/* Advances FILE to its next case. If BY is nonnull, then FILE's is_minimal
member is updated based on whether the new case's BY values still match
those in BY. */
static void
advance_file (struct comb_file *file, union value by[])
{
case_unref (file->data);
file->data = casereader_read (file->reader);
if (by)
file->is_minimal = (file->data != NULL
&& subcase_equal_cx (&file->by_vars, file->data, by));
}
static bool
read_input_case (struct merge *m, size_t idx)
{
struct merge_input *i = &m->inputs[idx];
i->c = casereader_read (i->reader);
if (i->c)
return true;
else
{
casereader_destroy (i->reader);
remove_element (m->inputs, m->input_cnt, sizeof *m->inputs, idx);
m->input_cnt--;
return false;
}
}
/* Reads FILE, which must be of type COMB_TABLE, until it
encounters a case with BY or greater for its BY variables.
Returns true if a case with exactly BY for its BY variables
was found, otherwise false. */
static bool
scan_table (struct comb_file *file, union value by[])
{
while (file->data != NULL)
{
int cmp = subcase_compare_3way_xc (&file->by_vars, by, file->data);
if (cmp > 0)
{
case_unref (file->data);
file->data = casereader_read (file->reader);
}
else
return cmp == 0;
}
return false;
}
/*
Update IA according to the contents of DICT and CREADER.
CREADER will be destroyed by this function.
*/
void
update_assistant (struct import_assistant *ia)
{
struct sheet_spec_page *ssp = ia->sheet_spec;
int rows = 0;
if (ssp->dict)
{
struct ccase *c;
int col;
ia->column_cnt = dict_get_var_cnt (ssp->dict);
ia->columns = xcalloc (ia->column_cnt, sizeof (*ia->columns));
for (col = 0; col < ia->column_cnt ; ++col)
{
const struct variable *var = dict_get_var (ssp->dict, col);
ia->columns[col].name = xstrdup (var_get_name (var));
ia->columns[col].contents = NULL;
}
for (; (c = casereader_read (ssp->reader)) != NULL; case_unref (c))
{
rows++;
for (col = 0; col < ia->column_cnt ; ++col)
{
char *ss;
const struct variable *var = dict_get_var (ssp->dict, col);
ia->columns[col].contents = xrealloc (ia->columns[col].contents,
sizeof (struct substring) * rows);
ss = data_out (case_data (c, var), dict_get_encoding (ssp->dict),
var_get_print_format (var));
ia->columns[col].contents[rows - 1] = ss_cstr (ss);
}
if (rows > MAX_PREVIEW_LINES)
{
case_unref (c);
break;
}
}
}
ia->file.line_cnt = rows;
}
void
order_stats_accumulate_idx (struct order_stats **os, size_t nos,
struct casereader *reader,
int wt_idx,
int val_idx)
{
struct ccase *cx;
struct ccase *prev_cx = NULL;
double prev_value = -DBL_MAX;
double cc_i = 0;
double c_i = 0;
for (; (cx = casereader_read (reader)) != NULL; case_unref (cx))
{
const double weight = (wt_idx == -1) ? 1.0 : case_data_idx (cx, wt_idx)->f;
const double this_value = case_data_idx (cx, val_idx)->f;
/* The casereader MUST be sorted */
assert (this_value >= prev_value);
if ( prev_value == -DBL_MAX || prev_value == this_value)
c_i += weight;
if ( prev_value > -DBL_MAX && this_value > prev_value)
{
update_k_values (prev_cx, prev_value, c_i, cc_i, os, nos);
c_i = weight;
}
case_unref (prev_cx);
cc_i += weight;
prev_value = this_value;
prev_cx = case_ref (cx);
}
update_k_values (prev_cx, prev_value, c_i, cc_i, os, nos);
case_unref (prev_cx);
casereader_destroy (reader);
}
static int
combine_files (enum comb_command_type command,
struct lexer *lexer, struct dataset *ds)
{
struct comb_proc proc;
bool saw_by = false;
bool saw_sort = false;
struct casereader *active_file = NULL;
char *first_name = NULL;
char *last_name = NULL;
struct taint *taint = NULL;
size_t n_tables = 0;
size_t allocated_files = 0;
size_t i;
proc.files = NULL;
proc.n_files = 0;
proc.dict = dict_create (get_default_encoding ());
