT_LONG_LABELS] != NULL)
578 parse_long_string_value_labels (r, extensions[EXT_LONG_LABELS], dict);
579 if (extensions[EXT_LONG_MISSING] != NULL)
580 parse_long_string_missing_values (r, extensions[EXT_LONG_MISSING], dict);
582 /* Warn if the actual amount of data per case differs from the
583 amount that the header claims. SPSS version 13 gets this
584 wrong when very long strings are involved, so don't warn in
586 if (header.nominal_case_size != -1 && header.nominal_case_size != n_vars
587 && info->version_major != 13)
588 sys_warn (r, -1, _("File header claims %d variable positions but "
589 "%zu were read from file."),
590 header.nominal_case_size, n_vars);
592 /* Create an index of dictionary variable widths for
593 sfm_read_case to use. We cannot use the `struct variable's
594 from the dictionary we created, because the caller owns the
595 dictionary and may destroy or modify its variables. */
596 sfm_dictionary_to_sfm_vars (dict, &r->sfm_vars, &r->sfm_var_cnt);
597 pool_register (r->pool, free, r->sfm_vars);
598 r->proto = caseproto_ref_pool (dict_get_proto (dict), r->pool);
603 sfm_read_info_destroy (info);
607 return casereader_create_sequential
609 r->case_cnt == -1 ? CASENUMBER_MAX: r->case_cnt,
610 &sys_file_casereader_class, r);
615 sfm_read_info_destroy (info);
625 /* Closes a system file after we're done with it.
626 Returns true if an I/O error has occurred on READER, false
629 close_reader (struct sfm_reader *r)
638 if (fn_close (fh_get_file_name (r->fh), r->file) == EOF)
640 msg (ME, _("Error closing system file `%s': %s."),
641 fh_get_file_name (r->fh), strerror (errno));
651 pool_destroy (r->pool);
656 /* Destroys READER. */
658 sys_file_casereader_destroy (struct casereader *reader UNUSED, void *r_)
660 struct sfm_reader *r = r_;
664 /* Returns true if FILE is an SPSS system file,
667 sfm_detect (FILE *file)
671 if (fread (magic, 4, 1, file) != 1)
675 return (!strcmp (ASCII_MAGIC, magic)
676 || !strcmp (ASCII_ZMAGIC, magic)
677 || !strcmp (EBCDIC_MAGIC, magic));
680 /* Reads the global header of the system file. Initializes *HEADER and *INFO,
681 except for the string fields in *INFO, which parse_header() will initialize
682 later once the file's encoding is known. */
684 read_header (struct sfm_reader *r, struct sfm_read_info *info,
685 struct sfm_header_record *header)
687 uint8_t raw_layout_code[4];
692 read_string (r, header->magic, sizeof header->magic);
693 read_string (r, header->eye_catcher, sizeof header->eye_catcher);
695 if (!strcmp (ASCII_MAGIC, header->magic)
696 || !strcmp (EBCDIC_MAGIC, header->magic))
698 else if (!strcmp (ASCII_ZMAGIC, header->magic))
701 sys_error (r, 0, _("This is not an SPSS system file."));
703 /* Identify integer format. */
704 read_bytes (r, raw_layout_code, sizeof raw_layout_code);
705 if ((!integer_identify (2, raw_layout_code, sizeof raw_layout_code,
707 && !integer_identify (3, raw_layout_code, sizeof raw_layout_code,
709 || (r->integer_format != INTEGER_MSB_FIRST
710 && r->integer_format != INTEGER_LSB_FIRST))
711 sys_error (r, 64, _("This is not an SPSS system file."));
713 header->nominal_case_size = read_int (r);
714 if (header->nominal_case_size < 0
715 || header->nominal_case_size > INT_MAX / 16)
716 header->nominal_case_size = -1;
718 compressed = read_int (r);
722 r->compression = SFM_COMP_NONE;
723 else if (compressed == 1)
724 r->compression = SFM_COMP_SIMPLE;
725 else if (compressed != 0)
726 sys_error (r, 0, "System file header has invalid compression "
727 "value %d.", compressed);
732 r->compression = SFM_COMP_ZLIB;
734 sys_error (r, 0, "ZLIB-compressed system file header has invalid "
735 "compression value %d.", compressed);
738 header->weight_idx = read_int (r);
740 r->case_cnt = read_int (r);
741 if ( r->case_cnt > INT_MAX / 2)
744 /* Identify floating-point format and obtain compression bias. */
745 read_bytes (r, raw_bias, sizeof raw_bias);
746 if (float_identify (100.0, raw_bias, sizeof raw_bias, &r->float_format) == 0)
748 uint8_t zero_bias[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
750 if (memcmp (raw_bias, zero_bias, 8))
751 sys_warn (r, r->pos - 8,
752 _("Compression bias is not the usual "
753 "value of 100, or system file uses unrecognized "
754 "floating-point format."));
757 /* Some software is known to write all-zeros to this
758 field. Such software also writes floating-point
759 numbers in the format that we expect by default
760 (it seems that all software most likely does, in
761 reality), so don't warn in this case. */
764 if (r->integer_format == INTEGER_MSB_FIRST)
765 r->float_format = FLOAT_IEEE_DOUBLE_BE;
767 r->float_format = FLOAT_IEEE_DOUBLE_LE;
769 float_convert (r->float_format, raw_bias, FLOAT_NATIVE_DOUBLE, &r->bias);
771 read_string (r, header->creation_date, sizeof header->creation_date);
772 read_string (r, header->creation_time, sizeof header->creation_time);
773 read_string (r, header->file_label, sizeof header->file_label);
776 info->integer_format = r->integer_format;
777 info->float_format = r->float_format;
778 info->compression = r->compression;
779 info->case_cnt = r->case_cnt;
782 /* Reads a variable (type 2) record from R into RECORD. */
784 read_variable_record (struct sfm_reader *r, struct sfm_var_record *record)
786 int has_variable_label;
788 memset (record, 0, sizeof *record);
790 record->pos = r->pos;
791 record->width = read_int (r);
792 has_variable_label = read_int (r);
793 record->missing_value_code = read_int (r);
794 record->print_format = read_int (r);
795 record->write_format = read_int (r);
796 read_bytes (r, record->name, sizeof record->name);
798 if (has_variable_label == 1)
800 enum { MAX_LABEL_LEN = 255 };
801 size_t len, read_len;
805 /* Read up to MAX_LABEL_LEN bytes of label. */
806 read_len = MIN (MAX_LABEL_LEN, len);
807 record->label = pool_malloc (r->pool, read_len + 1);
808 read_string (r, record->label, read_len + 1);
810 /* Skip unread label bytes. */
811 skip_bytes (r, len - read_len);
813 /* Skip label padding up to multiple of 4 bytes. */
814 skip_bytes (r, ROUND_UP (len, 4) - len);
816 else if (has_variable_label != 0)
817 sys_error (r, record->pos,
818 _("Variable label indicator field is not 0 or 1."));
820 /* Set missing values. */
821 if (record->missing_value_code != 0)
823 int code = record->missing_value_code;
824 if (record->width == 0)
826 if (code < -3 || code > 3 || code == -1)
827 sys_error (r, record->pos,
828 _("Numeric missing value indicator field is not "
829 "-3, -2, 0, 1, 2, or 3."));
833 if (code < 1 || code > 3)
834 sys_error (r, record->pos,
835 _("String missing value indicator field is not "
839 read_bytes (r, record->missing, 8 * abs (code));
843 /* Reads value labels from R into RECORD. */
845 read_value_label_record (struct sfm_reader *r,
846 struct sfm_value_label_record *record,
851 /* Read type 3 record. */
852 record->pos = r->pos;
853 record->n_labels = read_int (r);
854 if (record->n_labels > SIZE_MAX / sizeof *record->labels)
855 sys_error (r, r->pos - 4, _("Invalid number of labels %zu."),
857 record->labels = pool_nmalloc (r->pool, record->n_labels,
858 sizeof *record->labels);
859 for (i = 0; i < record->n_labels; i++)
861 struct sfm_value_label *label = &record->labels[i];
862 unsigned char label_len;
865 read_bytes (r, label->value, sizeof label->value);
867 /* Read label length. */
868 read_bytes (r, &label_len, sizeof label_len);
869 padded_len = ROUND_UP (label_len + 1, 8);
871 /* Read label, padding. */
872 label->label = pool_malloc (r->pool, padded_len + 1);
873 read_bytes (r, label->label, padded_len - 1);
874 label->label[label_len] = '\0';
877 /* Read record type of type 4 record. */
878 if (read_int (r) != 4)
879 sys_error (r, r->pos - 4,
