X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fdata%2Fsys-file-writer.c;h=c78e04d55f24e41ef5ecce805664c896fc15c44e;hb=128f9567c68a176ab88fd0f54b2398de1f23ba9d;hp=9586cbc2e38e503c75ede6be7ea754bd93d80a11;hpb=fe8dc2171009e90d2335f159d05f7e6660e24780;p=pspp diff --git a/src/data/sys-file-writer.c b/src/data/sys-file-writer.c index 9586cbc2e3..c78e04d55f 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, 2010, 2011 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 @@ -74,18 +74,20 @@ 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. - Compressed data is output as groups of 8 1-byte opcodes - followed by up to 8 (depending on the opcodes) 8-byte data - items. Data items and opcodes arrive at the same time but - must be reordered for writing to disk, thus a small amount - of buffering here. */ - uint8_t opcodes[8]; /* Buffered opcodes. */ - int opcode_cnt; /* Number of buffered opcodes. */ - uint8_t data[8][8]; /* Buffered data. */ - int data_cnt; /* Number of buffered data items. */ + Compressed data is output as a series of 8-byte elements, with 1 to 9 + such elements clustered together. The first element in a cluster is 8 + 1-byte opcodes. Some opcodes call for an additional element in the + cluster (hence, if there are eight such opcodes, then the cluster + contains a full 9 elements). + + cbuf[] holds a cluster at a time. */ + uint8_t cbuf[9][8]; + int n_opcodes; /* Number of opcodes in cbuf[0] so far. */ + int n_elements; /* Number of elements in cbuf[] so far. */ /* Variables. */ struct sfm_var *sfm_vars; /* Variables. */ @@ -115,6 +117,8 @@ static void write_vls_length_table (struct sfm_writer *w, static void write_long_string_value_labels (struct sfm_writer *, const struct dictionary *); +static void write_long_string_missing_values (struct sfm_writer *, + const struct dictionary *); static void write_mrsets (struct sfm_writer *, const struct dictionary *, bool pre_v14); @@ -176,6 +180,7 @@ 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 i; @@ -198,7 +203,8 @@ sfm_open_writer (struct file_handle *fh, struct dictionary *d, w->compress = opts.compress; w->case_cnt = 0; - w->opcode_cnt = w->data_cnt = 0; + w->n_opcodes = w->n_elements = 0; + memset (w->cbuf[0], 0, 8); /* Figure out how to map in-memory case data to on-disk case data. Also count the number of segments. Very long strings @@ -227,6 +233,9 @@ sfm_open_writer (struct file_handle *fh, struct dictionary *d, 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); @@ -253,10 +262,14 @@ sfm_open_writer (struct file_handle *fh, struct dictionary *d, write_vls_length_table (w, d); write_long_string_value_labels (w, d); + write_long_string_missing_values (w, d); - if (attrset_count (dict_get_attributes (d))) - write_data_file_attributes (w, d); - write_variable_attributes (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); @@ -311,6 +324,7 @@ 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]; @@ -320,12 +334,15 @@ 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); @@ -374,14 +391,14 @@ 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_utf8_string (w, dict_get_encoding (d), file_label, 64); + write_utf8_string (w, dict_encoding, file_label, 64); /* Padding. */ write_zeros (w, 3); @@ -430,7 +447,6 @@ write_variable (struct sfm_writer *w, const struct variable *v) int segment_cnt = sfm_width_to_segments (width); int seg0_width = sfm_segment_alloc_width (width, 0); const char *encoding = var_get_encoding (v); - struct missing_values mv; int i; /* Record type. */ @@ -444,14 +460,20 @@ 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. */ - 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)); + negated. + + Missing values for long string variables are written in a separate + record. */ + if (width <= MAX_SHORT_STRING) + { + const struct missing_values *mv = var_get_missing_values (v); + if (mv_has_range (mv)) + write_int (w, -2 - mv_n_values (mv)); + else + write_int (w, mv_n_values (mv)); + } else - write_int (w, mv_n_values (&mv)); + write_int (w, 0); /* Print and write formats. */ write_format (w, *var_get_print_format (v), seg0_width); @@ -474,15 +496,19 @@ write_variable (struct sfm_writer *w, const struct variable *v) } /* Write the missing values, if any, range first. */ - if (mv_has_range (&mv)) + if (width <= MAX_SHORT_STRING) { - double x, y; - mv_get_range (&mv, &x, &y); - write_float (w, x); - write_float (w, y); + const struct missing_values *mv = var_get_missing_values (v); + if (mv_has_range (mv)) + { + double 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++) + write_value (w, mv_get_value (mv, i), 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); @@ -502,8 +528,6 @@ write_variable (struct sfm_writer *w, const struct variable *v) write_variable_continuation_records (w, seg_width); } - - mv_destroy (&mv); } /* Writes the value labels to system file W. @@ -671,6 +695,46 @@ write_data_file_attributes (struct sfm_writer *w, 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_TARGET: + 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) { @@ -682,14 +746,16 @@ 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)) - { - if (n_attrsets++) - ds_put_byte (&s, '/'); - ds_put_format (&s, "%s:", var_get_name (v)); - put_attrset (&s, 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); @@ -708,6 +774,12 @@ write_mrsets (struct sfm_writer *w, const struct dictionary *dict, 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; @@ -759,11 +831,12 @@ write_mrsets (struct sfm_writer *w, const struct dictionary *dict, 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", - short_name_utf8, -1); - str_lowercase (short_name); + lower_name_utf8, -1); ds_put_format (&s, " %s", short_name); free (short_name); + free (lower_name_utf8); } ds_put_byte (&s, '\n'); } @@ -830,11 +903,11 @@ write_vls_length_table (struct sfm_writer *w, ds_destroy (&map); } - static void write_long_string_value_labels (struct sfm_writer *w, const struct dictionary *dict) { + const char *encoding = dict_get_encoding (dict); size_t n_vars = dict_get_var_cnt (dict); size_t size, i; off_t start UNUSED; @@ -845,7 +918,6 @@ write_long_string_value_labels (struct sfm_writer *w, { 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; @@ -875,7 +947,6 @@ write_long_string_value_labels (struct sfm_writer *w, { 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; @@ -911,6 +982,71 @@ write_long_string_value_labels (struct sfm_writer *w, assert (ftello (w->file) == start + size); } +static void +write_long_string_missing_values (struct sfm_writer *w, + const struct dictionary *dict) +{ + const char *encoding = dict_get_encoding (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++) + { + struct variable *var = dict_get_var (dict, i); + const struct missing_values *mv = var_get_missing_values (var); + int width = var_get_width (var); + + if (mv_is_empty (mv) || width < 9) + continue; + + size += 4; + size += recode_string_len (encoding, "UTF-8", var_get_name (var), -1); + size += 1; + size += mv_n_values (mv) * (4 + 8); + } + if (size == 0) + return; + + write_int (w, 7); /* Record type. */ + write_int (w, 22); /* 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 missing_values *mv = var_get_missing_values (var); + int width = var_get_width (var); + uint8_t n_missing_values; + char *var_name; + int j; + + if (mv_is_empty (mv) || 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); + + n_missing_values = mv_n_values (mv); + write_bytes (w, &n_missing_values, 1); + + for (j = 0; j < n_missing_values; j++) + { + const union value *value = mv_get_value (mv, j); + + write_int (w, 8); + write_bytes (w, value_str (value, width), 8); + } + } + assert (ftello (w->file) == start + size); +} + static void write_encoding_record (struct sfm_writer *w, const struct dictionary *d) @@ -952,6 +1088,7 @@ static void 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; @@ -973,13 +1110,21 @@ write_integer_info_record (struct sfm_writer *w, abort (); /* Choose codepage. */ - codepage = sys_get_codepage_from_encoding (dict_get_encoding (d)); + codepage = sys_get_codepage_from_encoding (dict_encoding); if (codepage == 0) { - /* Default to "7-bit ASCII" if the codepage number is unknown, because + /* 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. */ - codepage = 2; + encoding. + */ + if (is_encoding_ascii_compatible (dict_encoding)) + codepage = 2; + else if (is_encoding_ebcdic_compatible (dict_encoding)) + codepage = 1; } /* Write record. */ @@ -1064,8 +1209,7 @@ close_writer (struct sfm_writer *w) if (w->file != NULL) { /* Flush buffer. */ - if (w->opcode_cnt > 0) - flush_compressed (w); + flush_compressed (w); fflush (w->file); ok = !write_error (w); @@ -1148,10 +1292,7 @@ write_case_compressed (struct sfm_writer *w, const struct ccase *c) && d == (int) d) put_cmp_opcode (w, (int) d + COMPRESSION_BIAS); else - { - put_cmp_opcode (w, 253); - put_cmp_number (w, d); - } + put_cmp_number (w, d); } else { @@ -1169,10 +1310,7 @@ write_case_compressed (struct sfm_writer *w, const struct ccase *c) if (!memcmp (data, " ", chunk_size)) put_cmp_opcode (w, 254); else - { - put_cmp_opcode (w, 253); - put_cmp_string (w, data, chunk_size); - } + put_cmp_string (w, data, chunk_size); } /* This code deals properly with padding that is not a @@ -1186,19 +1324,16 @@ write_case_compressed (struct sfm_writer *w, const struct ccase *c) } } -/* Flushes buffered compressed opcodes and data to W. - The compression buffer must not be empty. */ +/* Flushes buffered compressed opcodes and data to W. */ static void flush_compressed (struct sfm_writer *w) { - assert (w->opcode_cnt > 0 && w->opcode_cnt <= 8); - - write_bytes (w, w->opcodes, w->opcode_cnt); - write_zeros (w, 8 - w->opcode_cnt); - - write_bytes (w, w->data, w->data_cnt * sizeof *w->data); - - w->opcode_cnt = w->data_cnt = 0; + if (w->n_opcodes) + { + write_bytes (w, w->cbuf, 8 * (1 + w->n_elements)); + w->n_opcodes = w->n_elements = 0; + memset (w->cbuf[0], 0, 8); + } } /* Appends OPCODE to the buffered set of compression opcodes in @@ -1206,41 +1341,32 @@ flush_compressed (struct sfm_writer *w) static void put_cmp_opcode (struct sfm_writer *w, uint8_t opcode) { - if (w->opcode_cnt >= 8) + if (w->n_opcodes >= 8) flush_compressed (w); - w->opcodes[w->opcode_cnt++] = opcode; + w->cbuf[0][w->n_opcodes++] = opcode; } -/* Appends NUMBER to the buffered compression data in W. The - buffer must not be full; the way to assure that is to call - this function only just after a call to put_cmp_opcode, which - will flush the buffer as necessary. */ +/* Appends NUMBER to the buffered compression data in W. */ static void put_cmp_number (struct sfm_writer *w, double number) { - assert (w->opcode_cnt > 0); - assert (w->data_cnt < 8); - - convert_double_to_output_format (number, w->data[w->data_cnt++]); + put_cmp_opcode (w, 253); + convert_double_to_output_format (number, w->cbuf[++w->n_elements]); } /* Appends SIZE bytes of DATA to the buffered compression data in W, followed by enough spaces to pad the output data to exactly - 8 bytes (thus, SIZE must be no greater than 8). The buffer - must not be full; the way to assure that is to call this - function only just after a call to put_cmp_opcode, which will - flush the buffer as necessary. */ + 8 bytes (thus, SIZE must be no greater than 8). */ static void put_cmp_string (struct sfm_writer *w, const void *data, size_t size) { - assert (w->opcode_cnt > 0); - assert (w->data_cnt < 8); assert (size <= 8); - memset (w->data[w->data_cnt], ' ', 8); - memcpy (w->data[w->data_cnt], data, size); - w->data_cnt++; + put_cmp_opcode (w, 253); + w->n_elements++; + memset (w->cbuf[w->n_elements], w->space, 8); + memcpy (w->cbuf[w->n_elements], data, size); } /* Writes 32-bit integer X to the output file for writer W. */ @@ -1300,7 +1426,7 @@ 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 @@ -1361,5 +1487,5 @@ static void write_spaces (struct sfm_writer *w, size_t n) { while (n-- > 0) - putc (' ', w->file); + putc (w->space, w->file); }