/* PSPP - computes sample statistics.
Copyright (C) 1997-9, 2000, 2006 Free Software Foundation, Inc.
- Written by Ben Pfaff <blp@gnu.org>.
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
#include <data/case.h>
#include <data/casefile.h>
#include <data/fastfile.h>
+#include <data/format.h>
#include <data/dictionary.h>
#include <data/por-file-writer.h>
#include <data/procedure.h>
}
}
-/* Clears internal state related to case reader input procedure. */
-static void
-case_reader_source_destroy (struct case_source *source)
+/* Reads one case into C.
+ Returns true if successful, false at end of file or if an
+ I/O error occurred. */
+static bool
+case_reader_source_read (struct case_source *source, struct ccase *c)
{
struct case_reader_pgm *pgm = source->aux;
- case_reader_pgm_free (pgm);
+ if (any_reader_read (pgm->reader, pgm->map == NULL ? c : &pgm->bounce))
+ {
+ if (pgm->map != NULL)
+ map_case (pgm->map, &pgm->bounce, c);
+ return true;
+ }
+ else
+ return false;
}
-/* Reads all the cases from the data file into C and passes them
- to WRITE_CASE one by one, passing WC_DATA.
- Returns true if successful, false if an I/O error occurred. */
+/* Destroys the source.
+ Returns true if successful read, false if an I/O occurred
+ during destruction or previously. */
static bool
-case_reader_source_read (struct case_source *source,
- struct ccase *c,
- write_case_func *write_case, write_case_data wc_data)
+case_reader_source_destroy (struct case_source *source)
{
struct case_reader_pgm *pgm = source->aux;
- bool ok = true;
-
- do
- {
- bool got_case;
- if (pgm->map == NULL)
- got_case = any_reader_read (pgm->reader, c);
- else
- {
- got_case = any_reader_read (pgm->reader, &pgm->bounce);
- if (got_case)
- map_case (pgm->map, &pgm->bounce, c);
- }
- if (!got_case)
- break;
-
- ok = write_case (wc_data);
- }
- while (ok);
-
- return ok && !any_reader_error (pgm->reader);
+ bool ok = !any_reader_error (pgm->reader);
+ case_reader_pgm_free (pgm);
+ return ok;
}
static const struct case_source_class case_reader_source_class =
\f
/* SAVE and EXPORT. */
-static bool output_proc (const struct ccase *, void *, const struct dataset *);
-
/* Parses and performs the SAVE or EXPORT procedure. */
static int
parse_output_proc (struct lexer *lexer, struct dataset *ds, enum writer_type writer_type)
bool retain_unselected;
struct variable *saved_filter_variable;
struct case_writer *aw;
- bool ok;
+ struct ccase *c;
+ bool ok = true;
aw = parse_write_command (lexer, ds, writer_type, PROC_CMD, &retain_unselected);
if (aw == NULL)
saved_filter_variable = dict_get_filter (dataset_dict (ds));
if (retain_unselected)
dict_set_filter (dataset_dict (ds), NULL);
- ok = procedure (ds, output_proc, aw);
+
+ proc_open (ds);
+ while (ok && proc_read (ds, &c))
+ ok = case_writer_write_case (aw, c);
+ ok = proc_close (ds) && ok;
+
dict_set_filter (dataset_dict (ds), saved_filter_variable);
case_writer_destroy (aw);
return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
}
-/* Writes case C to file. */
-static bool
-output_proc (const struct ccase *c, void *aw_, const struct dataset *ds UNUSED)
-{
- struct case_writer *aw = aw_;
- return case_writer_write_case (aw, c);
-}
-
int
cmd_save (struct lexer *lexer, struct dataset *ds)
{
char *in_name; /* Variable name. */
struct variable *in_var; /* Variable (in master dictionary). */
- struct ccase input; /* Input record. */
+ struct ccase input_storage; /* Input record storage. */
+ struct ccase *input; /* Input record. */
};
/* MATCH FILES procedure. */
static bool mtf_free (struct mtf_proc *);
static bool mtf_close_file (struct mtf_file *);
static int mtf_merge_dictionary (struct dictionary *const, struct mtf_file *);
+static bool mtf_read_records (struct mtf_proc *, struct dataset *);
static bool mtf_delete_file_in_place (struct mtf_proc *, struct mtf_file **);
-static bool mtf_read_nonactive_records (void *);
-static bool mtf_processing_finish (void *, const struct dataset *);
-static bool mtf_processing (const struct ccase *, void *, const struct dataset *);
+static bool mtf_processing (struct mtf_proc *, struct dataset *);
static char *var_type_description (struct variable *);
bool saw_table = false;
bool saw_in = false;
- bool ok;
-
mtf.head = mtf.tail = NULL;
mtf.by_cnt = 0;
mtf.first[0] = '\0';
file->dict = NULL;
file->in_name = NULL;
file->in_var = NULL;
- case_nullify (&file->input);
+ case_nullify (&file->input_storage);
+ file->input = &file->input_storage;
/* FILEs go first, then TABLEs. */
if (file->type == MTF_TABLE || first_table == NULL)
if (file->reader == NULL)
goto error;
- case_create (&file->input, dict_get_next_value_idx (file->dict));
+ case_create (&file->input_storage,
+ dict_get_next_value_idx (file->dict));
}
while (lex_match (lexer, '/'))