880 _("Variable index record (type 4) does not immediately "
881 "follow value label record (type 3) as it should."));
883 /* Read number of variables associated with value label from type 4
885 record->n_vars = read_int (r);
886 if (record->n_vars < 1 || record->n_vars > n_vars)
887 sys_error (r, r->pos - 4,
888 _("Number of variables associated with a value label (%zu) "
889 "is not between 1 and the number of variables (%zu)."),
890 record->n_vars, n_vars);
891 record->vars = pool_nmalloc (r->pool, record->n_vars, sizeof *record->vars);
892 for (i = 0; i < record->n_vars; i++)
893 record->vars[i] = read_int (r);
896 /* Reads a document record from R and returns it. */
897 static struct sfm_document_record *
898 read_document_record (struct sfm_reader *r)
900 struct sfm_document_record *record;
903 record = pool_malloc (r->pool, sizeof *record);
904 record->pos = r->pos;
906 n_lines = read_int (r);
907 if (n_lines <= 0 || n_lines >= INT_MAX / DOC_LINE_LENGTH)
908 sys_error (r, record->pos,
909 _("Number of document lines (%d) "
910 "must be greater than 0 and less than %d."),
911 n_lines, INT_MAX / DOC_LINE_LENGTH);
913 record->n_lines = n_lines;
914 record->documents = pool_malloc (r->pool, DOC_LINE_LENGTH * n_lines);
915 read_bytes (r, record->documents, DOC_LINE_LENGTH * n_lines);
921 read_extension_record_header (struct sfm_reader *r, int subtype,
922 struct sfm_extension_record *record)
924 record->subtype = subtype;
925 record->pos = r->pos;
926 record->size = read_int (r);
927 record->count = read_int (r);
929 /* Check that SIZE * COUNT + 1 doesn't overflow. Adding 1
930 allows an extra byte for a null terminator, used by some
931 extension processing routines. */
932 if (record->size != 0
933 && size_overflow_p (xsum (1, xtimes (record->count, record->size))))
934 sys_error (r, record->pos, "Record type 7 subtype %d too large.", subtype);
937 /* Reads an extension record from R into RECORD. */
938 static struct sfm_extension_record *
939 read_extension_record (struct sfm_reader *r, int subtype)
941 struct extension_record_type
948 static const struct extension_record_type types[] =
950 /* Implemented record types. */
951 { EXT_INTEGER, 4, 8 },
953 { EXT_MRSETS, 1, 0 },
954 { EXT_PRODUCT_INFO, 1, 0 },
955 { EXT_DISPLAY, 4, 0 },
956 { EXT_LONG_NAMES, 1, 0 },
957 { EXT_LONG_STRINGS, 1, 0 },
958 { EXT_NCASES, 8, 2 },
959 { EXT_FILE_ATTRS, 1, 0 },
960 { EXT_VAR_ATTRS, 1, 0 },
961 { EXT_MRSETS2, 1, 0 },
962 { EXT_ENCODING, 1, 0 },
963 { EXT_LONG_LABELS, 1, 0 },
964 { EXT_LONG_MISSING, 1, 0 },
966 /* Ignored record types. */
967 { EXT_VAR_SETS, 0, 0 },
969 { EXT_DATA_ENTRY, 0, 0 },
970 { EXT_DATAVIEW, 0, 0 },
973 const struct extension_record_type *type;
974 struct sfm_extension_record *record;
977 record = pool_malloc (r->pool, sizeof *record);
978 read_extension_record_header (r, subtype, record);
979 n_bytes = record->count * record->size;
981 for (type = types; type < &types[sizeof types / sizeof *types]; type++)
982 if (subtype == type->subtype)
984 if (type->size > 0 && record->size != type->size)
985 sys_warn (r, record->pos,
986 _("Record type 7, subtype %d has bad size %zu "
987 "(expected %d)."), subtype, record->size, type->size);
988 else if (type->count > 0 && record->count != type->count)
989 sys_warn (r, record->pos,
990 _("Record type 7, subtype %d has bad count %zu "
991 "(expected %d)."), subtype, record->count, type->count);
992 else if (type->count == 0 && type->size == 0)
994 /* Ignore this record. */
998 char *data = pool_malloc (r->pool, n_bytes + 1);
999 data[n_bytes] = '\0';
1001 record->data = data;
1002 read_bytes (r, record->data, n_bytes);
1009 sys_warn (r, record->pos,
1010 _("Unrecognized record type 7, subtype %d. Please send a "
1011 "copy of this file, and the syntax which created it to %s."),
1012 subtype, PACKAGE_BUGREPORT);
1015 skip_bytes (r, n_bytes);
1020 skip_extension_record (struct sfm_reader *r, int subtype)
1022 struct sfm_extension_record record;
1024 read_extension_record_header (r, subtype, &record);
1025 skip_bytes (r, record.count * record.size);
1029 parse_header (struct sfm_reader *r, const struct sfm_header_record *header,
1030 struct sfm_read_info *info, struct dictionary *dict)
1032 const char *dict_encoding = dict_get_encoding (dict);
1033 struct substring product;
1034 struct substring label;
1037 /* Convert file label to UTF-8 and put it into DICT. */
1038 label = recode_substring_pool ("UTF-8", dict_encoding,
1039 ss_cstr (header->file_label), r->pool);
1040 ss_trim (&label, ss_cstr (" "));
1041 label.string[label.length] = '\0';
1042 fixed_label = fix_line_ends (label.string);
1043 dict_set_label (dict, fixed_label);
1046 /* Put creation date and time in UTF-8 into INFO. */
1047 info->creation_date = recode_string ("UTF-8", dict_encoding,
1048 header->creation_date, -1);
1049 info->creation_time = recode_string ("UTF-8", dict_encoding,
1050 header->creation_time, -1);
1052 /* Put product name into INFO, dropping eye-catcher string if present. */
1053 product = recode_substring_pool ("UTF-8", dict_encoding,
1054 ss_cstr (header->eye_catcher), r->pool);
1055 ss_match_string (&product, ss_cstr ("@(#) SPSS DATA FILE"));
1056 ss_trim (&product, ss_cstr (" "));
1057 info->product = ss_xstrdup (product);
1060 /* Reads a variable (type 2) record from R and adds the
1061 corresponding variable to DICT.
1062 Also skips past additional variable records for long string
1065 parse_variable_records (struct sfm_reader *r, struct dictionary *dict,
1066 struct sfm_var_record *var_recs, size_t n_var_recs)
1068 const char *dict_encoding = dict_get_encoding (dict);
1069 struct sfm_var_record *rec;
1072 for (rec = var_recs; rec < &var_recs[n_var_recs]; )
1074 struct variable *var;
1079 name = recode_string_pool ("UTF-8", dict_encoding,
1080 rec->name, 8, r->pool);
1081 name[strcspn (name, " ")] = '\0';
1083 if (!dict_id_is_valid (dict, name, false)
1084 || name[0] == '$' || name[0] == '#')
1085 sys_error (r, rec->pos, _("Invalid variable name `%s'."), name);
1087 if (rec->width < 0 || rec->width > 255)
1088 sys_error (r, rec->pos,
1089 _("Bad width %d for variable %s."), rec->width, name);
1091 var = rec->var = dict_create_var (dict, name, rec->width);
1093 sys_error (r, rec->pos, _("Duplicate variable name `%s'."), name);
1095 /* Set the short name the same as the long name. */
1096 var_set_short_name (var, 0, name);
1098 /* Get variable label, if any. */
1103 utf8_label = recode_string_pool ("UTF-8", dict_encoding,
1104 rec->label, -1, r->pool);
1105 var_set_label (var, utf8_label, false);
1108 /* Set missing values. */
1109 if (rec->missing_value_code != 0)
1111 int width = var_get_width (var);
1112 struct missing_values mv;
1114 mv_init_pool (r->pool, &mv, width);
1115 if (var_is_numeric (var))
1117 bool has_range = rec->missing_value_code < 0;
1118 int n_discrete = (has_range
1119 ? rec->missing_value_code == -3
1120 : rec->missing_value_code);
1125 double low = parse_float (r, rec->missing, 0);
1126 double high = parse_float (r, rec->missing, 8);
1128 /* Deal with SPSS 21 change in representation. */
1132 mv_add_range (&mv, low, high);
1136 for (i = 0; i < n_discrete; i++)
1138 mv_add_num (&mv, parse_float (r, rec->missing, ofs));
1146 value_init_pool (r->pool, &value, width);
1147 value_set_missing (&value, width);
1148 for (i = 0; i < rec->missing_value_code; i++)
1149 mv_add_str (&mv, rec->missing + 8 * i, MIN (width, 8));
1151 var_set_missing_values (var, &mv);
1155 parse_format_spec (r, rec->pos + 12, rec->print_format,
1156 PRINT_FORMAT, var, &n_warnings);
1157 parse_format_spec (r, rec->pos + 16, rec->write_format,
1158 WRITE_FORMAT, var, &n_warnings);
1160 /* Account for values.