proc.output = NULL;
proc.matcher = NULL;
subcase_init_empty (&proc.by_vars);
proc.first = NULL;
proc.last = NULL;
proc.buffered_case = NULL;
proc.prev_BY = NULL;
dict_set_case_limit (proc.dict, dict_get_case_limit (dataset_dict (ds)));
lex_match (lexer, T_SLASH);
for (;;)
{
struct comb_file *file;
enum comb_file_type type;
if (lex_match_id (lexer, "FILE"))
type = COMB_FILE;
else if (command == COMB_MATCH && lex_match_id (lexer, "TABLE"))
{
type = COMB_TABLE;
n_tables++;
}
else
break;
lex_match (lexer, T_EQUALS);
if (proc.n_files >= allocated_files)
proc.files = x2nrealloc (proc.files, &allocated_files,
sizeof *proc.files);
file = &proc.files[proc.n_files++];
file->type = type;
subcase_init_empty (&file->by_vars);
subcase_init_empty (&file->src);
subcase_init_empty (&file->dst);
file->mv = NULL;
file->handle = NULL;
file->dict = NULL;
file->reader = NULL;
file->data = NULL;
file->is_sorted = true;
file->in_name = NULL;
file->in_var = NULL;
if (lex_match (lexer, T_ASTERISK))
{
if (!dataset_has_source (ds))
{
msg (SE, _("Cannot specify the active dataset since none "
"has been defined."));
goto error;
}
if (proc_make_temporary_transformations_permanent (ds))
msg (SE, _("This command may not be used after TEMPORARY when "
"the active dataset is an input source. "
"Temporary transformations will be made permanent."));
file->dict = dict_clone (dataset_dict (ds));
}
else
{
file->handle = fh_parse (lexer, FH_REF_FILE, dataset_session (ds));
if (file->handle == NULL)
goto error;
file->reader = any_reader_open (file->handle, NULL, &file->dict);
if (file->reader == NULL)
goto error;
}
while (lex_match (lexer, T_SLASH))
if (lex_match_id (lexer, "RENAME"))
{
if (!parse_dict_rename (lexer, file->dict))
goto error;
}
else if (lex_match_id (lexer, "IN"))
{
lex_match (lexer, T_EQUALS);
if (lex_token (lexer) != T_ID)
{
lex_error (lexer, NULL);
goto error;
}
if (file->in_name)
{
msg (SE, _("Multiple IN subcommands for a single FILE or "
"TABLE."));
goto error;
}
file->in_name = xstrdup (lex_tokcstr (lexer));
lex_get (lexer);
}
else if (lex_match_id (lexer, "SORT"))
{
file->is_sorted = false;
saw_sort = true;
}
if (!merge_dictionary (proc.dict, file))
goto error;
}
while (lex_token (lexer) != T_ENDCMD)
{
if (lex_match (lexer, T_BY))
{
const struct variable **by_vars;
size_t i;
bool ok;
if (saw_by)
{
lex_sbc_only_once ("BY");
goto error;
}
saw_by = true;
lex_match (lexer, T_EQUALS);
if (!parse_sort_criteria (lexer, proc.dict, &proc.by_vars,
&by_vars, NULL))
goto error;
ok = true;
for (i = 0; i < proc.n_files; i++)
{
struct comb_file *file = &proc.files[i];
size_t j;
for (j = 0; j < subcase_get_n_fields (&proc.by_vars); j++)
{
const char *name = var_get_name (by_vars[j]);
struct variable *var = dict_lookup_var (file->dict, name);
if (var != NULL)
subcase_add_var (&file->by_vars, var,
subcase_get_direction (&proc.by_vars, j));
else
{
if (file->handle != NULL)
msg (SE, _("File %s lacks BY variable %s."),
fh_get_name (file->handle), name);
else
msg (SE, _("Active dataset lacks BY variable %s."),
name);
ok = false;
}
}
assert (!ok || subcase_conformable (&file->by_vars,
&proc.files[0].by_vars));
}
free (by_vars);
if (!ok)
goto error;
}
else if (command != COMB_UPDATE && lex_match_id (lexer, "FIRST"))
{
if (first_name != NULL)
{
lex_sbc_only_once ("FIRST");
goto error;
}
lex_match (lexer, T_EQUALS);
if (!lex_force_id (lexer))
goto error;
first_name = xstrdup (lex_tokcstr (lexer));
lex_get (lexer);
}
else if (command != COMB_UPDATE && lex_match_id (lexer, "LAST"))
{
if (last_name != NULL)
{
lex_sbc_only_once ("LAST");
goto error;
}
lex_match (lexer, T_EQUALS);
if (!lex_force_id (lexer))
goto error;
