/* PSPP - computes sample statistics.
- Copyright (C) 1997-9, 2000 Free Software Foundation, Inc.
- Written by Ben Pfaff <blp@gnu.org>.
+ Copyright (C) 1997-9, 2000, 2006 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
02110-1301, USA. */
#include <config.h>
+
#include "sys-file-writer.h"
-#include "sfm-private.h"
-#include <libpspp/message.h>
-#include <stdlib.h>
+#include "sys-file-private.h"
+
#include <ctype.h>
#include <errno.h>
#include <fcntl.h>
+#include <stdlib.h>
#include <sys/stat.h>
#include <time.h>
#include <unistd.h>
+
#include <libpspp/alloc.h>
-#include "case.h"
-#include "dictionary.h"
-#include <libpspp/message.h>
-#include "file-handle-def.h"
#include <libpspp/hash.h>
#include <libpspp/magic.h>
+#include <libpspp/message.h>
#include <libpspp/misc.h>
-#include "settings.h"
-#include "stat-macros.h"
#include <libpspp/str.h>
-#include "value-labels.h"
-#include "variable.h"
#include <libpspp/version.h>
-#include <minmax.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/format.h>
+#include <data/missing-values.h>
+#include <data/settings.h>
+#include <data/value-labels.h>
+#include <data/variable.h>
+
+#include "minmax.h"
#include "gettext.h"
#define _(msgid) gettext (msgid)
+/* Find 64-bit floating-point type. */
+#if SIZEOF_FLOAT == 8
+ #define flt64 float
+ #define FLT64_MAX FLT_MAX
+#elif SIZEOF_DOUBLE == 8
+ #define flt64 double
+ #define FLT64_MAX DBL_MAX
+#elif SIZEOF_LONG_DOUBLE == 8
+ #define flt64 long double
+ #define FLT64_MAX LDBL_MAX
+#else
+ #error Which one of your basic types is 64-bit floating point?
+#endif
+
+/* Figure out SYSMIS value for flt64. */
+#include <libpspp/magic.h>
+#if SIZEOF_DOUBLE == 8
+#define second_lowest_flt64 second_lowest_value
+#else
+#error Must define second_lowest_flt64 for your architecture.
+#endif
+
+/* Record Type 1: General Information. */
+struct sysfile_header
+ {
+ char rec_type[4] ; /* 00: Record-type code, "$FL2". */
+ char prod_name[60] ; /* 04: Product identification. */
+ int32_t layout_code ; /* 40: 2. */
+ int32_t nominal_case_size ; /* 44: Number of `value's per case.
+ Note: some systems set this to -1 */
+ int32_t compress ; /* 48: 1=compressed, 0=not compressed. */
+ int32_t weight_idx ; /* 4c: 1-based index of weighting var, or 0. */
+ int32_t case_cnt ; /* 50: Number of cases, -1 if unknown. */
+ flt64 bias ; /* 54: Compression bias (100.0). */
+ char creation_date[9] ; /* 5c: `dd mmm yy' creation date of file. */
+ char creation_time[8] ; /* 65: `hh:mm:ss' 24-hour creation time. */
+ char file_label[64] ; /* 6d: File label. */
+ char padding[3] ; /* ad: Ignored padding. */
+ } ATTRIBUTE((packed)) ;
+
+/* Record Type 2: Variable. */
+struct sysfile_variable
+ {
+ int32_t rec_type ; /* 2. */
+ int32_t type ; /* 0=numeric, 1-255=string width,
+ -1=continued string. */
+ int32_t has_var_label ; /* 1=has a variable label, 0=doesn't. */
+ int32_t n_missing_values ; /* Missing value code of -3,-2,0,1,2, or 3. */
+ int32_t print ; /* Print format. */
+ int32_t write ; /* Write format. */
+ char name[SHORT_NAME_LEN] ; /* Variable name. */
+ /* The rest of the structure varies. */
+ } ATTRIBUTE((packed)) ;
+
/* Compression bias used by PSPP. Values between (1 -
COMPRESSION_BIAS) and (251 - COMPRESSION_BIAS) inclusive can be
compressed. */
size_t flt64_cnt; /* Number of flt64 elements. */
};
+static struct casewriter_class sys_file_casewriter_class;
+
static char *append_string_max (char *, const char *, const char *);
static void write_header (struct sfm_writer *, const struct dictionary *);
static void buf_write (struct sfm_writer *, const void *, size_t);