1161 Skip long string continuation records, if any. */
1162 n_values = rec->width == 0 ? 1 : DIV_RND_UP (rec->width, 8);
1163 for (i = 1; i < n_values; i++)
1164 if (i + (rec - var_recs) >= n_var_recs || rec[i].width != -1)
1165 sys_error (r, rec->pos, _("Missing string continuation record."));
1170 /* Translates the format spec from sysfile format to internal
1173 parse_format_spec (struct sfm_reader *r, off_t pos, unsigned int format,
1174 enum which_format which, struct variable *v,
1177 const int max_warnings = 8;
1178 uint8_t raw_type = format >> 16;
1179 uint8_t w = format >> 8;
1188 ok = (fmt_from_io (raw_type, &f.type)
1189 && fmt_check_output (&f)
1190 && fmt_check_width_compat (&f, var_get_width (v)));
1195 if (which == PRINT_FORMAT)
1196 var_set_print_format (v, &f);
1198 var_set_write_format (v, &f);
1200 else if (format == 0)
1202 /* Actually observed in the wild. No point in warning about it. */
1204 else if (++*n_warnings <= max_warnings)
1206 if (which == PRINT_FORMAT)
1207 sys_warn (r, pos, _("Variable %s with width %d has invalid print "
1209 var_get_name (v), var_get_width (v), format);
1211 sys_warn (r, pos, _("Variable %s with width %d has invalid write "
1213 var_get_name (v), var_get_width (v), format);
1215 if (*n_warnings == max_warnings)
1216 sys_warn (r, -1, _("Suppressing further invalid format warnings."));
1221 parse_document (struct dictionary *dict, struct sfm_document_record *record)
1225 for (p = record->documents;
1226 p < record->documents + DOC_LINE_LENGTH * record->n_lines;
1227 p += DOC_LINE_LENGTH)
1229 struct substring line;
1231 line = recode_substring_pool ("UTF-8", dict_get_encoding (dict),
1232 ss_buffer (p, DOC_LINE_LENGTH), NULL);
1233 ss_rtrim (&line, ss_cstr (" "));
1234 line.string[line.length] = '\0';
1236 dict_add_document_line (dict, line.string, false);
1242 /* Parses record type 7, subtype 3. */
1244 parse_machine_integer_info (struct sfm_reader *r,
1245 const struct sfm_extension_record *record,
1246 struct sfm_read_info *info)
1248 int float_representation, expected_float_format;
1249 int integer_representation, expected_integer_format;
1251 /* Save version info. */
1252 info->version_major = parse_int (r, record->data, 0);
1253 info->version_minor = parse_int (r, record->data, 4);
1254 info->version_revision = parse_int (r, record->data, 8);
1256 /* Check floating point format. */
1257 float_representation = parse_int (r, record->data, 16);
1258 if (r->float_format == FLOAT_IEEE_DOUBLE_BE
1259 || r->float_format == FLOAT_IEEE_DOUBLE_LE)
1260 expected_float_format = 1;
1261 else if (r->float_format == FLOAT_Z_LONG)
1262 expected_float_format = 2;
1263 else if (r->float_format == FLOAT_VAX_G || r->float_format == FLOAT_VAX_D)
1264 expected_float_format = 3;
1267 if (float_representation != expected_float_format)
1268 sys_error (r, record->pos, _("Floating-point representation indicated by "
1269 "system file (%d) differs from expected (%d)."),
1270 float_representation, expected_float_format);
1272 /* Check integer format. */
1273 integer_representation = parse_int (r, record->data, 24);
1274 if (r->integer_format == INTEGER_MSB_FIRST)
1275 expected_integer_format = 1;
1276 else if (r->integer_format == INTEGER_LSB_FIRST)
1277 expected_integer_format = 2;
1280 if (integer_representation != expected_integer_format)
1281 sys_warn (r, record->pos,
1282 _("Integer format indicated by system file (%d) "
1283 "differs from expected (%d)."),
1284 integer_representation, expected_integer_format);
1289 choose_encoding (struct sfm_reader *r,
1290 const struct sfm_header_record *header,
1291 const struct sfm_extension_record *ext_integer,
1292 const struct sfm_extension_record *ext_encoding)
1294 /* The EXT_ENCODING record is a more reliable way to determine dictionary
1297 return ext_encoding->data;
1299 /* But EXT_INTEGER is better than nothing as a fallback. */
1302 int codepage = parse_int (r, ext_integer->data, 7 * 4);
1303 const char *encoding;
1312 /* These ostensibly mean "7-bit ASCII" and "8-bit ASCII"[sic]
1313 respectively. However, there are known to be many files in the wild
1314 with character code 2, yet have data which are clearly not ASCII.
1315 Therefore we ignore these values. */
1322 encoding = sys_get_encoding_from_codepage (codepage);
1323 if (encoding != NULL)
1329 /* If the file magic number is EBCDIC then its character data is too. */
1330 if (!strcmp (header->magic, EBCDIC_MAGIC))
1333 return locale_charset ();
1336 /* Parses record type 7, subtype 4. */
1338 parse_machine_float_info (struct sfm_reader *r,
1339 const struct sfm_extension_record *record)
1341 double sysmis = parse_float (r, record->data, 0);
1342 double highest = parse_float (r, record->data, 8);
1343 double lowest = parse_float (r, record->data, 16);
1345 if (sysmis != SYSMIS)
1346 sys_warn (r, record->pos,
1347 _("File specifies unexpected value %g (%a) as %s, "
1348 "instead of %g (%a)."),
1349 sysmis, sysmis, "SYSMIS", SYSMIS, SYSMIS);
1351 if (highest != HIGHEST)
1352 sys_warn (r, record->pos,
1353 _("File specifies unexpected value %g (%a) as %s, "
1354 "instead of %g (%a)."),
1355 highest, highest, "HIGHEST", HIGHEST, HIGHEST);
1357 /* SPSS before version 21 used a unique value just bigger than SYSMIS as
1358 LOWEST. SPSS 21 uses SYSMIS for LOWEST, which is OK because LOWEST only
1359 appears in a context (missing values) where SYSMIS cannot. */
1360 if (lowest != LOWEST && lowest != SYSMIS)
1361 sys_warn (r, record->pos,
1362 _("File specifies unexpected value %g (%a) as %s, "
1363 "instead of %g (%a) or %g (%a)."),
1364 lowest, lowest, "LOWEST", LOWEST, LOWEST, SYSMIS, SYSMIS);
1367 /* Parses record type 7, subtype 10. */
1369 parse_extra_product_info (struct sfm_reader *r,
1370 const struct sfm_extension_record *record,
1371 struct sfm_read_info *info)
1373 struct text_record *text;
1375 text = open_text_record (r, record, true);
1376 info->product_ext = fix_line_ends (text_get_all (text));
1377 close_text_record (r, text);
1380 /* Parses record type 7, subtype 7 or 19. */
1382 parse_mrsets (struct sfm_reader *r, const struct sfm_extension_record *record,
1383 struct dictionary *dict)
1385 struct text_record *text;
1386 struct mrset *mrset;
1388 text = open_text_record (r, record, false);
1391 const char *counted = NULL;
1394 struct stringi_set var_names;
1395 size_t allocated_vars;
1399 mrset = xzalloc (sizeof *mrset);
1401 name = text_get_token (text, ss_cstr ("="), NULL);
1404 mrset->name = recode_string ("UTF-8", r->encoding, name, -1);
1406 if (mrset->name[0] != '$')
1408 sys_warn (r, record->pos,
1409 _("`%s' does not begin with `$' at offset %zu "
1410 "in MRSETS record."), mrset->name, text_pos (text));
1414 if (text_match (text, 'C'))
1416 mrset->type = MRSET_MC;
1417 if (!text_match (text, ' '))
1419 sys_warn (r, record->pos,
1420 _("Missing space following `%c' at offset %zu "
1421 "in MRSETS record."), 'C', text_pos (text));
1425 else if (text_match (text, 'D'))
1427 mrset->type = MRSET_MD;
1428 mrset->cat_source = MRSET_VARLABELS;
1430 else if (text_match (text, 'E'))
1434 mrset->type = MRSET_MD;
1435 mrset->cat_source = MRSET_COUNTEDVALUES;
1436 if (!text_match (text, ' '))
1438 sys_warn (r, record->pos,
1439 _("Missing space following `%c' at offset %zu "
1440 "in MRSETS record."), 'E', text_pos (text));
1444 number = text_get_token (text, ss_cstr (" "), NULL);
1445 if (!strcmp (number, "11"))
1446 mrset->label_from_var_label = true;
1447 else if (strcmp (number, "1"))
1448 sys_warn (r, record->pos,
1449 _("Unexpected label source value `%s' following `E' "
1450 "at offset %zu in MRSETS record."),
1451 number, text_pos (text));
1455 sys_warn (r, record->pos,
1456 _("Missing `C', `D', or `E' at offset %zu "
1457 "in MRSETS record."),
1462 if (mrset->type == MRSET_MD)
1464 counted = text_parse_counted_string (r, text);
1465 if (counted == NULL)
1469 label = text_parse_counted_string (r, text);
1472 if (label[0] != '\0')
1473 mrset->label = recode_string ("UTF-8", r->encoding, label, -1);
1475 stringi_set_init (&var_names);
1480 const char *raw_var_name;
1481 struct variable *var;
1484 raw_var_name = text_get_token (text, ss_cstr (" \n"), &delimiter);
1485 if (raw_var_name == NULL)
1487 sys_warn (r, record->pos,
1488 _("Missing new-line parsing variable names "
1489 "at offset %zu in MRSETS record."),
1493 var_name = recode_string ("UTF-8", r->encoding, raw_var_name, -1);
1495 var = dict_lookup_var (dict, var_name);
1501 if (!stringi_set_insert (&var_names, var_name))
1503 sys_warn (r, record->pos,
1504 _("Duplicate variable name %s "
1505 "at offset %zu in MRSETS record."),
1506 var_name, text_pos (text));
1512 if (mrset->label == NULL && mrset->label_from_var_label
1513 && var_has_label (var))
1514 mrset->label = xstrdup (var_get_label (var));
1517 && var_get_type (var) != var_get_type (mrset->vars[0]))
1519 sys_warn (r, record->pos,
1520 _("MRSET %s contains both string and "
1521 "numeric variables."), name);
1524 width = MIN (width, var_get_width (var));
1526 if (mrset->n_vars >= allocated_vars)
1527 mrset->vars = x2nrealloc (mrset->vars, &allocated_vars,
1528 sizeof *mrset->vars);
1529 mrset->vars[mrset->n_vars++] = var;
