X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fdata%2Fsys-file-writer.c;h=2369370ca10a2fd09f4a0b80169912033542f665;hb=1076f8e88e2eee1bc780edd9d79d696a38af936d;hp=393be4e1265292b58fc37daa64bfaf4a4ff40a15;hpb=8af88c0b7ea2fe75df7e45497988ed0371006a86;p=pspp diff --git a/src/data/sys-file-writer.c b/src/data/sys-file-writer.c index 393be4e126..2369370ca1 100644 --- a/src/data/sys-file-writer.c +++ b/src/data/sys-file-writer.c @@ -1,5 +1,5 @@ /* PSPP - a program for statistical analysis. - Copyright (C) 1997-9, 2000, 2006, 2007, 2009 Free Software Foundation, Inc. + Copyright (C) 1997-2000, 2006-2013 Free Software Foundation, Inc. This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by @@ -16,8 +16,8 @@ #include -#include "sys-file-writer.h" -#include "sys-file-private.h" +#include "data/sys-file-writer.h" +#include "data/sys-file-private.h" #include #include @@ -26,31 +26,34 @@ #include #include -#include -#include -#include -#include -#include -#include - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include "minmax.h" -#include "unlocked-io.h" -#include "xalloc.h" +#include "data/attributes.h" +#include "data/case.h" +#include "data/casewriter-provider.h" +#include "data/casewriter.h" +#include "data/dictionary.h" +#include "data/file-handle-def.h" +#include "data/file-name.h" +#include "data/format.h" +#include "data/make-file.h" +#include "data/missing-values.h" +#include "data/mrset.h" +#include "data/settings.h" +#include "data/short-names.h" +#include "data/value-labels.h" +#include "data/variable.h" +#include "libpspp/float-format.h" +#include "libpspp/i18n.h" +#include "libpspp/integer-format.h" +#include "libpspp/message.h" +#include "libpspp/misc.h" +#include "libpspp/str.h" +#include "libpspp/string-array.h" +#include "libpspp/version.h" + +#include "gl/xmemdup0.h" +#include "gl/minmax.h" +#include "gl/unlocked-io.h" +#include "gl/xalloc.h" #include "gettext.h" #define _(msgid) gettext (msgid) @@ -71,6 +74,7 @@ struct sfm_writer bool compress; /* 1=compressed, 0=not compressed. */ casenumber case_cnt; /* Number of cases written so far. */ + uint8_t space; /* ' ' in the file's character encoding. */ /* Compression buffering. @@ -96,16 +100,25 @@ static const struct casewriter_class sys_file_casewriter_class; static void write_header (struct sfm_writer *, const struct dictionary *); static void write_variable (struct sfm_writer *, const struct variable *); static void write_value_labels (struct sfm_writer *, - struct variable *, int idx); -static void write_integer_info_record (struct sfm_writer *); + const struct dictionary *); +static void write_integer_info_record (struct sfm_writer *, + const struct dictionary *); static void write_float_info_record (struct sfm_writer *); static void write_longvar_table (struct sfm_writer *w, const struct dictionary *dict); +static void write_encoding_record (struct sfm_writer *w, + const struct dictionary *); + static void write_vls_length_table (struct sfm_writer *w, const struct dictionary *dict); +static void write_long_string_value_labels (struct sfm_writer *, + const struct dictionary *); + +static void write_mrsets (struct sfm_writer *, const struct dictionary *, + bool pre_v14); static void write_variable_display_parameters (struct sfm_writer *w, const struct dictionary *dict); @@ -121,6 +134,12 @@ static void write_int (struct sfm_writer *, int32_t); static inline void convert_double_to_output_format (double, uint8_t[8]); static void write_float (struct sfm_writer *, double); static void write_string (struct sfm_writer *, const char *, size_t); +static void write_utf8_string (struct sfm_writer *, const char *encoding, + const char *string, size_t width); +static void write_utf8_record (struct sfm_writer *, const char *encoding, + const struct string *content, int subtype); +static void write_string_record (struct