static void write_documents (struct sfm_writer *, const struct dictionary *);
+bool write_error (const struct sfm_writer *);
+bool close_writer (struct sfm_writer *);
+
static inline int
var_flt64_cnt (const struct variable *v)
{
assert(sizeof(flt64) == MAX_SHORT_STRING);
- return width_to_bytes(v->width) / MAX_SHORT_STRING ;
+ return sfm_width_to_bytes(var_get_width (v)) / MAX_SHORT_STRING ;
}
static inline int
var_flt64_cnt_nom (const struct variable *v)
{
- return v->type == NUMERIC ? 1 : DIV_RND_UP (v->width, sizeof (flt64));
+ return (var_is_numeric (v)
+ ? 1 : DIV_RND_UP (var_get_width (v), sizeof (flt64)));
}
No reference to D is retained, so it may be modified or
destroyed at will after this function returns. D is not
modified by this function, except to assign short names. */
-struct sfm_writer *
+struct casewriter *
sfm_open_writer (struct file_handle *fh, struct dictionary *d,
struct sfm_write_options opts)
{
{
const struct variable *dv = dict_get_var (d, i);
struct sfm_var *sv = &w->vars[i];
- sv->width = dv->width;
+ sv->width = var_get_width (dv);
/* spss compatibility nonsense */
- if ( dv->width > MAX_LONG_STRING )
+ if ( var_get_width (dv) >= MIN_VERY_LONG_STRING )
w->has_vls = true;
- sv->fv = dv->fv;
+ sv->fv = var_get_case_index (dv);
sv->flt64_cnt = var_flt64_cnt (dv);
}
{
int count = 0;
const struct variable *v = dict_get_var(d, i);
- int wcount = v->width;
+ int wcount = var_get_width (v);
do {
- struct variable var_cont = *v;
- if ( v->type == ALPHA)
+ struct variable *var_cont = var_clone (v);
+ var_set_short_name (var_cont, var_get_short_name (v));
+ if ( var_is_alpha (v))
{
if ( 0 != count )
{
- mv_init(&var_cont.miss, 0);
- strcpy(var_cont.short_name,
- cont_var_name(v->short_name, count));
- var_cont.label = NULL;
+ var_clear_missing_values (var_cont);
+ var_set_short_name (var_cont,
+ cont_var_name (var_get_short_name (v),
+ count));
+ var_clear_label (var_cont);
w->var_cnt_vls++;
}
count++;
- if ( wcount > MAX_LONG_STRING )
+ if ( wcount >= MIN_VERY_LONG_STRING )
{
- var_cont.width = MAX_LONG_STRING;
+ var_set_width (var_cont, MIN_VERY_LONG_STRING - 1);
wcount -= EFFECTIVE_LONG_STRING_LENGTH;
}
else
{
- var_cont.width = wcount;
- wcount -= var_cont.width;
+ var_set_width (var_cont, wcount);
+ wcount -= var_get_width (var_cont);
}
-
- var_cont.write.w = var_cont.width;
- var_cont.print.w = var_cont.width;
}
- write_variable (w, &var_cont);
+ write_variable (w, var_cont);
+ var_destroy (var_cont);
} while(wcount > 0);
}
w->y = (unsigned char *) w->ptr;
}
- if (sfm_write_error (w))
+ if (write_error (w))
goto error;
- return w;
+ return casewriter_create (&sys_file_casewriter_class, w);
error:
- sfm_close_writer (w);
+ close_writer (w);
return NULL;
open_error:
if (dict_get_weight (d) != NULL)
{
- struct variable *weight_var;
+ const struct variable *weight_var;
int recalc_weight_idx = 1;
int i;
static inline void
write_format_spec (const struct fmt_spec *src, int32_t *dest)
{
- assert(check_output_specifier(src, true));
- *dest = (formats[src->type].spss << 16) | (src->w << 8) | src->d;
+ assert (fmt_check_output (src));
+ *dest = (fmt_to_io (src->type) << 16) | (src->w << 8) | src->d;
}
/* Write the variable record(s) for primary variable P and secondary
struct missing_values mv;
flt64 m[3]; /* Missing value values. */
int nm; /* Number of missing values, possibly negative. */
+ const char *label = var_get_label (v);
sv.rec_type = 2;
- sv.type = min(v->width, MAX_LONG_STRING);
- sv.has_var_label = (v->label != NULL);
+ sv.type = MIN (var_get_width (v), MIN_VERY_LONG_STRING - 1);
+ sv.has_var_label = label != NULL;