1531 while (delimiter != '\n');
1533 if (mrset->n_vars < 2)
1535 sys_warn (r, record->pos,
1536 _("MRSET %s has only %zu variables."), mrset->name,
1538 mrset_destroy (mrset);
1539 stringi_set_destroy (&var_names);
1543 if (mrset->type == MRSET_MD)
1545 mrset->width = width;
1546 value_init (&mrset->counted, width);
1548 mrset->counted.f = c_strtod (counted, NULL);
1550 value_copy_str_rpad (&mrset->counted, width,
1551 (const uint8_t *) counted, ' ');
1554 dict_add_mrset (dict, mrset);
1556 stringi_set_destroy (&var_names);
1558 mrset_destroy (mrset);
1559 close_text_record (r, text);
1562 /* Read record type 7, subtype 11, which specifies how variables
1563 should be displayed in GUI environments. */
1565 parse_display_parameters (struct sfm_reader *r,
1566 const struct sfm_extension_record *record,
1567 struct dictionary *dict)
1569 bool includes_width;
1570 bool warned = false;
1575 n_vars = dict_get_var_cnt (dict);
1576 if (record->count == 3 * n_vars)
1577 includes_width = true;
1578 else if (record->count == 2 * n_vars)
1579 includes_width = false;
1582 sys_warn (r, record->pos,
1583 _("Extension 11 has bad count %zu (for %zu variables)."),
1584 record->count, n_vars);
1589 for (i = 0; i < n_vars; ++i)
1591 struct variable *v = dict_get_var (dict, i);
1592 int measure, width, align;
1594 measure = parse_int (r, record->data, ofs);
1599 width = parse_int (r, record->data, ofs);
1605 align = parse_int (r, record->data, ofs);
1608 /* SPSS sometimes seems to set variables' measure to zero. */
1612 if (measure < 1 || measure > 3 || align < 0 || align > 2)
1615 sys_warn (r, record->pos,
1616 _("Invalid variable display parameters for variable "
1617 "%zu (%s). Default parameters substituted."),
1618 i, var_get_name (v));
1623 var_set_measure (v, (measure == 1 ? MEASURE_NOMINAL
1624 : measure == 2 ? MEASURE_ORDINAL
1626 var_set_alignment (v, (align == 0 ? ALIGN_LEFT
1627 : align == 1 ? ALIGN_RIGHT
1630 /* Older versions (SPSS 9.0) sometimes set the display
1631 width to zero. This causes confusion in the GUI, so
1632 only set the width if it is nonzero. */
1634 var_set_display_width (v, width);
1639 rename_var_and_save_short_names (struct dictionary *dict, struct variable *var,
1640 const char *new_name)
1642 size_t n_short_names;
1646 /* Renaming a variable may clear its short names, but we
1647 want to retain them, so we save them and re-set them
1649 n_short_names = var_get_short_name_cnt (var);
1650 short_names = xnmalloc (n_short_names, sizeof *short_names);
1651 for (i = 0; i < n_short_names; i++)
1653 const char *s = var_get_short_name (var, i);
1654 short_names[i] = s != NULL ? xstrdup (s) : NULL;
1657 /* Set long name. */
1658 dict_rename_var (dict, var, new_name);
1660 /* Restore short names. */
1661 for (i = 0; i < n_short_names; i++)
1663 var_set_short_name (var, i, short_names[i]);
1664 free (short_names[i]);
1669 /* Parses record type 7, subtype 13, which gives the long name that corresponds
1670 to each short name. Modifies variable names in DICT accordingly. */
1672 parse_long_var_name_map (struct sfm_reader *r,
1673 const struct sfm_extension_record *record,
1674 struct dictionary *dict)
1676 struct text_record *text;
1677 struct variable *var;
1682 /* There are no long variable names. Use the short variable names,
1683 converted to lowercase, as the long variable names. */
1686 for (i = 0; i < dict_get_var_cnt (dict); i++)
1688 struct variable *var = dict_get_var (dict, i);
1691 new_name = utf8_to_lower (var_get_name (var));
1692 rename_var_and_save_short_names (dict, var, new_name);
1699 /* Rename each of the variables, one by one. (In a correctly constructed
1700 system file, this cannot create any intermediate duplicate variable names,
1701 because all of the new variable names are longer than any of the old
1702 variable names and thus there cannot be any overlaps.) */
1703 text = open_text_record (r, record, true);
1704 while (read_variable_to_value_pair (r, dict, text, &var, &long_name))
1706 /* Validate long name. */
1707 if (!dict_id_is_valid (dict, long_name, false))
1709 sys_warn (r, record->pos,
1710 _("Long variable mapping from %s to invalid "
1711 "variable name `%s'."),
1712 var_get_name (var), long_name);
1716 /* Identify any duplicates. */
1717 if (utf8_strcasecmp (var_get_short_name (var, 0), long_name)
1718 && dict_lookup_var (dict, long_name) != NULL)
1720 sys_warn (r, record->pos,
1721 _("Duplicate long variable name `%s'."), long_name);
1725 rename_var_and_save_short_names (dict, var, long_name);
1727 close_text_record (r, text);
1730 /* Reads record type 7, subtype 14, which gives the real length
1731 of each very long string. Rearranges DICT accordingly. */
1733 parse_long_string_map (struct sfm_reader *r,
1734 const struct sfm_extension_record *record,
1735 struct dictionary *dict)
1737 struct text_record *text;
1738 struct variable *var;
1741 text = open_text_record (r, record, true);
1742 while (read_variable_to_value_pair (r, dict, text, &var, &length_s))
1744 size_t idx = var_get_dict_index (var);
1750 length = strtol (length_s, NULL, 10);
1751 if (length < 1 || length > MAX_STRING)
1753 sys_warn (r, record->pos,
1754 _("%s listed as string of invalid length %s "
1755 "in very long string record."),
1756 var_get_name (var), length_s);
1760 /* Check segments. */
1761 segment_cnt = sfm_width_to_segments (length);
1762 if (segment_cnt == 1)
1764 sys_warn (r, record->pos,
1765 _("%s listed in very long string record with width %s, "
1766 "which requires only one segment."),
1767 var_get_name (var), length_s);
1770 if (idx + segment_cnt > dict_get_var_cnt (dict))
1771 sys_error (r, record->pos,
1772 _("Very long string %s overflows dictionary."),
1773 var_get_name (var));
1775 /* Get the short names from the segments and check their
1777 for (i = 0; i < segment_cnt; i++)
1779 struct variable *seg = dict_get_var (dict, idx + i);
1780 int alloc_width = sfm_segment_alloc_width (length, i);
1781 int width = var_get_width (seg);
1784 var_set_short_name (var, i, var_get_short_name (seg, 0));
1785 if (ROUND_UP (width, 8) != ROUND_UP (alloc_width, 8))
1786 sys_error (r, record->pos,
1787 _("Very long string with width %ld has segment %d "
1788 "of width %d (expected %d)."),
1789 length, i, width, alloc_width);
1791 dict_delete_consecutive_vars (dict, idx + 1, segment_cnt - 1);
1792 var_set_width (var, length);
1794 close_text_record (r, text);
1795 dict_compact_values (dict);
1799 parse_value_labels (struct sfm_reader *r, struct dictionary *dict,
1800 const struct sfm_var_record *var_recs, size_t n_var_recs,
1801 const struct sfm_value_label_record *record)
1803 struct variable **vars;
1807 utf8_labels = pool_nmalloc (r->pool, record->n_labels, sizeof *utf8_labels);
1808 for (i = 0; i < record->n_labels; i++)
1809 utf8_labels[i] = recode_string_pool ("UTF-8", dict_get_encoding (dict),
1810 record->labels[i].label, -1,
1813 vars = pool_nmalloc (r->pool, record->n_vars, sizeof *vars);
1814 for (i = 0; i < record->n_vars; i++)
1815 vars[i] = lookup_var_by_index (r, record->pos,
1816 var_recs, n_var_recs, record->vars[i]);
1818 for (i = 1; i < record->n_vars; i++)
1819 if (var_get_type (vars[i]) != var_get_type (vars[0]))
1820 sys_error (r, record->pos,
1821 _("Variables associated with value label are not all of "
1822 "identical type. Variable %s is %s, but variable "
1824 var_get_name (vars[0]),
1825 var_is_numeric (vars[0]) ? _("numeric") : _("string"),
1826 var_get_name (vars[i]),
1827 var_is_numeric (vars[i]) ? _("numeric") : _("string"));
1829 for (i = 0; i < record->n_vars; i++)
1831 struct variable *var = vars[i];
1835 width = var_get_width (var);
1837 sys_error (r, record->pos,
1838 _("Value labels may not be added to long string "
1839 "variables (e.g. %s) using records types 3 and 4."),
1840 var_get_name (var));
1842 for (j = 0; j < record->n_labels; j++)
1844 struct sfm_value_label *label = &record->labels[j];
1847 value_init (&value, width);
1849 value.f = parse_float (r, label->value, 0);
1851 memcpy (value_str_rw (&value, width), label->value, width);
1853 if (!var_add_value_label (var, &value, utf8_labels[j]))
1855 if (var_is_numeric (var))
1856 sys_warn (r, record->pos,
1857 _("Duplicate value label for %g on %s."),
1858 value.f, var_get_name (var));
1860 sys_warn (r, record->pos,
1861 _("Duplicate value label for `%.*s' on %s."),
1862 width, value_str (&value, width),
1863 var_get_name (var));
1866 value_destroy (&value, width);
1870 pool_free (r->pool, vars);
1871 for (i = 0; i < record->n_labels; i++)
1872 pool_free (r->pool, utf8_labels[i]);
1873 pool_free (r->pool, utf8_labels);
1876 static struct variable *
1877 lookup_var_by_index (struct sfm_reader *r, off_t offset,
1878 const struct sfm_var_record *var_recs, size_t n_var_recs,
1881 const struct sfm_var_record *rec;
1883 if (idx < 1 || idx > n_var_recs)
1885 sys_error (r, offset,
1886 _("Variable index %d not in valid range 1...%zu."),
1891 rec = &var_recs[idx - 1];
1892 if (rec->var == NULL)
1894 sys_error (r, offset,
1895 _("Variable index %d refers to long string continuation."),
1903 /* Parses a set of custom attributes from TEXT into ATTRS.