last_name = xstrdup (lex_tokcstr (lexer));
lex_get (lexer);
}
else if (lex_match_id (lexer, "MAP"))
{
/* FIXME. */
}
else if (lex_match_id (lexer, "DROP"))
{
if (!parse_dict_drop (lexer, proc.dict))
goto error;
}
else if (lex_match_id (lexer, "KEEP"))
{
if (!parse_dict_keep (lexer, proc.dict))
goto error;
}
else
{
lex_error (lexer, NULL);
goto error;
}
if (!lex_match (lexer, T_SLASH) && lex_token (lexer) != T_ENDCMD)
{
lex_end_of_command (lexer);
goto error;
}
}
if (!saw_by)
{
if (command == COMB_UPDATE)
{
lex_sbc_missing ("BY");
goto error;
}
if (n_tables)
{
msg (SE, _("BY is required when %s is specified."), "TABLE");
goto error;
}
if (saw_sort)
{
msg (SE, _("BY is required when %s is specified."), "SORT");
goto error;
}
}
/* Add IN, FIRST, and LAST variables to master dictionary. */
for (i = 0; i < proc.n_files; i++)
{
struct comb_file *file = &proc.files[i];
if (!create_flag_var ("IN", file->in_name, proc.dict, &file->in_var))
goto error;
}
if (!create_flag_var ("FIRST", first_name, proc.dict, &proc.first)
|| !create_flag_var ("LAST", last_name, proc.dict, &proc.last))
goto error;
dict_delete_scratch_vars (proc.dict);
dict_compact_values (proc.dict);
/* Set up mapping from each file's variables to master
variables. */
for (i = 0; i < proc.n_files; i++)
{
struct comb_file *file = &proc.files[i];
size_t src_var_cnt = dict_get_var_cnt (file->dict);
size_t j;
file->mv = xnmalloc (src_var_cnt, sizeof *file->mv);
for (j = 0; j < src_var_cnt; j++)
{
struct variable *src_var = dict_get_var (file->dict, j);
struct variable *dst_var = dict_lookup_var (proc.dict,
var_get_name (src_var));
if (dst_var != NULL)
{
size_t n = subcase_get_n_fields (&file->src);
file->mv[n] = var_get_missing_values (src_var);
subcase_add_var (&file->src, src_var, SC_ASCEND);
subcase_add_var (&file->dst, dst_var, SC_ASCEND);
}
}
}
proc.output = autopaging_writer_create (dict_get_proto (proc.dict));
taint = taint_clone (casewriter_get_taint (proc.output));
/* Set up case matcher. */
proc.matcher = case_matcher_create ();
for (i = 0; i < proc.n_files; i++)
{
struct comb_file *file = &proc.files[i];
if (file->reader == NULL)
{
if (active_file == NULL)
{
proc_discard_output (ds);
file->reader = active_file = proc_open_filtering (ds, false);
}
else
file->reader = casereader_clone (active_file);
}
if (!file->is_sorted)
file->reader = sort_execute (file->reader, &file->by_vars);
taint_propagate (casereader_get_taint (file->reader), taint);
file->data = casereader_read (file->reader);
if (file->type == COMB_FILE)
case_matcher_add_input (proc.matcher, &file->by_vars,
&file->data, &file->is_minimal);
}
if (command == COMB_ADD)
execute_add_files (&proc);
else if (command == COMB_MATCH)
execute_match_files (&proc);
else if (command == COMB_UPDATE)
execute_update (&proc);
else
NOT_REACHED ();
case_matcher_destroy (proc.matcher);
proc.matcher = NULL;
close_all_comb_files (&proc);
if (active_file != NULL)
proc_commit (ds);
dataset_set_dict (ds, proc.dict);
dataset_set_source (ds, casewriter_make_reader (proc.output));
proc.dict = NULL;
proc.output = NULL;
free_comb_proc (&proc);
free (first_name);
free (last_name);
return taint_destroy (taint) ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
error:
if (active_file != NULL)
proc_commit (ds);
free_comb_proc (&proc);
taint_destroy (taint);
free (first_name);
free (last_name);
return CMD_CASCADING_FAILURE;
}
/* Parses and executes the AGGREGATE procedure. */
int
cmd_aggregate (struct lexer *lexer, struct dataset *ds)
{
struct dictionary *dict = dataset_dict (ds);
struct agr_proc agr;
struct file_handle *out_file = NULL;
struct casereader *input = NULL, *group;