7. Repeat from step 2.
- Unfortunately, this algorithm can't be implemented in a
- straightforward way because there's no function to read a
- record from the active file. Instead, it has to be written
- as a state machine.
-
FIXME: For merging large numbers of files (more than 10?) a
better algorithm would use a heap for finding minimum
values. */
- if (!used_active_file)
+ if (used_active_file)
+ {
+ proc_set_sink (ds, create_case_sink (&null_sink_class,
+ dataset_dict (ds),
+ dataset_get_casefile_factory (ds),
+ NULL));
+ proc_open (ds);
+ }
+ else
discard_variables (ds);
dict_compact_values (mtf.dict);
- mtf.output = fastfile_create (dict_get_next_value_idx (mtf.dict));
+ mtf.output = dataset_get_casefile_factory (ds)->create_casefile
+ (dataset_get_casefile_factory (ds),
+ dict_get_next_value_idx (mtf.dict));
+
mtf.seq_nums = xcalloc (dict_get_var_cnt (mtf.dict), sizeof *mtf.seq_nums);
case_create (&mtf.mtf_case, dict_get_next_value_idx (mtf.dict));
- if (!mtf_read_nonactive_records (&mtf))
+ if (!mtf_read_records (&mtf, ds))
+ goto error;
+ while (mtf.head && mtf.head->type == MTF_FILE)
+ if (!mtf_processing (&mtf, ds))
+ goto error;
+ if (!proc_close (ds))
goto error;
-
- if (used_active_file)
- {
- proc_set_sink (ds,
- create_case_sink (&null_sink_class,
- dataset_dict (ds), NULL));
- ok =
- ( procedure (ds, mtf_processing, &mtf) &&
- mtf_processing_finish (&mtf, ds) );
- }
- else
- ok = mtf_processing_finish (&mtf, ds);
discard_variables (ds);
proc_set_source (ds, storage_source_create (mtf.output));
mtf.output = NULL;
- if (!mtf_free (&mtf))
- ok = false;
- return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
+ return mtf_free (&mtf) ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
error:
+ proc_close (ds);
mtf_free (&mtf);
return CMD_CASCADING_FAILURE;
}
-/* Repeats 2...7 an arbitrary number of times. */
-static bool
-mtf_processing_finish (void *mtf_, const struct dataset *ds)
-{
- struct mtf_proc *mtf = mtf_;
- struct mtf_file *iter;
-
- /* Find the active file and delete it. */
- for (iter = mtf->head; iter; iter = iter->next)
- if (iter->handle == NULL)
- {
- if (!mtf_delete_file_in_place (mtf, &iter))
- NOT_REACHED ();
- break;
- }
-
- while (mtf->head && mtf->head->type == MTF_FILE)
- if (!mtf_processing (NULL, mtf, ds))
- return false;
-
- return true;
-}
-
/* Return a string in a static buffer describing V's variable type and
width. */
static char *
any_reader_close (file->reader);
if (file->handle != NULL)
dict_destroy (file->dict);
- case_destroy (&file->input);
+ case_destroy (&file->input_storage);
free (file->in_name);
free (file);
return ok;
*file = f->next;
if (f->in_var != NULL)
- case_data_rw (&mtf->mtf_case, f->in_var->fv)->f = 0.;
+ case_data_rw (&mtf->mtf_case, f->in_var)->f = 0.;
for (i = 0; i < dict_get_var_cnt (f->dict); i++)
{
struct variable *v = dict_get_var (f->dict, i);
struct variable *mv = get_master (v);
if (mv != NULL)
{
- union value *out = case_data_rw (&mtf->mtf_case, mv->fv);
+ union value *out = case_data_rw (&mtf->mtf_case, mv);
if (var_is_numeric (v))
out->f = SYSMIS;
return mtf_close_file (f);
}
-/* Read a record from every input file except the active file.