1904 ATTRS may be a null pointer, in which case the attributes are
1905 read but discarded. */
1907 parse_attributes (struct sfm_reader *r, struct text_record *text,
1908 struct attrset *attrs)
1912 struct attribute *attr;
1916 /* Parse the key. */
1917 key = text_get_token (text, ss_cstr ("("), NULL);
1921 attr = attribute_create (key);
1922 for (index = 1; ; index++)
1924 /* Parse the value. */
1928 value = text_get_token (text, ss_cstr ("\n"), NULL);
1931 text_warn (r, text, _("Error parsing attribute value %s[%d]."),
1936 length = strlen (value);
1937 if (length >= 2 && value[0] == '\'' && value[length - 1] == '\'')
1939 value[length - 1] = '\0';
1940 attribute_add_value (attr, value + 1);
1945 _("Attribute value %s[%d] is not quoted: %s."),
1947 attribute_add_value (attr, value);
1950 /* Was this the last value for this attribute? */
1951 if (text_match (text, ')'))
1955 attrset_add (attrs, attr);
1957 attribute_destroy (attr);
1959 while (!text_match (text, '/'));
1962 /* Reads record type 7, subtype 17, which lists custom
1963 attributes on the data file. */
1965 parse_data_file_attributes (struct sfm_reader *r,
1966 const struct sfm_extension_record *record,
1967 struct dictionary *dict)
1969 struct text_record *text = open_text_record (r, record, true);
1970 parse_attributes (r, text, dict_get_attributes (dict));
1971 close_text_record (r, text);
1974 /* Parses record type 7, subtype 18, which lists custom
1975 attributes on individual variables. */
1977 parse_variable_attributes (struct sfm_reader *r,
1978 const struct sfm_extension_record *record,
1979 struct dictionary *dict)
1981 struct text_record *text;
1982 struct variable *var;
1984 text = open_text_record (r, record, true);
1985 while (text_read_variable_name (r, dict, text, ss_cstr (":"), &var))
1986 parse_attributes (r, text, var != NULL ? var_get_attributes (var) : NULL);
1987 close_text_record (r, text);
1991 assign_variable_roles (struct sfm_reader *r, struct dictionary *dict)
1993 size_t n_warnings = 0;
1996 for (i = 0; i < dict_get_var_cnt (dict); i++)
1998 struct variable *var = dict_get_var (dict, i);
1999 struct attrset *attrs = var_get_attributes (var);
2000 const struct attribute *attr = attrset_lookup (attrs, "$@Role");
2003 int value = atoi (attribute_get_value (attr, 0));
2025 role = ROLE_PARTITION;
2034 if (n_warnings++ == 0)
2035 sys_warn (r, -1, _("Invalid role for variable %s."),
2036 var_get_name (var));
2039 var_set_role (var, role);
2044 sys_warn (r, -1, _("%zu other variables had invalid roles."),
2049 check_overflow (struct sfm_reader *r,
2050 const struct sfm_extension_record *record,
2051 size_t ofs, size_t length)
2053 size_t end = record->size * record->count;
2054 if (length >= end || ofs + length > end)
2055 sys_error (r, record->pos + end,
2056 _("Extension record subtype %d ends unexpectedly."),
2061 parse_long_string_value_labels (struct sfm_reader *r,
2062 const struct sfm_extension_record *record,
2063 struct dictionary *dict)
2065 const char *dict_encoding = dict_get_encoding (dict);
2066 size_t end = record->size * record->count;
2073 struct variable *var;
2078 /* Parse variable name length. */
2079 check_overflow (r, record, ofs, 4);
2080 var_name_len = parse_int (r, record->data, ofs);
2083 /* Parse variable name, width, and number of labels. */
2084 check_overflow (r, record, ofs, var_name_len + 8);
2085 var_name = recode_string_pool ("UTF-8", dict_encoding,
2086 (const char *) record->data + ofs,
2087 var_name_len, r->pool);
2088 width = parse_int (r, record->data, ofs + var_name_len);
2089 n_labels = parse_int (r, record->data, ofs + var_name_len + 4);
2090 ofs += var_name_len + 8;
2092 /* Look up 'var' and validate. */
2093 var = dict_lookup_var (dict, var_name);
2095 sys_warn (r, record->pos + ofs,
2096 _("Ignoring long string value label record for "
2097 "unknown variable %s."), var_name);
2098 else if (var_is_numeric (var))
2100 sys_warn (r, record->pos + ofs,
2101 _("Ignoring long string value label record for "
2102 "numeric variable %s."), var_name);
2105 else if (width != var_get_width (var))
2107 sys_warn (r, record->pos + ofs,
2108 _("Ignoring long string value label record for variable "
2109 "%s because the record's width (%d) does not match the "
2110 "variable's width (%d)."),
2111 var_name, width, var_get_width (var));
2116 value_init_pool (r->pool, &value, width);
2117 for (i = 0; i < n_labels; i++)
2119 size_t value_length, label_length;
2120 bool skip = var == NULL;
2122 /* Parse value length. */
2123 check_overflow (r, record, ofs, 4);
2124 value_length = parse_int (r, record->data, ofs);
2128 check_overflow (r, record, ofs, value_length);
2131 if (value_length == width)
2132 memcpy (value_str_rw (&value, width),
2133 (const uint8_t *) record->data + ofs, width);
2136 sys_warn (r, record->pos + ofs,
2137 _("Ignoring long string value label %zu for "
2138 "variable %s, with width %d, that has bad value "
2140 i, var_get_name (var), width, value_length);
2144 ofs += value_length;
2146 /* Parse label length. */
2147 check_overflow (r, record, ofs, 4);
2148 label_length = parse_int (r, record->data, ofs);
2152 check_overflow (r, record, ofs, label_length);
2157 label = recode_string_pool ("UTF-8", dict_encoding,
2158 (const char *) record->data + ofs,
2159 label_length, r->pool);
2160 if (!var_add_value_label (var, &value, label))
2161 sys_warn (r, record->pos + ofs,
2162 _("Duplicate value label for `%.*s' on %s."),
2163 width, value_str (&value, width),
2164 var_get_name (var));
2165 pool_free (r->pool, label);
2167 ofs += label_length;
2173 parse_long_string_missing_values (struct sfm_reader *r,
2174 const struct sfm_extension_record *record,
2175 struct dictionary *dict)
2177 const char *dict_encoding = dict_get_encoding (dict);
2178 size_t end = record->size * record->count;
2183 struct missing_values mv;
2185 struct variable *var;
2186 int n_missing_values;
2190 /* Parse variable name length. */
2191 check_overflow (r, record, ofs, 4);
2192 var_name_len = parse_int (r, record->data, ofs);
2195 /* Parse variable name. */
2196 check_overflow (r, record, ofs, var_name_len + 1);
2197 var_name = recode_string_pool ("UTF-8", dict_encoding,
2198 (const char *) record->data + ofs,
2199 var_name_len, r->pool);
2200 ofs += var_name_len;
2202 /* Parse number of missing values. */
2203 n_missing_values = ((const uint8_t *) record->data)[ofs];
2204 if (n_missing_values < 1 || n_missing_values > 3)
2205 sys_warn (r, record->pos + ofs,
2206 _("Long string missing values record says variable %s "
2207 "has %d missing values, but only 1 to 3 missing values "
2209 var_name, n_missing_values);
2212 /* Look up 'var' and validate. */
2213 var = dict_lookup_var (dict, var_name);
2215 sys_warn (r, record->pos + ofs,
2216 _("Ignoring long string missing value record for "
2217 "unknown variable %s."), var_name);
2218 else if (var_is_numeric (var))
2220 sys_warn (r, record->pos + ofs,
2221 _("Ignoring long string missing value record for "
2222 "numeric variable %s."), var_name);
2227 mv_init_pool (r->pool, &mv, var ? var_get_width (var) : 8);
2228 for (i = 0; i < n_missing_values; i++)
2230 size_t value_length;
2232 /* Parse value length. */
2233 check_overflow (r, record, ofs, 4);
2234 value_length = parse_int (r, record->data, ofs);
2238 check_overflow (r, record, ofs, value_length);
2241 && !mv_add_str (&mv, (const uint8_t *) record->data + ofs,