- mv_copy (&mv, &v->miss);
+ mv_copy (&mv, var_get_missing_values (v));
nm = 0;
if (mv_has_range (&mv))
{
{
union value value;
mv_pop_value (&mv, &value);
- if (v->type == NUMERIC)
+ if (var_is_numeric (v))
m[nm] = value.f;
else
- buf_copy_rpad ((char *) &m[nm], sizeof m[nm], value.s, v->width);
+ buf_copy_rpad ((char *) &m[nm], sizeof m[nm], value.s,
+ var_get_width (v));
nm++;
}
- if (mv_has_range (&v->miss))
+ if (mv_has_range (var_get_missing_values (v)))
nm = -nm;
sv.n_missing_values = nm;
- write_format_spec (&v->print, &sv.print);
- write_format_spec (&v->write, &sv.write);
- buf_copy_str_rpad (sv.name, sizeof sv.name, v->short_name);
+ write_format_spec (var_get_print_format (v), &sv.print);
+ write_format_spec (var_get_write_format (v), &sv.write);
+ buf_copy_str_rpad (sv.name, sizeof sv.name, var_get_short_name (v));
buf_write (w, &sv, sizeof sv);
- if (v->label)
+ if (label != NULL)
{
struct label
{
int ext_len;
- l.label_len = min (strlen (v->label), 255);
+ l.label_len = MIN (strlen (label), 255);
ext_len = ROUND_UP (l.label_len, sizeof l.label_len);
- memcpy (l.label, v->label, l.label_len);
+ memcpy (l.label, label, l.label_len);
memset (&l.label[l.label_len], ' ', ext_len - l.label_len);
buf_write (w, &l, offsetof (struct label, label) + ext_len);
if (nm)
buf_write (w, m, sizeof *m * abs (nm));
- if (v->type == ALPHA && v->width > (int) sizeof (flt64))
+ if (var_is_alpha (v) && var_get_width (v) > (int) sizeof (flt64))
{
int i;
int pad_count;
memset (&sv.write, 0, sizeof sv.write);
memset (&sv.name, 0, sizeof sv.name);
- pad_count = DIV_RND_UP (min(v->width, MAX_LONG_STRING),
+ pad_count = DIV_RND_UP (MIN(var_get_width (v), MIN_VERY_LONG_STRING - 1),
(int) sizeof (flt64)) - 1;
for (i = 0; i < pad_count; i++)
buf_write (w, &sv, sizeof sv);
int32_t vars[1] ;
} ATTRIBUTE((packed));
+ const struct val_labs *val_labs;
struct val_labs_iterator *i;
struct value_label_rec *vlr;
struct var_idx_rec vir;
size_t vlr_size;
flt64 *loc;
- if (!val_labs_count (v->val_labs))
+ val_labs = var_get_value_labels (v);
+ if (val_labs == NULL)
return;
/* Pass 1: Count bytes. */
vlr_size = (sizeof (struct value_label_rec)
- + sizeof (flt64) * (val_labs_count (v->val_labs) - 1));
- for (vl = val_labs_first (v->val_labs, &i); vl != NULL;
- vl = val_labs_next (v->val_labs, &i))
+ + sizeof (flt64) * (val_labs_count (val_labs) - 1));
+ for (vl = val_labs_first (val_labs, &i); vl != NULL;
+ vl = val_labs_next (val_labs, &i))
vlr_size += ROUND_UP (strlen (vl->label) + 1, sizeof (flt64));
/* Pass 2: Copy bytes. */
vlr = xmalloc (vlr_size);
vlr->rec_type = 3;
- vlr->n_labels = val_labs_count (v->val_labs);
+ vlr->n_labels = val_labs_count (val_labs);
loc = vlr->labels;
- for (vl = val_labs_first_sorted (v->val_labs, &i); vl != NULL;
- vl = val_labs_next (v->val_labs, &i))
+ for (vl = val_labs_first_sorted (val_labs, &i); vl != NULL;
+ vl = val_labs_next (val_labs, &i))
{
size_t len = strlen (vl->label);
int32_t n_lines ; /* Number of lines of documents. */
} ATTRIBUTE((packed)) rec_6;
- const char *documents;
- size_t n_lines;
+ const char * documents = dict_get_documents (d);
+ size_t doc_bytes = strlen (documents);
- documents = dict_get_documents (d);
- n_lines = strlen (documents) / 80;
+ assert (doc_bytes % 80 == 0);
rec_6.rec_type = 6;
- rec_6.n_lines = n_lines;
+ rec_6.n_lines = doc_bytes / 80;
buf_write (w, &rec_6, sizeof rec_6);
- buf_write (w, documents, 80 * n_lines);
+ buf_write (w, documents, 80 * rec_6.n_lines);
}
/* Write the alignment, width and scale values */
v = dict_get_var(dict, i);
- params.measure = v->measure;
- params.width = v->display_width;