+/* Read a record from every input file.
Returns true if successful, false if an I/O error occurred. */
static bool
-mtf_read_nonactive_records (void *mtf_)
+mtf_read_records (struct mtf_proc *mtf, struct dataset *ds)
{
- struct mtf_proc *mtf = mtf_;
struct mtf_file *iter, *next;
bool ok = true;
for (iter = mtf->head; ok && iter != NULL; iter = next)
{
next = iter->next;
- if (iter->handle && !any_reader_read (iter->reader, &iter->input))
- if (!mtf_delete_file_in_place (mtf, &iter))
- ok = false;
+ if (iter->handle
+ ? !any_reader_read (iter->reader, iter->input)
+ : !proc_read (ds, &iter->input))
+ {
+ if (!mtf_delete_file_in_place (mtf, &iter))
+ ok = false;
+ }
}
return ok;
}
if A == B, 1 if A > B. */
static inline int
mtf_compare_BY_values (struct mtf_proc *mtf,
- struct mtf_file *a, struct mtf_file *b,
- const struct ccase *c)
+ struct mtf_file *a, struct mtf_file *b)
{
- const struct ccase *ca = case_is_null (&a->input) ? c : &a->input;
- const struct ccase *cb = case_is_null (&b->input) ? c : &b->input;
- assert ((a == NULL) + (b == NULL) + (c == NULL) <= 1);
- return case_compare_2dict (ca, cb, a->by, b->by, mtf->by_cnt);
+ return case_compare_2dict (a->input, b->input, a->by, b->by, mtf->by_cnt);
}
/* Perform one iteration of steps 3...7 above.
Returns true if successful, false if an I/O error occurred. */
static bool
-mtf_processing (const struct ccase *c, void *mtf_, const struct dataset *ds UNUSED)
+mtf_processing (struct mtf_proc *mtf, struct dataset *ds)
{
- struct mtf_proc *mtf = mtf_;
-
- /* Do we need another record from the active file? */
- bool read_active_file;
+ struct mtf_file *min_head, *min_tail; /* Files with minimum BY values. */
+ struct mtf_file *max_head, *max_tail; /* Files with non-minimum BYs. */
+ struct mtf_file *iter, *next;
- assert (mtf->head != NULL);
- if (mtf->head->type == MTF_TABLE)
- return true;
-
- do
+ /* 3. Find the FILE input record(s) that have minimum BY
+ values. Store all the values from these input records into
+ the output record. */
+ min_head = min_tail = mtf->head;
+ max_head = max_tail = NULL;
+ for (iter = mtf->head->next; iter && iter->type == MTF_FILE;
+ iter = iter->next)
{
- struct mtf_file *min_head, *min_tail; /* Files with minimum BY values. */
- struct mtf_file *max_head, *max_tail; /* Files with non-minimum BYs. */
- struct mtf_file *iter, *next;
-
- read_active_file = false;
+ int cmp = mtf_compare_BY_values (mtf, min_head, iter);
+ if (cmp < 0)
+ {
+ if (max_head)
+ max_tail = max_tail->next_min = iter;
+ else
+ max_head = max_tail = iter;
+ }
+ else if (cmp == 0)
+ min_tail = min_tail->next_min = iter;
+ else /* cmp > 0 */
+ {
+ if (max_head)
+ {
+ max_tail->next_min = min_head;
+ max_tail = min_tail;
+ }
+ else
+ {
+ max_head = min_head;
+ max_tail = min_tail;
+ }
+ min_head = min_tail = iter;
+ }
+ }
- /* 3. Find the FILE input record(s) that have minimum BY
- values. Store all the values from these input records into
- the output record. */
- min_head = min_tail = mtf->head;
- max_head = max_tail = NULL;
- for (iter = mtf->head->next; iter && iter->type == MTF_FILE;
- iter = iter->next)
+ /* 4. For every TABLE, read another record as long as the BY
+ values on the TABLE's input record are less than the FILEs'
+ BY values. If an exact match is found, store all the values
+ from the TABLE input record into the output record. */
+ for (; iter != NULL; iter = next)
+ {
+ assert (iter->type == MTF_TABLE);
+
+ next = iter->next;
+ for (;;)
{
- int cmp = mtf_compare_BY_values (mtf, min_head, iter, c);