struct casegrouper *grouper;
struct casewriter *output = NULL;
bool copy_documents = false;
bool presorted = false;
bool saw_direction;
bool ok;
memset(&agr, 0 , sizeof (agr));
agr.missing = ITEMWISE;
agr.src_dict = dict;
subcase_init_empty (&agr.sort);
/* OUTFILE subcommand must be first. */
lex_match (lexer, T_SLASH);
if (!lex_force_match_id (lexer, "OUTFILE"))
goto error;
lex_match (lexer, T_EQUALS);
if (!lex_match (lexer, T_ASTERISK))
{
out_file = fh_parse (lexer, FH_REF_FILE, dataset_session (ds));
if (out_file == NULL)
goto error;
}
if (out_file == NULL && lex_match_id (lexer, "MODE"))
{
lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "ADDVARIABLES"))
{
agr.add_variables = true;
/* presorted is assumed in ADDVARIABLES mode */
presorted = true;
}
else if (lex_match_id (lexer, "REPLACE"))
{
agr.add_variables = false;
}
else
goto error;
}
if ( agr.add_variables )
agr.dict = dict_clone (dict);
else
agr.dict = dict_create (dict_get_encoding (dict));
dict_set_label (agr.dict, dict_get_label (dict));
dict_set_documents (agr.dict, dict_get_documents (dict));
/* Read most of the subcommands. */
for (;;)
{
lex_match (lexer, T_SLASH);
if (lex_match_id (lexer, "MISSING"))
{
lex_match (lexer, T_EQUALS);
if (!lex_match_id (lexer, "COLUMNWISE"))
{
lex_error_expecting (lexer, "COLUMNWISE", NULL);
goto error;
}
agr.missing = COLUMNWISE;
}
else if (lex_match_id (lexer, "DOCUMENT"))
copy_documents = true;
else if (lex_match_id (lexer, "PRESORTED"))
presorted = true;
else if (lex_force_match_id (lexer, "BREAK"))
{
int i;
lex_match (lexer, T_EQUALS);
if (!parse_sort_criteria (lexer, dict, &agr.sort, &agr.break_vars,
&saw_direction))
goto error;
agr.break_var_cnt = subcase_get_n_fields (&agr.sort);
if (! agr.add_variables)
for (i = 0; i < agr.break_var_cnt; i++)
dict_clone_var_assert (agr.dict, agr.break_vars[i]);
/* BREAK must follow the options. */
break;
}
else
goto error;
}
if (presorted && saw_direction)
msg (SW, _("When PRESORTED is specified, specifying sorting directions "
"with (A) or (D) has no effect. Output data will be sorted "
"the same way as the input data."));
/* Read in the aggregate functions. */
lex_match (lexer, T_SLASH);
if (!parse_aggregate_functions (lexer, dict, &agr))
goto error;
/* Delete documents. */
if (!copy_documents)
dict_clear_documents (agr.dict);
/* Cancel SPLIT FILE. */
dict_set_split_vars (agr.dict, NULL, 0);
/* Initialize. */
agr.case_cnt = 0;
if (out_file == NULL)
{
/* The active dataset will be replaced by the aggregated data,
so TEMPORARY is moot. */
proc_cancel_temporary_transformations (ds);
proc_discard_output (ds);
output = autopaging_writer_create (dict_get_proto (agr.dict));
}
else
{
output = any_writer_open (out_file, agr.dict);
if (output == NULL)
goto error;
}
input = proc_open (ds);
if (!subcase_is_empty (&agr.sort) && !presorted)
{
input = sort_execute (input, &agr.sort);
subcase_clear (&agr.sort);
}
for (grouper = casegrouper_create_vars (input, agr.break_vars,
agr.break_var_cnt);
casegrouper_get_next_group (grouper, &group);
casereader_destroy (group))
{
struct casereader *placeholder = NULL;
struct ccase *c = casereader_peek (group, 0);
if (c == NULL)
{
casereader_destroy (group);
continue;
}
initialize_aggregate_info (&agr);
if ( agr.add_variables )
placeholder = casereader_clone (group);
{
struct ccase *cg;
for (; (cg = casereader_read (group)) != NULL; case_unref (cg))
accumulate_aggregate_info (&agr, cg);
}
if (agr.add_variables)
{
struct ccase *cg;
for (; (cg = casereader_read (placeholder)) != NULL; case_unref (cg))
dump_aggregate_info (&agr, output, cg);
casereader_destroy (placeholder);
}
else
{
dump_aggregate_info (&agr, output, c);