- params.align = v->alignment;
+ params.measure = (var_get_measure (v) == MEASURE_NOMINAL ? 1
+ : var_get_measure (v) == MEASURE_ORDINAL ? 2
+ : 3);
+ params.width = var_get_display_width (v);
+ params.align = (var_get_alignment (v) == ALIGN_LEFT ? 0
+ : var_get_alignment (v) == ALIGN_RIGHT ? 1
+ : 2);
buf_write (w, ¶ms, sizeof(params));
- if ( v->width > MAX_LONG_STRING )
+ if (var_is_long_string (v))
{
- int wcount = v->width - EFFECTIVE_LONG_STRING_LENGTH ;
+ int wcount = var_get_width (v) - EFFECTIVE_LONG_STRING_LENGTH ;
while (wcount > 0)
{
- params.width = wcount > MAX_LONG_STRING ? 32 : wcount;
+ params.width = wcount >= MIN_VERY_LONG_STRING ? 32 : wcount;
buf_write (w, ¶ms, sizeof(params));
{
const struct variable *v = dict_get_var (dict, i);
- if ( v->width <= MAX_LONG_STRING )
+ if ( var_get_width (v) < MIN_VERY_LONG_STRING )
continue;
- ds_put_format (&vls_length_map, "%s=%05d", v->short_name, v->width);
+ ds_put_format (&vls_length_map, "%s=%05d",
+ var_get_short_name (v), var_get_width (v));
ds_put_char (&vls_length_map, '\0');
ds_put_char (&vls_length_map, '\t');
}
if (i)
ds_put_char (&long_name_map, '\t');
- ds_put_format (&long_name_map, "%s=%s", v->short_name, v->name);
+ ds_put_format (&long_name_map, "%s=%s",
+ var_get_short_name (v), var_get_name (v));
}
lv_hdr.rec_type = 7;
static char *
append_string_max (char *dest, const char *src, const char *end)
{
- int nbytes = min (end - dest, (int) strlen (src));
+ int nbytes = MIN (end - dest, (int) strlen (src));
memcpy (dest, src, nbytes);
return dest + nbytes;
}
static void write_compressed_data (struct sfm_writer *w, const flt64 *elem);
-/* Writes case C to system file W.
- Returns 1 if successful, 0 if an I/O error occurred. */
-int
-sfm_write_case (struct sfm_writer *w, const struct ccase *c)
+/* Writes case C to system file W. */
+static void
+sys_file_casewriter_write (struct casewriter *writer, void *w_,
+ struct ccase *c)
{
- if (ferror (w->file))
- return 0;
+ struct sfm_writer *w = w_;
+ if (ferror (w->file))
+ {
+ casewriter_force_error (writer);
+ case_destroy (c);
+ return;
+ }
w->case_cnt++;
if (v->width == 0)
{
- *bounce_cur = case_num (c, v->fv);
+ *bounce_cur = case_num_idx (c, v->fv);
bounce_cur += v->flt64_cnt;
}
else
{ int ofs = 0;
while (ofs < v->width)
{
- int chunk = MIN (MAX_LONG_STRING, v->width - ofs);
+ int chunk = MIN (MIN_VERY_LONG_STRING - 1, v->width - ofs);
int nv = DIV_RND_UP (chunk, sizeof (flt64));
buf_copy_rpad ((char *) bounce_cur, nv * sizeof (flt64),
- case_data (c, v->fv)->s + ofs, chunk);
+ case_data_idx (c, v->fv)->s + ofs, chunk);
bounce_cur += nv;
ofs += chunk;
}
local_free (bounce);
}
-
- return !sfm_write_error (w);
+
+ case_destroy (c);
+}
+
+static void
+sys_file_casewriter_destroy (struct casewriter *writer, void *w_)
+{
+ struct sfm_writer *w = w_;
+ if (!close_writer (w))
+ casewriter_force_error (writer);
}
static void
/* Returns true if an I/O error has occurred on WRITER, false otherwise. */
bool
-sfm_write_error (const struct sfm_writer *writer)
+write_error (const struct sfm_writer *writer)
{
return ferror (writer->file);
}
/* Closes a system file after we're done with it.
Returns true if successful, false if an I/O error occurred. */
bool
-sfm_close_writer (struct sfm_writer *w)
+close_writer (struct sfm_writer *w)
{
bool ok;
}
fflush (w->file);
- ok = !sfm_write_error (w);
+ ok = !write_error (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
return ok;
}
+\f
+static struct casewriter_class sys_file_casewriter_class =
+ {
+ sys_file_casewriter_write,
+ sys_file_casewriter_destroy,
+ NULL,
+ };
projects / pspp / blobdiff