+ int cmp = mtf_compare_BY_values (mtf, min_head, iter);
if (cmp < 0)
{
if (max_head)
else
max_head = max_tail = iter;
}
- else if (cmp == 0)
- min_tail = min_tail->next_min = iter;
+ else if (cmp == 0)
+ min_tail = min_tail->next_min = iter;
else /* cmp > 0 */
{
- if (max_head)
- {
- max_tail->next_min = min_head;
- max_tail = min_tail;
- }
- else
- {
- max_head = min_head;
- max_tail = min_tail;
- }
- min_head = min_tail = iter;
+ if (iter->handle
+ ? any_reader_read (iter->reader, iter->input)
+ : proc_read (ds, &iter->input))
+ continue;
+ if (!mtf_delete_file_in_place (mtf, &iter))
+ return false;
}
+ break;
}
-
- /* 4. For every TABLE, read another record as long as the BY
- values on the TABLE's input record are less than the FILEs'
- BY values. If an exact match is found, store all the values
- from the TABLE input record into the output record. */
- for (; iter != NULL; iter = next)
- {
- assert (iter->type == MTF_TABLE);
-
- next = iter->next;
- for (;;)
- {
- int cmp = mtf_compare_BY_values (mtf, min_head, iter, c);
- if (cmp < 0)
- {
- if (max_head)
- max_tail = max_tail->next_min = iter;
- else
- max_head = max_tail = iter;
- }
- else if (cmp == 0)
- min_tail = min_tail->next_min = iter;
- else /* cmp > 0 */
- {
- if (iter->handle == NULL)
- return true;
- if (any_reader_read (iter->reader, &iter->input))
- continue;
- if (!mtf_delete_file_in_place (mtf, &iter))
- return false;
- }
- break;
- }
- }
+ }
- /* Next sequence number. */
- mtf->seq_num++;
+ /* Next sequence number. */
+ mtf->seq_num++;
- /* Store data to all the records we are using. */
- if (min_tail)
- min_tail->next_min = NULL;
- for (iter = min_head; iter; iter = iter->next_min)
- {
- int i;
+ /* Store data to all the records we are using. */
+ if (min_tail)
+ min_tail->next_min = NULL;
+ for (iter = min_head; iter; iter = iter->next_min)
+ {
+ int i;
- for (i = 0; i < dict_get_var_cnt (iter->dict); i++)
- {
- struct variable *v = dict_get_var (iter->dict, i);
- struct variable *mv = get_master (v);
+ for (i = 0; i < dict_get_var_cnt (iter->dict); i++)
+ {
+ struct variable *v = dict_get_var (iter->dict, i);
+ struct variable *mv = get_master (v);
+ size_t mv_index = mv ? var_get_dict_index (mv) : 0;
- if (mv != NULL && mtf->seq_nums[mv->index] != mtf->seq_num)
- {
- const struct ccase *record
- = case_is_null (&iter->input) ? c : &iter->input;
- union value *out = case_data_rw (&mtf->mtf_case, mv->fv);
-
- mtf->seq_nums[mv->index] = mtf->seq_num;
- if (var_is_numeric (v))
- out->f = case_num (record, v->fv);
- else
- memcpy (out->s, case_str (record, v->fv),
- var_get_width (v));
- }
- }
- if (iter->in_var != NULL)
- case_data_rw (&mtf->mtf_case, iter->in_var->fv)->f = 1.;
+ if (mv != NULL && mtf->seq_nums[mv_index] != mtf->seq_num)
+ {
+ const struct ccase *record = iter->input;
+ union value *out = case_data_rw (&mtf->mtf_case, mv);
- if (iter->type == MTF_FILE && iter->handle == NULL)
- read_active_file = true;
- }
+ mtf->seq_nums[mv_index] = mtf->seq_num;
+ if (var_is_numeric (v))
+ out->f = case_num (record, v);
+ else
+ memcpy (out->s, case_str (record, v), var_get_width (v));
+ }
+ }
+ if (iter->in_var != NULL)
+ case_data_rw (&mtf->mtf_case, iter->in_var)->f = 1.;
+ }
- /* Store missing values to all the records we're not
- using. */
- if (max_tail)
- max_tail->next_min = NULL;
- for (iter = max_head; iter; iter = iter->next_min)
- {
- int i;
+ /* Store missing values to all the records we're not using. */
+ if (max_tail)
+ max_tail->next_min = NULL;
+ for (iter = max_head; iter; iter = iter->next_min)
+ {
+ int i;