2243 sys_warn (r, record->pos + ofs,
2244 _("Ignoring long string missing value %zu for variable "
2245 "%s, with width %d, that has bad value width %zu."),
2246 i, var_get_name (var), var_get_width (var),
2248 ofs += value_length;
2251 var_set_missing_values (var, &mv);
2257 static void partial_record (struct sfm_reader *r)
2260 static void read_error (struct casereader *, const struct sfm_reader *);
2262 static bool read_case_number (struct sfm_reader *, double *);
2263 static bool read_case_string (struct sfm_reader *, uint8_t *, size_t);
2264 static int read_opcode (struct sfm_reader *);
2265 static bool read_compressed_number (struct sfm_reader *, double *);
2266 static bool read_compressed_string (struct sfm_reader *, uint8_t *);
2267 static bool read_whole_strings (struct sfm_reader *, uint8_t *, size_t);
2268 static bool skip_whole_strings (struct sfm_reader *, size_t);
2270 /* Reads and returns one case from READER's file. Returns a null
2271 pointer if not successful. */
2272 static struct ccase *
2273 sys_file_casereader_read (struct casereader *reader, void *r_)
2275 struct sfm_reader *r = r_;
2276 struct ccase *volatile c;
2282 c = case_create (r->proto);
2283 if (setjmp (r->bail_out))
2285 casereader_force_error (reader);
2290 for (i = 0; i < r->sfm_var_cnt; i++)
2292 struct sfm_var *sv = &r->sfm_vars[i];
2293 union value *v = case_data_rw_idx (c, sv->case_index);
2295 if (sv->var_width == 0)
2297 if (!read_case_number (r, &v->f))
2302 uint8_t *s = value_str_rw (v, sv->var_width);
2303 if (!read_case_string (r, s + sv->offset, sv->segment_width))
2305 if (!skip_whole_strings (r, ROUND_DOWN (sv->padding, 8)))
2314 if (r->case_cnt != -1)
2315 read_error (reader, r);
2320 /* Issues an error that R ends in a partial record. */
2322 partial_record (struct sfm_reader *r)
2324 sys_error (r, r->pos, _("File ends in partial case."));
2327 /* Issues an error that an unspecified error occurred SFM, and
2330 read_error (struct casereader *r, const struct sfm_reader *sfm)
2332 msg (ME, _("Error reading case from file %s."), fh_get_name (sfm->fh));
2333 casereader_force_error (r);
2336 /* Reads a number from R and stores its value in *D.
2337 If R is compressed, reads a compressed number;
2338 otherwise, reads a number in the regular way.
2339 Returns true if successful, false if end of file is
2340 reached immediately. */
2342 read_case_number (struct sfm_reader *r, double *d)
2344 if (r->compression == SFM_COMP_NONE)
2347 if (!try_read_bytes (r, number, sizeof number))
2349 float_convert (r->float_format, number, FLOAT_NATIVE_DOUBLE, d);
2353 return read_compressed_number (r, d);
2356 /* Reads LENGTH string bytes from R into S.
2357 Always reads a multiple of 8 bytes; if LENGTH is not a
2358 multiple of 8, then extra bytes are read and discarded without
2360 Reads compressed strings if S is compressed.
2361 Returns true if successful, false if end of file is
2362 reached immediately. */
2364 read_case_string (struct sfm_reader *r, uint8_t *s, size_t length)
2366 size_t whole = ROUND_DOWN (length, 8);
2367 size_t partial = length % 8;
2371 if (!read_whole_strings (r, s, whole))
2378 if (!read_whole_strings (r, bounce, sizeof bounce))
2384 memcpy (s + whole, bounce, partial);
2390 /* Reads and returns the next compression opcode from R. */
2392 read_opcode (struct sfm_reader *r)
2394 assert (r->compression != SFM_COMP_NONE);
2398 if (r->opcode_idx >= sizeof r->opcodes)
2400 if (!try_read_compressed_bytes (r, r->opcodes, sizeof r->opcodes))
2404 opcode = r->opcodes[r->opcode_idx++];
2411 /* Reads a compressed number from R and stores its value in D.
2412 Returns true if successful, false if end of file is
2413 reached immediately. */
2415 read_compressed_number (struct sfm_reader *r, double *d)
2417 int opcode = read_opcode (r);
2425 *d = read_compressed_float (r);
2429 float_convert (r->float_format, " ", FLOAT_NATIVE_DOUBLE, d);
2430 if (!r->corruption_warning)
2432 r->corruption_warning = true;
2433 sys_warn (r, r->pos,
2434 _("Possible compressed data corruption: "
2435 "compressed spaces appear in numeric field."));
2444 *d = opcode - r->bias;
2451 /* Reads a compressed 8-byte string segment from R and stores it
2453 Returns true if successful, false if end of file is
2454 reached immediately. */
2456 read_compressed_string (struct sfm_reader *r, uint8_t *dst)
2458 int opcode = read_opcode (r);
2466 read_compressed_bytes (r, dst, 8);
2470 memset (dst, ' ', 8);
2475 double value = opcode - r->bias;
2476 float_convert (FLOAT_NATIVE_DOUBLE, &value, r->float_format, dst);
2479 /* This has actually been seen "in the wild". The submitter of the
2480 file that showed that the contents decoded as spaces, but they
2481 were at the end of the field so it's possible that the null
2482 bytes just acted as null terminators. */
2484 else if (!r->corruption_warning)
2486 r->corruption_warning = true;
2487 sys_warn (r, r->pos,
2488 _("Possible compressed data corruption: "
2489 "string contains compressed integer (opcode %d)."),
2499 /* Reads LENGTH string bytes from R into S.
2500 LENGTH must be a multiple of 8.
2501 Reads compressed strings if S is compressed.
2502 Returns true if successful, false if end of file is
2503 reached immediately. */
2505 read_whole_strings (struct sfm_reader *r, uint8_t *s, size_t length)
2507 assert (length % 8 == 0);
2508 if (r->compression == SFM_COMP_NONE)
2509 return try_read_bytes (r, s, length);
2513 for (ofs = 0; ofs < length; ofs += 8)
2514 if (!read_compressed_string (r, s + ofs))
2524 /* Skips LENGTH string bytes from R.
2525 LENGTH must be a multiple of 8.
2526 (LENGTH is also limited to 1024, but that's only because the
2527 current caller never needs more than that many bytes.)
2528 Returns true if successful, false if end of file is
2529 reached immediately. */
2531 skip_whole_strings (struct sfm_reader *r, size_t length)
2533 uint8_t buffer[1024];
2534 assert (length < sizeof buffer);
2535 return read_whole_strings (r, buffer, length);
2538 /* Helpers for reading records that contain structured text
2541 /* Maximum number of warnings to issue for a single text
2543 #define MAX_TEXT_WARNINGS 5
2548 struct substring buffer; /* Record contents. */
2549 off_t start; /* Starting offset in file. */
2550 size_t pos; /* Current position in buffer. */
2551 int n_warnings; /* Number of warnings issued or suppressed. */
2552 bool recoded; /* Recoded into UTF-8? */
2555 static struct text_record *
2556 open_text_record (struct sfm_reader *r,
2557 const struct sfm_extension_record *record,
2558 bool recode_to_utf8)
2560 struct text_record *text;
2561 struct substring raw;
2563 text = pool_alloc (r->pool, sizeof *text);
2564 raw = ss_buffer (record->data, record->size * record->count);
2565 text->start = record->pos;
2566 text->buffer = (recode_to_utf8
2567 ? recode_substring_pool ("UTF-8", r->encoding, raw, r->pool)
2570 text->n_warnings = 0;
2571 text->recoded = recode_to_utf8;
2576 /* Closes TEXT, frees its storage, and issues a final warning
2577 about suppressed warnings if necesary. */
2579 close_text_record (struct sfm_reader *r, struct text_record *text)
2581 if (text->n_warnings > MAX_TEXT_WARNINGS)
2582 sys_warn (r, -1, _("Suppressed %d additional related warnings."),
2583 text->n_warnings - MAX_TEXT_WARNINGS);
2585 pool_free (r->pool, ss_data (text->buffer));
2588 /* Reads a variable=value pair from TEXT.