sfm_writer *, + const struct substring content, int subtype); static void write_bytes (struct sfm_writer *, const void *, size_t); static void write_zeros (struct sfm_writer *, size_t); static void write_spaces (struct sfm_writer *, size_t); @@ -134,8 +153,8 @@ static void put_cmp_opcode (struct sfm_writer *, uint8_t); static void put_cmp_number (struct sfm_writer *, double); static void put_cmp_string (struct sfm_writer *, const void *, size_t); -bool write_error (const struct sfm_writer *); -bool close_writer (struct sfm_writer *); +static bool write_error (const struct sfm_writer *); +static bool close_writer (struct sfm_writer *); /* Returns default options for writing a system file. */ struct sfm_write_options @@ -149,8 +168,7 @@ sfm_writer_default_options (void) } /* Opens the system file designated by file handle FH for writing - cases from dictionary D according to the given OPTS. If - COMPRESS is nonzero, the system file will be compressed. + cases from dictionary D according to the given OPTS. No reference to D is retained, so it may be modified or destroyed at will after this function returns. D is not @@ -159,9 +177,9 @@ struct casewriter * sfm_open_writer (struct file_handle *fh, struct dictionary *d, struct sfm_write_options opts) { + struct encoding_info encoding_info; struct sfm_writer *w; mode_t mode; - int idx; int i; /* Check version. */ @@ -199,18 +217,21 @@ sfm_open_writer (struct file_handle *fh, struct dictionary *d, goto error; /* Create the file on disk. */ - mode = S_IRUSR | S_IRGRP | S_IROTH; + mode = 0444; if (opts.create_writeable) - mode |= S_IWUSR | S_IWGRP | S_IWOTH; + mode |= 0222; w->rf = replace_file_start (fh_get_file_name (fh), "wb", mode, &w->file, NULL); if (w->rf == NULL) { - msg (ME, _("Error opening \"%s\" for writing as a system file: %s."), + msg (ME, _("Error opening `%s' for writing as a system file: %s."), fh_get_file_name (fh), strerror (errno)); goto error; } + get_encoding_info (&encoding_info, dict_get_encoding (d)); + w->space = encoding_info.space[0]; + /* Write the file header. */ write_header (w, d); @@ -219,22 +240,16 @@ sfm_open_writer (struct file_handle *fh, struct dictionary *d, for (i = 0; i < dict_get_var_cnt (d); i++) write_variable (w, dict_get_var (d, i)); - /* Write out value labels. */ - idx = 0; - for (i = 0; i < dict_get_var_cnt (d); i++) - { - struct variable *v = dict_get_var (d, i); - - write_value_labels (w, v, idx); - idx += sfm_width_to_octs (var_get_width (v)); - } + write_value_labels (w, d); - if (dict_get_documents (d) != NULL) + if (dict_get_document_line_cnt (d) > 0) write_documents (w, d); - write_integer_info_record (w); + write_integer_info_record (w, d); write_float_info_record (w); + write_mrsets (w, d, true); + write_variable_display_parameters (w, d); if (opts.version >= 3) @@ -242,22 +257,27 @@ sfm_open_writer (struct file_handle *fh, struct dictionary *d, write_vls_length_table (w, d); - if (attrset_count (dict_get_attributes (d))) - write_data_file_attributes (w, d); - write_variable_attributes (w, d); + write_long_string_value_labels (w, d); + + if (opts.version >= 3) + { + if (attrset_count (dict_get_attributes (d))) + write_data_file_attributes (w, d); + write_variable_attributes (w, d); + } + + write_mrsets (w, d, false); + + write_encoding_record (w, d); /* Write end-of-headers record. */ write_int (w, 999); write_int (w, 0); if (write_error (w)) - { - close_writer (w); - return NULL; - } + goto error; - return casewriter_create (dict_get_next_value_idx (d), - &sys_file_casewriter_class, w); + return casewriter_create (dict_get_proto (d), &sys_file_casewriter_class, w); error: close_writer (w); @@ -299,6 +319,7 @@ calc_oct_idx (const struct dictionary *d, struct variable *target_var) static void write_header (struct sfm_writer *w, const struct dictionary *d) { + const char *dict_encoding = dict_get_encoding (d); char prod_name[61]; char creation_date[10]; char creation_time[9]; @@ -308,12 +329,15 @@ write_header (struct sfm_writer *w, const struct dictionary *d) time_t t; /* Record-type code. */ - write_string (w, "$FL2", 4); + if (is_encoding_ebcdic_compatible (dict_encoding)) + write_string (w, EBCDIC_MAGIC, 4); + else + write_string (w, ASCII_MAGIC, 4); /* Product identification. */ snprintf (prod_name, sizeof prod_name, "@(#) SPSS DATA FILE %s - %s", version, host_system); - write_string (w, prod_name, 60); + write_utf8_string (w, dict_encoding, prod_name, 60); /* Layout code. */ write_int (w, 2); @@ -362,14 +386,14 @@ write_header (struct sfm_writer *w, const struct dictionary *d) snprintf (creation_time, sizeof creation_time, "%02d:%02d:%02d", hour - 1, min - 1, sec - 1); } - write_string (w, creation_date, 9); - write_string (w, creation_time, 8); + write_utf8_string (w, dict_encoding, creation_date, 9); + write_utf8_string (w, dict_encoding, creation_time, 8); /* File label. */ file_label = dict_get_label (d); if (file_label == NULL) file_label = ""; - write_string (w, file_label, 64); + write_utf8_string (w, dict_encoding, file_label, 64); /* Padding. */ write_zeros (w, 3); @@ -417,7 +441,8 @@ write_variable (struct sfm_writer *w, const struct variable *v) int width = var_get_width (v); int segment_cnt = sfm_width_to_segments (width); int seg0_width = sfm_segment_alloc_width (width, 0); - const struct missing_values *mv = var_get_missing_values (v); + const char *encoding = var_get_encoding (v); + struct missing_values mv; int i; /* Record type. */ @@ -432,7 +457,13 @@ write_variable (struct sfm_writer *w, const struct variable *v) /* Number of missing values. If there is a range, then the range counts as 2 missing values and causes the number to be negated. */ - write_int (w, mv_has_range (mv) ? -2 - mv_n_values (mv) : mv_n_values (mv)); + mv_copy (&mv, var_get_missing_values (v)); + if (mv_get_width (&mv) > 8) + mv_resize (&mv, 8); + if (mv_has_range (&mv)) + write_int (w, -2 - mv_n_values (&mv)); + else + write_int (w, mv_n_values (&mv)); /* Print and write formats. */ write_format (w, *var_get_print_format (v), seg0_width); @@ -441,31 +472,29 @@ write_variable (struct sfm_writer *w, const struct variable *v) /* Short name. The full name is in a translation table written separately. */ - write_string (w, var_get_short_name (v, 0), 8); + write_utf8_string (w, encoding, var_get_short_name (v, 0), 8); /* Value label. */ if (var_has_label (v)) { - const char *label = var_get_label (v); - size_t padded_len = ROUND_UP (MIN (strlen (label), 255), 4); - write_int (w, padded_len); + char *label = recode_string (encoding, UTF8, var_get_label (v), -1); + size_t label_len = MIN (strlen (label), 255); + size_t padded_len = ROUND_UP (label_len, 4); + write_int (w, label_len); write_string (w, label, padded_len); + free (label); } /* Write the missing values, if any, range first. */ - if (mv_has_range (mv)) + if (mv_has_range (&mv)) { double x, y; - mv_get_range (mv, &x, &y); + mv_get_range (&mv, &x, &y); write_float (w, x); write_float (w, y); } - for (i = 0; i < mv_n_values (mv); i++) - { - union value value; - mv_get_value (mv, &value, i); - write_value (w, &value, seg0_width); - } + for (i = 0; i < mv_n_values (&mv); i++) + write_value (w, mv_get_value (&mv, i), mv_get_width (&mv)); write_variable_continuation_records (w, seg0_width); @@ -481,54 +510,147 @@ write_variable (struct sfm_writer *w, const struct variable *v) write_int (w, 0); /* No missing values. */ write_format (w, fmt, seg_width); /* Print format. */ write_format (w, fmt, seg_width); /* Write format. */ - write_string (w, var_get_short_name (v, i), 8); + write_utf8_string (w, encoding, var_get_short_name (v, i), 8); write_variable_continuation_records (w, seg_width); } + + mv_destroy (&mv); } -/* Writes