- for (i = 0; i < dict_get_var_cnt (iter->dict); i++)
- {
- struct variable *v = dict_get_var (iter->dict, i);
- struct variable *mv = get_master (v);
+ for (i = 0; i < dict_get_var_cnt (iter->dict); i++)
+ {
+ struct variable *v = dict_get_var (iter->dict, i);
+ struct variable *mv = get_master (v);
+ size_t mv_index = mv ? var_get_dict_index (mv) : 0;
- if (mv != NULL && mtf->seq_nums[mv->index] != mtf->seq_num)
- {
- union value *out = case_data_rw (&mtf->mtf_case, mv->fv);
- mtf->seq_nums[mv->index] = mtf->seq_num;
+ if (mv != NULL && mtf->seq_nums[mv_index] != mtf->seq_num)
+ {
+ union value *out = case_data_rw (&mtf->mtf_case, mv);
+ mtf->seq_nums[mv_index] = mtf->seq_num;
- if (var_is_numeric (v))
- out->f = SYSMIS;
- else
- memset (out->s, ' ', var_get_width (v));
- }
+ if (var_is_numeric (v))
+ out->f = SYSMIS;
+ else
+ memset (out->s, ' ', var_get_width (v));
}
- if (iter->in_var != NULL)
- case_data_rw (&mtf->mtf_case, iter->in_var->fv)->f = 0.;
- }
+ }
+ if (iter->in_var != NULL)
+ case_data_rw (&mtf->mtf_case, iter->in_var)->f = 0.;
+ }
- /* 5. Write the output record. */
- casefile_append (mtf->output, &mtf->mtf_case);
+ /* 5. Write the output record. */
+ casefile_append (mtf->output, &mtf->mtf_case);
- /* 6. Read another record from each input file FILE and TABLE
- that we stored values from above. If we come to the end of
- one of the input files, remove it from the list of input
- files. */
- for (iter = min_head; iter && iter->type == MTF_FILE; iter = next)
- {
- next = iter->next_min;
- if (iter->reader != NULL
- && !any_reader_read (iter->reader, &iter->input))
- if (!mtf_delete_file_in_place (mtf, &iter))
- return false;
- }
+ /* 6. Read another record from each input file FILE and TABLE
+ that we stored values from above. If we come to the end of
+ one of the input files, remove it from the list of input
+ files. */
+ for (iter = min_head; iter && iter->type == MTF_FILE; iter = next)
+ {
+ next = iter->next_min;
+ if (iter->reader != NULL
+ ? !any_reader_read (iter->reader, iter->input)
+ : !proc_read (ds, &iter->input))
+ if (!mtf_delete_file_in_place (mtf, &iter))
+ return false;
}
- while (!read_active_file
- && mtf->head != NULL && mtf->head->type == MTF_FILE);
-
return true;
}
if (var_get_width (dv) == var_get_width (mv))
{
- if (val_labs_count (dv->val_labs)
- && !val_labs_count (mv->val_labs))
- {
- val_labs_destroy (mv->val_labs);
- mv->val_labs = val_labs_copy (dv->val_labs);
- }
+ if (var_has_value_labels (dv) && !var_has_value_labels (mv))
+ var_set_value_labels (mv, var_get_value_labels (dv));
if (var_has_missing_values (dv) && !var_has_missing_values (mv))
var_set_missing_values (mv, var_get_missing_values (dv));
}
static struct variable *
get_master (struct variable *v)
{
- return v->aux;
+ return var_get_aux (v);
}
\f
/* Case map.
{
struct variable *v = dict_get_var (d, i);
int *src_fv = xmalloc (sizeof *src_fv);
- *src_fv = v->fv;
+ *src_fv = var_get_case_index (v);
var_attach_aux (v, src_fv, var_dtor_free);
}
}
int *src_fv = (int *) var_detach_aux (v);
size_t idx;
- if (v->fv != *src_fv)
+ if (var_get_case_index (v) != *src_fv)
identity_map = 0;
for (idx = 0; idx < value_cnt; idx++)
{
int src_idx = *src_fv + idx;
- int dst_idx = v->fv + idx;
+ int dst_idx = var_get_case_index (v) + idx;
assert (map->map[dst_idx] == -1);
map->map[dst_idx] = src_idx;
{
int src_idx = map->map[dst_idx];
if (src_idx != -1)
- *case_data_rw (dst, dst_idx) = *case_data (src, src_idx);
+ *case_data_rw_idx (dst, dst_idx) = *case_data_idx (src, src_idx);
}
}
projects / pspp-builds.git / blobdiff