2589 Looks up the variable in DICT and stores it into *VAR.
2590 Stores a null-terminated value into *VALUE. */
2592 read_variable_to_value_pair (struct sfm_reader *r, struct dictionary *dict,
2593 struct text_record *text,
2594 struct variable **var, char **value)
2598 if (!text_read_short_name (r, dict, text, ss_cstr ("="), var))
2601 *value = text_get_token (text, ss_buffer ("\t\0", 2), NULL);
2605 text->pos += ss_span (ss_substr (text->buffer, text->pos, SIZE_MAX),
2606 ss_buffer ("\t\0", 2));
2614 text_read_variable_name (struct sfm_reader *r, struct dictionary *dict,
2615 struct text_record *text, struct substring delimiters,
2616 struct variable **var)
2620 name = text_get_token (text, delimiters, NULL);
2624 *var = dict_lookup_var (dict, name);
2628 text_warn (r, text, _("Dictionary record refers to unknown variable %s."),
2635 text_read_short_name (struct sfm_reader *r, struct dictionary *dict,
2636 struct text_record *text, struct substring delimiters,
2637 struct variable **var)
2639 char *short_name = text_get_token (text, delimiters, NULL);
2640 if (short_name == NULL)
2643 *var = dict_lookup_var (dict, short_name);
2645 text_warn (r, text, _("Dictionary record refers to unknown variable %s."),
2650 /* Displays a warning for the current file position, limiting the
2651 number to MAX_TEXT_WARNINGS for TEXT. */
2653 text_warn (struct sfm_reader *r, struct text_record *text,
2654 const char *format, ...)
2656 if (text->n_warnings++ < MAX_TEXT_WARNINGS)
2660 va_start (args, format);
2661 sys_msg (r, text->start + text->pos, MW, format, args);
2667 text_get_token (struct text_record *text, struct substring delimiters,
2670 struct substring token;
2673 if (!ss_tokenize (text->buffer, delimiters, &text->pos, &token))
2676 end = &ss_data (token)[ss_length (token)];
2677 if (delimiter != NULL)
2680 return ss_data (token);
2683 /* Reads a integer value expressed in decimal, then a space, then a string that
2684 consists of exactly as many bytes as specified by the integer, then a space,
2685 from TEXT. Returns the string, null-terminated, as a subset of TEXT's
2686 buffer (so the caller should not free the string). */
2688 text_parse_counted_string (struct sfm_reader *r, struct text_record *text)
2696 while (text->pos < text->buffer.length)
2698 int c = text->buffer.string[text->pos];
2699 if (c < '0' || c > '9')
2701 n = (n * 10) + (c - '0');
2704 if (text->pos >= text->buffer.length || start == text->pos)
2706 sys_warn (r, text->start,
2707 _("Expecting digit at offset %zu in MRSETS record."),
2712 if (!text_match (text, ' '))
2714 sys_warn (r, text->start,
2715 _("Expecting space at offset %zu in MRSETS record."),
2720 if (text->pos + n > text->buffer.length)
2722 sys_warn (r, text->start,
2723 _("%zu-byte string starting at offset %zu "
2724 "exceeds record length %zu."),
2725 n, text->pos, text->buffer.length);
2729 s = &text->buffer.string[text->pos];
2732 sys_warn (r, text->start,
2733 _("Expecting space at offset %zu following %zu-byte string."),
2743 text_match (struct text_record *text, char c)
2745 if (text->buffer.string[text->pos] == c)
2754 /* Returns the current byte offset (as converted to UTF-8, if it was converted)
2755 inside the TEXT's string. */
2757 text_pos (const struct text_record *text)
2763 text_get_all (const struct text_record *text)
2765 return text->buffer.string;
2770 /* Displays a corruption message. */
2772 sys_msg (struct sfm_reader *r, off_t offset,
2773 int class, const char *format, va_list args)
2778 ds_init_empty (&text);
2780 ds_put_format (&text, _("`%s' near offset 0x%llx: "),
2781 fh_get_file_name (r->fh), (long long int) offset);
2783 ds_put_format (&text, _("`%s': "), fh_get_file_name (r->fh));
2784 ds_put_vformat (&text, format, args);
2786 m.category = msg_class_to_category (class);
2787 m.severity = msg_class_to_severity (class);
2793 m.text = ds_cstr (&text);
2798 /* Displays a warning for offset OFFSET in the file. */
2800 sys_warn (struct sfm_reader *r, off_t offset, const char *format, ...)
2804 va_start (args, format);
2805 sys_msg (r, offset, MW, format, args);
2809 /* Displays an error for the current file position,
2810 marks it as in an error state,
2811 and aborts reading it using longjmp. */
2813 sys_error (struct sfm_reader *r, off_t offset, const char *format, ...)
2817 va_start (args, format);
2818 sys_msg (r, offset, ME, format, args);
2822 longjmp (r->bail_out, 1);
2825 /* Reads BYTE_CNT bytes into BUF.
2826 Returns true if exactly BYTE_CNT bytes are successfully read.
2827 Aborts if an I/O error or a partial read occurs.
2828 If EOF_IS_OK, then an immediate end-of-file causes false to be
2829 returned; otherwise, immediate end-of-file causes an abort
2832 read_bytes_internal (struct sfm_reader *r, bool eof_is_ok,
2833 void *buf, size_t byte_cnt)
2835 size_t bytes_read = fread (buf, 1, byte_cnt, r->file);
2836 r->pos += bytes_read;
2837 if (bytes_read == byte_cnt)
2839 else if (ferror (r->file))
2840 sys_error (r, r->pos, _("System error: %s."), strerror (errno));
2841 else if (!eof_is_ok || bytes_read != 0)
2842 sys_error (r, r->pos, _("Unexpected end of file."));
2847 /* Reads BYTE_CNT into BUF.
2848 Aborts upon I/O error or if end-of-file is encountered. */
2850 read_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
2852 read_bytes_internal (r, false, buf, byte_cnt);
2855 /* Reads BYTE_CNT bytes into BUF.
2856 Returns true if exactly BYTE_CNT bytes are successfully read.
2857 Returns false if an immediate end-of-file is encountered.
2858 Aborts if an I/O error or a partial read occurs. */
2860 try_read_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
2862 return read_bytes_internal (r, true, buf, byte_cnt);
2865 /* Reads a 32-bit signed integer from R and returns its value in
2868 read_int (struct sfm_reader *r)
2871 read_bytes (r, integer, sizeof integer);
2872 return integer_get (r->integer_format, integer, sizeof integer);
2875 /* Reads a 64-bit signed integer from R and returns its value in
2877 static long long int
2878 read_int64 (struct sfm_reader *r)
2881 read_bytes (r, integer, sizeof integer);
2882 return integer_get (r->integer_format, integer, sizeof integer);
2886 parse_int (struct sfm_reader *r, const void *data, size_t ofs)
2888 return integer_get (r->integer_format, (const uint8_t *) data + ofs, 4);
2892 parse_float (struct sfm_reader *r, const void *data, size_t ofs)
2894 return float_get_double (r->float_format, (const uint8_t *) data + ofs);
2897 /* Reads exactly SIZE - 1 bytes into BUFFER
2898 and stores a null byte into BUFFER[SIZE - 1]. */
2900 read_string (struct sfm_reader *r, char *buffer, size_t size)
2903 read_bytes (r, buffer, size - 1);
2904 buffer[size - 1] = '\0';
2907 /* Skips BYTES bytes forward in R. */
2909 skip_bytes (struct sfm_reader *r, size_t bytes)
2914 size_t chunk = MIN (sizeof buffer, bytes);
2915 read_bytes (r, buffer, chunk);
2920 /* Returns a malloc()'d copy of S in which all lone CRs and CR LF pairs have
2921 been replaced by LFs.