the value labels for variable V having system file - variable index IDX to system file W. */ +/* Writes the value labels to system file W. + + Value labels for long string variables are written separately, + by write_long_string_value_labels. */ static void -write_value_labels (struct sfm_writer *w, struct variable *v, int idx) +write_value_labels (struct sfm_writer *w, const struct dictionary *d) { - const struct val_labs *val_labs; - struct val_labs_iterator *i; - struct val_lab *vl; + struct label_set + { + struct hmap_node hmap_node; + const struct val_labs *val_labs; + int *indexes; + size_t n_indexes, allocated_indexes; + }; + + size_t n_sets, allocated_sets; + struct label_set **sets; + struct hmap same_sets; + size_t i; + int idx; - val_labs = var_get_value_labels (v); - if (val_labs == NULL) - return; + n_sets = allocated_sets = 0; + sets = NULL; + hmap_init (&same_sets); - /* Value label record. */ - write_int (w, 3); /* Record type. */ - write_int (w, val_labs_count (val_labs)); - for (vl = val_labs_first_sorted (val_labs, &i); vl != NULL; - vl = val_labs_next (val_labs, &i)) + idx = 0; + for (i = 0; i < dict_get_var_cnt (d); i++) { - uint8_t len = MIN (strlen (vl->label), 255); + struct variable *v = dict_get_var (d, i); - write_value (w, &vl->value, var_get_width (v)); - write_bytes (w, &len, 1); - write_bytes (w, vl->label, len); - write_zeros (w, REM_RND_UP (len + 1, 8)); + if (var_has_value_labels (v) && var_get_width (v) <= 8) + { + const struct val_labs *val_labs = var_get_value_labels (v); + unsigned int hash = val_labs_hash (val_labs, 0); + struct label_set *set; + + HMAP_FOR_EACH_WITH_HASH (set, struct label_set, hmap_node, + hash, &same_sets) + { + if (val_labs_equal (set->val_labs, val_labs)) + { + if (set->n_indexes >= set->allocated_indexes) + set->indexes = x2nrealloc (set->indexes, + &set->allocated_indexes, + sizeof *set->indexes); + set->indexes[set->n_indexes++] = idx; + goto next_var; + } + } + + set = xmalloc (sizeof *set); + set->val_labs = val_labs; + set->indexes = xmalloc (sizeof *set->indexes); + set->indexes[0] = idx; + set->n_indexes = 1; + set->allocated_indexes = 1; + hmap_insert (&same_sets, &set->hmap_node, hash); + + if (n_sets >= allocated_sets) + sets = x2nrealloc (sets, &allocated_sets, sizeof *sets); + sets[n_sets++] = set; + } + + next_var: + idx += sfm_width_to_octs (var_get_width (v)); + } + + for (i = 0; i < n_sets; i++) + { + const struct label_set *set = sets[i]; + const struct val_labs *val_labs = set->val_labs; + size_t n_labels = val_labs_count (val_labs); + int width = val_labs_get_width (val_labs); + const struct val_lab **labels; + size_t j; + + /* Value label record. */ + write_int (w, 3); /* Record type. */ + write_int (w, n_labels); + labels = val_labs_sorted (val_labs); + for (j = 0; j < n_labels; j++) + { + const struct val_lab *vl = labels[j]; + char *label = recode_string (dict_get_encoding (d), UTF8, + val_lab_get_escaped_label (vl), -1); + uint8_t len = MIN (strlen (label), 255); + + write_value (w, val_lab_get_value (vl), width); + write_bytes (w, &len, 1); + write_bytes (w, label, len); + write_zeros (w, REM_RND_UP (len + 1, 8)); + free (label); + } + free (labels); + + /* Value label variable record. */ + write_int (w, 4); /* Record type. */ + write_int (w, set->n_indexes); + for (j = 0; j < set->n_indexes; j++) + write_int (w, set->indexes[j] + 1); } - /* Value label variable record. */ - write_int (w, 4); /* Record type. */ - write_int (w, 1); /* Number of variables. */ - write_int (w, idx + 1); /* Variable's dictionary index. */ + for (i = 0; i < n_sets; i++) + { + struct label_set *set = sets[i]; + + free (set->indexes); + free (set); + } + free (sets); + hmap_destroy (&same_sets); } /* Writes record type 6, document record. */ static void write_documents (struct sfm_writer *w, const struct dictionary *d) { - size_t line_cnt = dict_get_document_line_cnt (d); + const struct string_array *docs = dict_get_documents (d); + const char *enc = dict_get_encoding (d); + size_t i; write_int (w, 6); /* Record type. */ - write_int (w, line_cnt); - write_bytes (w, dict_get_documents (d), line_cnt * DOC_LINE_LENGTH); + write_int (w, docs->n); + for (i = 0; i < docs->n; i++) + { + char *s = recode_string (enc, "UTF-8", docs->strings[i], -1); + size_t s_len = strlen (s); + size_t write_len = MIN (s_len, DOC_LINE_LENGTH); + + write_bytes (w, s, write_len); + write_spaces (w, DOC_LINE_LENGTH - write_len); + free (s); + } } static void @@ -544,34 +666,63 @@ put_attrset (struct string *string, const struct attrset *attrs) size_t j; ds_put_cstr (string, attribute_get_name (attr)); - ds_put_char (string, '('); + ds_put_byte (string, '('); for (j = 0; j < n_values; j++) ds_put_format (string, "'%s'\n", attribute_get_value (attr, j)); - ds_put_char (string, ')'); + ds_put_byte (string, ')'); } } -static void -write_attribute_record (struct sfm_writer *w, const struct string *content, - int subtype) -{ - write_int (w, 7); - write_int (w, subtype); - write_int (w, 1); - write_int (w, ds_length (content)); - write_bytes (w, ds_data (content), ds_length (content)); -} - static void write_data_file_attributes (struct sfm_writer *w, const struct dictionary *d) { struct string s = DS_EMPTY_INITIALIZER; put_attrset (&s, dict_get_attributes (d)); - write_attribute_record (w, &s, 17); + write_utf8_record (w, dict_get_encoding (d), &s, 17); ds_destroy (&s); } +static void +add_role_attribute (enum var_role role, struct attrset *attrs) +{ + struct attribute *attr; + const char *s; + + switch (role) + { + case ROLE_INPUT: + default: + s = "0"; + break; + + case ROLE_OUTPUT: + s = "1"; + break; + + case ROLE_BOTH: + s = "2"; + break; + + case ROLE_NONE: + s = "3"; + break; + + case ROLE_PARTITION: + s = "4"; + break; + + case ROLE_SPLIT: + s = "5"; + break; + } + attrset_delete (attrs, "$@Role"); + + attr = attribute_create ("$@Role"); + attribute_add_value (attr, s); + attrset_add (attrs, attr); +} + static void write_variable_attributes (struct sfm_writer *w, const struct dictionary *d) { @@ -583,17 +734,103 @@ write_variable_attributes (struct sfm_writer *w, const struct dictionary *d) for (i = 0; i < n_vars; i++) { struct variable *v = dict_get_var (d, i); - struct attrset *attrs = var_get_attributes (v); - if (attrset_count (attrs)) + struct attrset attrs; + + attrset_clone (&attrs, var_get_attributes (v)); + + add_role_attribute (var_get_role (v), &attrs); + if (n_attrsets++) + ds_put_byte (&s, '/'); + ds_put_format (&s, "%s:", var_get_name (v)); + put_attrset (&s, &attrs); + attrset_destroy (&attrs); + } + if (n_attrsets) + write_utf8_record (w, dict_get_encoding (d), &s, 18); + ds_destroy (&s); +} + +/* Write multiple response sets. If PRE_V14 is true, writes sets supported by + SPSS before release 14, otherwise writes sets supported only by later + versions. */ +static void +write_mrsets (struct sfm_writer *w, const struct dictionary *dict, + bool pre_v14) +{ + const char *encoding = dict_get_encoding (dict); + struct string s = DS_EMPTY_INITIALIZER; + size_t n_mrsets; + size_t i; + + if (is_encoding_ebcdic_compatible (encoding)) + { + /* FIXME. */ + return; + } + + n_mrsets = dict_get_n_mrsets (dict); + if (n_mrsets == 0) + return; + + for (i = 0; i < n_mrsets; i++) + { + const struct mrset *mrset = dict_get_mrset (dict, i); + char *name; + size_t j; + + if ((mrset->type != MRSET_MD || mrset->cat_source != MRSET_COUNTEDVALUES) + != pre_v14) + continue; + + name = recode_string (encoding, "UTF-8", mrset->name, -1); + ds_put_format (&s, "%s=", name); + free (name); + + if (mrset->type == MRSET_MD) { - if (n_attrsets++) - ds_put_char (&s, '/'); - ds_put_format (&s, "%s:", var_get_short_name (v, 0)); - put_attrset (&s, attrs); + char *counted; + + if (mrset->cat_source == MRSET_COUNTEDVALUES) + ds_put_format (&s, "E %d ", mrset->label_from_var_label ? 