2923 (A product that identifies itself as VOXCO INTERVIEWER 4.3 produces system
2924 files that use CR-only line ends in the file label and extra product
2927 fix_line_ends (const char *s)
2931 d = dst = xmalloc (strlen (s) + 1);
2950 read_ztrailer (struct sfm_reader *r,
2951 long long int zheader_ofs,
2952 long long int ztrailer_len);
2955 zalloc (voidpf pool_, uInt items, uInt size)
2957 struct pool *pool = pool_;
2959 return (!size || xalloc_oversized (items, size)
2961 : pool_malloc (pool, items * size));
2965 zfree (voidpf pool_, voidpf address)
2967 struct pool *pool = pool_;
2969 pool_free (pool, address);
2973 read_zheader (struct sfm_reader *r)
2976 long long int zheader_ofs = read_int64 (r);
2977 long long int ztrailer_ofs = read_int64 (r);
2978 long long int ztrailer_len = read_int64 (r);
2980 if (zheader_ofs != pos)
2981 sys_error (r, pos, _("Wrong ZLIB data header offset %#llx "
2982 "(expected %#llx)."),
2983 zheader_ofs, (long long int) pos);
2985 if (ztrailer_ofs < r->pos)
2986 sys_error (r, pos, _("Impossible ZLIB trailer offset 0x%llx."),
2989 if (ztrailer_len < 24 || ztrailer_len % 24)
2990 sys_error (r, pos, _("Invalid ZLIB trailer length %lld."), ztrailer_len);
2992 r->ztrailer_ofs = ztrailer_ofs;
2993 read_ztrailer (r, zheader_ofs, ztrailer_len);
2995 if (r->zin_buf == NULL)
2997 r->zin_buf = pool_malloc (r->pool, ZIN_BUF_SIZE);
2998 r->zout_buf = pool_malloc (r->pool, ZOUT_BUF_SIZE);
2999 r->zstream.next_in = NULL;
3000 r->zstream.avail_in = 0;
3003 r->zstream.zalloc = zalloc;
3004 r->zstream.zfree = zfree;
3005 r->zstream.opaque = r->pool;
3011 seek (struct sfm_reader *r, off_t offset)
3013 if (fseeko (r->file, offset, SEEK_SET))
3014 sys_error (r, 0, _("%s: seek failed (%s)."),
3015 fh_get_file_name (r->fh), strerror (errno));
3019 /* Performs some additional consistency checks on the ZLIB compressed data
3022 read_ztrailer (struct sfm_reader *r,
3023 long long int zheader_ofs,
3024 long long int ztrailer_len)
3026 long long int expected_uncmp_ofs;
3027 long long int expected_cmp_ofs;
3030 unsigned int block_size;
3031 unsigned int n_blocks;
3035 if (fstat (fileno (r->file), &s))
3036 sys_error (ME, 0, _("%s: stat failed (%s)."),
3037 fh_get_file_name (r->fh), strerror (errno));
3039 if (!S_ISREG (s.st_mode))
3041 /* We can't seek to the trailer and then back to the data in this file,
3042 so skip doing extra checks. */
3046 if (r->ztrailer_ofs + ztrailer_len != s.st_size)
3047 sys_warn (r, r->pos,
3048 _("End of ZLIB trailer (0x%llx) is not file size (0x%llx)."),
3049 r->ztrailer_ofs + ztrailer_len, (long long int) s.st_size);
3051 seek (r, r->ztrailer_ofs);
3053 /* Read fixed header from ZLIB data trailer. */
3054 bias = read_int64 (r);
3055 if (-bias != r->bias)
3056 sys_error (r, r->pos, _("ZLIB trailer bias (%lld) differs from "
3057 "file header bias (%.2f)."),
3060 zero = read_int64 (r);
3062 sys_warn (r, r->pos,
3063 _("ZLIB trailer \"zero\" field has nonzero value %lld."), zero);
3065 block_size = read_int (r);
3066 if (block_size != ZBLOCK_SIZE)
3067 sys_warn (r, r->pos,
3068 _("ZLIB trailer specifies unexpected %u-byte block size."),
3071 n_blocks = read_int (r);
3072 if (n_blocks != (ztrailer_len - 24) / 24)
3073 sys_error (r, r->pos,
3074 _("%lld-byte ZLIB trailer specifies %u data blocks (expected "
3076 ztrailer_len, n_blocks, (ztrailer_len - 24) / 24);
3078 expected_uncmp_ofs = zheader_ofs;
3079 expected_cmp_ofs = zheader_ofs + 24;
3080 for (i = 0; i < n_blocks; i++)
3082 off_t desc_ofs = r->pos;
3083 unsigned long long int uncompressed_ofs = read_int64 (r);
3084 unsigned long long int compressed_ofs = read_int64 (r);
3085 unsigned int uncompressed_size = read_int (r);
3086 unsigned int compressed_size = read_int (r);
3088 if (uncompressed_ofs != expected_uncmp_ofs)
3089 sys_error (r, desc_ofs,
3090 _("ZLIB block descriptor %u reported uncompressed data "
3091 "offset %#llx, when %#llx was expected."),
3092 i, uncompressed_ofs, expected_uncmp_ofs);
3094 if (compressed_ofs != expected_cmp_ofs)
3095 sys_error (r, desc_ofs,
3096 _("ZLIB block descriptor %u reported compressed data "
3097 "offset %#llx, when %#llx was expected."),
3098 i, compressed_ofs, expected_cmp_ofs);
3100 if (i < n_blocks - 1)
3102 if (uncompressed_size != block_size)
3103 sys_warn (r, desc_ofs,
3104 _("ZLIB block descriptor %u reported block size %#x, "
3105 "when %#x was expected."),
3106 i, uncompressed_size, block_size);
3110 if (uncompressed_size > block_size)
3111 sys_warn (r, desc_ofs,
3112 _("ZLIB block descriptor %u reported block size %#x, "
3113 "when at most %#x was expected."),
3114 i, uncompressed_size, block_size);
3117 /* http://www.zlib.net/zlib_tech.html says that the maximum expansion
3118 from compression, with worst-case parameters, is 13.5% plus 11 bytes.
3119 This code checks for an expansion of more than 14.3% plus 11
3121 if (compressed_size > uncompressed_size + uncompressed_size / 7 + 11)
3122 sys_error (r, desc_ofs,
3123 _("ZLIB block descriptor %u reports compressed size %u "
3124 "and uncompressed size %u."),
3125 i, compressed_size, uncompressed_size);
3127 expected_uncmp_ofs += uncompressed_size;
3128 expected_cmp_ofs += compressed_size;
3131 if (expected_cmp_ofs != r->ztrailer_ofs)
3132 sys_error (r, r->pos, _("ZLIB trailer is at offset %#llx but %#llx "
3133 "would be expected from block descriptors."),
3134 r->ztrailer_ofs, expected_cmp_ofs);
3136 seek (r, zheader_ofs + 24);
3140 open_zstream (struct sfm_reader *r)
3144 r->zout_pos = r->zout_end = 0;
3145 error = inflateInit (&r->zstream);
3147 sys_error (r, r->pos, _("ZLIB initialization failed (%s)."),
3152 close_zstream (struct sfm_reader *r)
3156 error = inflateEnd (&r->zstream);
3158 sys_error (r, r->pos, _("Inconsistency at end of ZLIB stream (%s)."),
3163 read_bytes_zlib (struct sfm_reader *r, void *buf_, size_t byte_cnt)
3165 uint8_t *buf = buf_;
3174 /* Use already inflated data if there is any. */
3175 if (r->zout_pos < r->zout_end)
3177 unsigned int n = MIN (byte_cnt, r->zout_end - r->zout_pos);
3178 memcpy (buf, &r->zout_buf[r->zout_pos], n);
3187 /* We need to inflate some more data.
3188 Get some more input data if we don't have any. */
3189 if (r->zstream.avail_in == 0)
3191 unsigned int n = MIN (ZIN_BUF_SIZE, r->ztrailer_ofs - r->pos);
3192 if (n == 0 || !try_read_bytes (r, r->zin_buf, n))
3194 r->zstream.avail_in = n;
3195 r->zstream.next_in = r->zin_buf;
3198 /* Inflate the (remaining) input data. */
3199 r->zstream.avail_out = ZOUT_BUF_SIZE;
3200 r->zstream.next_out = r->zout_buf;
3201 error = inflate (&r->zstream, Z_SYNC_FLUSH);
3203 r->zout_end = r->zstream.next_out - r->zout_buf;
3204 if (r->zout_end == 0)
3206 if (error == Z_STREAM_END)
3212 sys_error (r, r->pos, _("ZLIB stream inconsistency (%s)."),
3217 /* Process the output data and ignore 'error' for now. ZLIB will
3218 present it to us again on the next inflate() call. */
3224 read_compressed_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
3226 if (r->compression == SFM_COMP_SIMPLE)
3227 return read_bytes (r, buf, byte_cnt);
3228 else if (!read_bytes_zlib (r, buf, byte_cnt))
3229 sys_error (r, r->pos, _("Unexpected end of ZLIB compressed data."));
3233 try_read_compressed_bytes (struct sfm_reader *r, void *buf, size_t byte_cnt)
3235 if (r->compression == SFM_COMP_SIMPLE)
3236 return try_read_bytes (r, buf, byte_cnt);
3238 return read_bytes_zlib (r, buf, byte_cnt);
3241 /* Reads a 64-bit floating-point number from R and returns its
3242 value in host format. */
3244 read_compressed_float (struct sfm_reader *r)
3247 read_compressed_bytes (r, number, sizeof number);
3248 return float_get_double (r->float_format, number);
3251 static const struct casereader_class sys_file_casereader_class =
3253 sys_file_casereader_read,
3254 sys_file_casereader_destroy,