11 : 1); + else + ds_put_byte (&s, 'D'); + + if (mrset->width == 0) + counted = xasprintf ("%.0f", mrset->counted.f); + else + counted = xmemdup0 (value_str (&mrset->counted, mrset->width), + mrset->width); + ds_put_format (&s, "%zu %s", strlen (counted), counted); + free (counted); + } + else + ds_put_byte (&s, 'C'); + ds_put_byte (&s, ' '); + + if (mrset->label && !mrset->label_from_var_label) + { + char *label = recode_string (encoding, "UTF-8", mrset->label, -1); + ds_put_format (&s, "%zu %s", strlen (label), label); + free (label); + } + else + ds_put_cstr (&s, "0 "); + + for (j = 0; j < mrset->n_vars; j++) + { + const char *short_name_utf8 = var_get_short_name (mrset->vars[j], 0); + char *lower_name_utf8 = utf8_to_lower (short_name_utf8); + char *short_name = recode_string (encoding, "UTF-8", + lower_name_utf8, -1); + ds_put_format (&s, " %s", short_name); + free (short_name); + free (lower_name_utf8); } + ds_put_byte (&s, '\n'); } - if (n_attrsets) - write_attribute_record (w, &s, 18); + + if (!ds_is_empty (&s)) + write_string_record (w, ds_ss (&s), pre_v14 ? 7 : 19); ds_destroy (&s); } @@ -650,14 +887,105 @@ write_vls_length_table (struct sfm_writer *w, var_get_short_name (v, 0), var_get_width (v), 0); } if (!ds_is_empty (&map)) + write_utf8_record (w, dict_get_encoding (dict), &map, 14); + ds_destroy (&map); +} + + +static void +write_long_string_value_labels (struct sfm_writer *w, + const struct dictionary *dict) +{ + size_t n_vars = dict_get_var_cnt (dict); + size_t size, i; + off_t start UNUSED; + + /* Figure out the size in advance. */ + size = 0; + for (i = 0; i < n_vars; i++) { - write_int (w, 7); /* Record type. */ - write_int (w, 14); /* Record subtype. */ - write_int (w, 1); /* Data item (char) size. */ - write_int (w, ds_length (&map)); /* Number of data items. */ - write_bytes (w, ds_data (&map), ds_length (&map)); + struct variable *var = dict_get_var (dict, i); + const struct val_labs *val_labs = var_get_value_labels (var); + const char *encoding = var_get_encoding (var); + int width = var_get_width (var); + const struct val_lab *val_lab; + + if (val_labs_count (val_labs) == 0 || width < 9) + continue; + + size += 12; + size += recode_string_len (encoding, "UTF-8", var_get_name (var), -1); + for (val_lab = val_labs_first (val_labs); val_lab != NULL; + val_lab = val_labs_next (val_labs, val_lab)) + { + size += 8 + width; + size += recode_string_len (encoding, "UTF-8", + val_lab_get_escaped_label (val_lab), -1); + } } - ds_destroy (&map); + if (size == 0) + return; + + write_int (w, 7); /* Record type. */ + write_int (w, 21); /* Record subtype */ + write_int (w, 1); /* Data item (byte) size. */ + write_int (w, size); /* Number of data items. */ + + start = ftello (w->file); + for (i = 0; i < n_vars; i++) + { + struct variable *var = dict_get_var (dict, i); + const struct val_labs *val_labs = var_get_value_labels (var); + const char *encoding = var_get_encoding (var); + int width = var_get_width (var); + const struct val_lab *val_lab; + char *var_name; + + if (val_labs_count (val_labs) == 0 || width < 9) + continue; + + var_name = recode_string (encoding, "UTF-8", var_get_name (var), -1); + write_int (w, strlen (var_name)); + write_bytes (w, var_name, strlen (var_name)); + free (var_name); + + write_int (w, width); + write_int (w, val_labs_count (val_labs)); + for (val_lab = val_labs_first (val_labs); val_lab != NULL; + val_lab = val_labs_next (val_labs, val_lab)) + { + char *label; + size_t len; + + write_int (w, width); + write_bytes (w, value_str (val_lab_get_value (val_lab), width), + width); + + label = recode_string (var_get_encoding (var), "UTF-8", + val_lab_get_escaped_label (val_lab), -1); + len = strlen (label); + write_int (w, len); + write_bytes (w, label, len); + free (label); + } + } + assert (ftello (w->file) == start + size); +} + +static void +write_encoding_record (struct sfm_writer *w, + const struct dictionary *d) +{ + /* IANA says "...character set names may be up to 40 characters taken + from the printable characters of US-ASCII," so character set names + don't need to be recoded to be in UTF-8. + + We convert encoding names to uppercase because SPSS writes encoding + names in uppercase. */ + char *encoding = xstrdup (dict_get_encoding (d)); + str_uppercase (encoding); + write_string_record (w, ss_cstr (encoding), 20); + free (encoding); } /* Writes the long variable name table. */ @@ -671,28 +999,24 @@ write_longvar_table (struct sfm_writer *w, const struct dictionary *dict) for (i = 0; i < dict_get_var_cnt (dict); i++) { struct variable *v = dict_get_var (dict, i); - if (i) - ds_put_char (&map, '\t'); + ds_put_byte (&map, '\t'); ds_put_format (&map, "%s=%s", var_get_short_name (v, 0), var_get_name (v)); } - - write_int (w, 7); /* Record type. */ - write_int (w, 13); /* Record subtype. */ - write_int (w, 1); /* Data item (char) size. */ - write_int (w, ds_length (&map)); /* Number of data items. */ - write_bytes (w, ds_data (&map), ds_length (&map)); - + write_utf8_record (w, dict_get_encoding (dict), &map, 13); ds_destroy (&map); } /* Write integer information record. */ static void -write_integer_info_record (struct sfm_writer *w) +write_integer_info_record (struct sfm_writer *w, + const struct dictionary *d) { + const char *dict_encoding = dict_get_encoding (d); int version_component[3]; int float_format; + int codepage; /* Parse the version string. */ memset (version_component, 0, sizeof version_component); @@ -710,6 +1034,24 @@ write_integer_info_record (struct sfm_writer *w) else abort (); + /* Choose codepage. */ + codepage = sys_get_codepage_from_encoding (dict_encoding); + if (codepage == 0) + { + /* The codepage is unknown. Choose a default. + + For an EBCDIC-compatible encoding, use the value for EBCDIC. + + For an ASCII-compatible encoding, default to "7-bit ASCII", because + many files use this codepage number regardless of their actual + encoding. + */ + if (is_encoding_ascii_compatible (dict_encoding)) + codepage = 2; + else if (is_encoding_ebcdic_compatible (dict_encoding)) + codepage = 1; + } + /* Write record. */ write_int (w, 7); /* Record type. */ write_int (w, 3); /* Record subtype. */ @@ -722,7 +1064,7 @@ write_integer_info_record (struct sfm_writer *w) write_int (w, float_format); write_int (w, 1); /* Compression code. */ write_int (w, INTEGER_NATIVE == INTEGER_MSB_FIRST ? 1 : 2); - write_int (w, 2); /* 7-bit ASCII. */ + write_int (w, codepage); } /* Write floating-point information record. */ @@ -772,7 +1114,7 @@ sys_file_casewriter_destroy (struct casewriter *writer, void *w_) } /* Returns true if an I/O error has occurred on WRITER, false otherwise. */ -bool +static bool write_error (const struct sfm_writer *writer) { return ferror (writer->file); @@ -780,7 +1122,7 @@ write_error (const struct sfm_writer *writer) /* Closes a system file after we're done with it. Returns true if successful, false if an I/O error occurred. */ -bool +static bool close_writer (struct sfm_writer *w) { bool ok; @@ -801,7 +1143,7 @@ close_writer (struct sfm_writer *w) /* Seek back to the beginning and update the number of cases. This is just a courtesy to later readers, so there's no need to check return values or report errors. */ - if (ok && w->case_cnt <= INT32_MAX && !fseek (w->file, 80, SEEK_SET)) + if (ok && w->case_cnt <= INT32_MAX && !fseeko (w->file, 80, SEEK_SET)) { write_int (w, w->case_cnt); clearerr (w->file); @@ -811,7 +1153,7 @@ close_writer (struct sfm_writer *w) ok = false; if (!ok) - msg (ME, _("An I/O error occurred writing system file \"%s\"."), + msg (ME, _("An I/O error occurred writing system file `%s'."), fh_get_file_name (w->fh)); if (ok ? !replace_file_commit (w->rf) : !replace_file_abort (w->rf)) @@ -845,12 +1187,12 @@ write_case_uncompressed (struct sfm_writer *w, const struct ccase *c) { struct sfm_var *v = &w->sfm_vars[i]; - if (v->width == 0) + if (v->var_width == 0) write_float (w, case_num_idx (c, v->case_index)); else { write_bytes (w, case_str_idx (c, v->case_index) + v->offset, - v->width); + v->segment_width); write_spaces (w, v->padding); } } @@ -866,7 +1208,7 @@ write_case_compressed (struct sfm_writer *w, const struct ccase *c) { struct sfm_var *v = &w->sfm_vars[i]; - if (v->width == 0) + if (v->var_width == 0) { double d = case_num_idx (c, v->case_index); if (d == SYSMIS) @@ -890,7 +1232,7 @@ write_case_compressed (struct sfm_writer *w, const struct ccase *c) multiple of 8, by ensuring that the final partial oct (8 byte unit) is treated as padded with spaces on the right. */ - for (width = v->width; width > 0; width -= 8, offset += 8) + for (width = v->segment_width; width > 0; width -= 8, offset += 8) { const void *data = case_str_idx (c, v->case_index) + offset; int chunk_size = MIN (width, 8); @@ -966,7 +1308,7 @@ put_cmp_string (struct sfm_writer *w, const void *data, size_t size) assert (w->data_cnt < 8); assert (size <= 8); - memset (w->data[w->data_cnt], ' ', 8); + memset (w->data[w->data_cnt], w->space, 8); memcpy (w->data[w->data_cnt], data, size); w->data_cnt++; } @@ -1013,14 +1355,14 @@ write_value (struct sfm_writer *w, const union value *value, int width) write_float (w, value->f); else { - write_bytes (w, value->s, width); + write_bytes (w, value_str (value, width), width); write_zeros (w, 8 - width); } } -/* Writes null-terminated STRING in a field of the given WIDTH to - W. If WIDTH is longer than WIDTH, it is truncated; if WIDTH - is narrowed, it is padded on the right with spaces. */ +/* Writes null-terminated STRING in a field of the given WIDTH to W. If STRING + is longer than WIDTH, it is truncated; if STRING is shorter than WIDTH, it + is padded on the right with spaces. */ static void write_string (struct sfm_writer *w, const char *string, size_t width) { @@ -1028,7 +1370,45 @@ write_string (struct sfm_writer *w, const char *string, size_t width) size_t pad_bytes = width - data_bytes; write_bytes (w, string, data_bytes); while (pad_bytes-- > 0) - putc (' ', w->file); + putc (w->space, w->file); +} + +/* Recodes null-terminated UTF-8 encoded STRING into ENCODING, and writes the + recoded version in a field of the given WIDTH to W. The string is truncated + or padded on the right with spaces to exactly WIDTH bytes. */ +static void +write_utf8_string (struct sfm_writer *w, const char *encoding, + const char *string, size_t width) +{ + char *s = recode_string (encoding, "UTF-8", string, -1); + write_string (w, s, width); + free (s); +} + +/* Writes a record with type 7, subtype SUBTYPE that contains CONTENT recoded + from UTF-8 encoded into ENCODING. */ +static void +write_utf8_record (struct sfm_writer *w, const char *encoding, + const struct string *content, int subtype) +{ + struct substring s; + + s = recode_substring_pool (encoding, "UTF-8", ds_ss (content), NULL); + write_string_record (w, s, subtype); + ss_dealloc (&s); +} + +/* Writes a record with type 7, subtype SUBTYPE that contains the string + CONTENT. */ +static void +write_string_record (struct sfm_writer *w, + const struct substring content, int subtype) +{ + write_int (w, 7); + write_int (w, subtype); + write_int (w, 1); + write_int (w, ss_length (content)); + write_bytes (w, ss_data (content), ss_length (content)); } /* Writes SIZE bytes of DATA to W's output file. */ @@ -1051,5 +1431,5 @@ static void write_spaces (struct sfm_writer *w, size_t n) { while (n-- > 0) - putc (' ', w->file); + putc (w->space, w->file); }