}
case_unref (c);
}
if (!casegrouper_destroy (grouper))
goto error;
if (!proc_commit (ds))
{
input = NULL;
goto error;
}
input = NULL;
if (out_file == NULL)
{
struct casereader *next_input = casewriter_make_reader (output);
if (next_input == NULL)
goto error;
dataset_set_dict (ds, agr.dict);
dataset_set_source (ds, next_input);
agr.dict = NULL;
}
else
{
ok = casewriter_destroy (output);
output = NULL;
if (!ok)
goto error;
}
agr_destroy (&agr);
fh_unref (out_file);
return CMD_SUCCESS;
error:
if (input != NULL)
proc_commit (ds);
casewriter_destroy (output);
agr_destroy (&agr);
fh_unref (out_file);
return CMD_CASCADING_FAILURE;
}
static void
do_reliability (struct casereader *input, struct dataset *ds,
const struct reliability *rel)
{
int i;
int si;
struct ccase *c;
casenumber n_missing ;
casenumber n_valid = 0;
for (si = 0 ; si < rel->n_sc; ++si)
{
struct cronbach *s = &rel->sc[si];
s->m = xzalloc (sizeof (s->m) * s->n_items);
s->total = moments1_create (MOMENT_VARIANCE);
for (i = 0 ; i < s->n_items ; ++i )
s->m[i] = moments1_create (MOMENT_VARIANCE);
}
input = casereader_create_filter_missing (input,
rel->variables,
rel->n_variables,
rel->exclude,
&n_missing,
NULL);
for (si = 0 ; si < rel->n_sc; ++si)
{
struct cronbach *s = &rel->sc[si];
s->totals_idx = caseproto_get_n_widths (casereader_get_proto (input));
input =
casereader_create_append_numeric (input, append_sum,
s, NULL);
}
for (; (c = casereader_read (input)) != NULL; case_unref (c))
{
double weight = 1.0;
n_valid ++;
for (si = 0; si < rel->n_sc; ++si)
{
struct cronbach *s = &rel->sc[si];
for (i = 0 ; i < s->n_items ; ++i )
moments1_add (s->m[i], case_data (c, s->items[i])->f, weight);
moments1_add (s->total, case_data_idx (c, s->totals_idx)->f, weight);
}
}
casereader_destroy (input);
for (si = 0; si < rel->n_sc; ++si)
{
struct cronbach *s = &rel->sc[si];
s->sum_of_variances = 0;
for (i = 0 ; i < s->n_items ; ++i )
{
double weight, mean, variance;
moments1_calculate (s->m[i], &weight, &mean, &variance, NULL, NULL);
s->sum_of_variances += variance;
}
moments1_calculate (s->total, NULL, NULL, &s->variance_of_sums,
NULL, NULL);
s->alpha =
alpha (s->n_items, s->sum_of_variances, s->variance_of_sums);
}
text_item_submit (text_item_create_format (TEXT_ITEM_PARAGRAPH, _("Scale: %s"),
ds_cstr (&rel->scale_name)));
case_processing_summary (n_valid, n_missing, dataset_dict (ds));
}
void
sign_execute (const struct dataset *ds,
struct casereader *input,
enum mv_class exclude,
const struct npar_test *test,
bool exact UNUSED,
double timer UNUSED)
{
int i;
bool warn = true;
const struct dictionary *dict = dataset_dict (ds);
const struct two_sample_test *t2s = UP_CAST (test, const struct two_sample_test, parent);
struct ccase *c;
struct sign_test_params *stp = xcalloc (t2s->n_pairs, sizeof *stp);
struct casereader *r = input;
for (; (c = casereader_read (r)) != NULL; case_unref (c))
{
const double weight = dict_get_case_weight (dict, c, &warn);
for (i = 0 ; i < t2s->n_pairs; ++i )
{
variable_pair *vp = &t2s->pairs[i];
const union value *value0 = case_data (c, (*vp)[0]);
const union value *value1 = case_data (c, (*vp)[1]);
const double diff = value0->f - value1->f;
if (var_is_value_missing ((*vp)[0], value0, exclude))
continue;
if (var_is_value_missing ((*vp)[1], value1, exclude))
continue;
if ( diff > 0)
stp[i].pos += weight;
else if (diff < 0)
stp[i].neg += weight;
else
stp[i].ties += weight;
}
}
casereader_destroy (r);
for (i = 0 ; i < t2s->n_pairs; ++i )
{
int r = MIN (stp[i].pos, stp[i].neg);
stp[i].one_tailed_sig = gsl_cdf_binomial_P (r,
0.5,
stp[i].pos + stp[i].neg);
stp[i].point_prob = gsl_ran_binomial_pdf (r, 0.5,
stp[i].pos + stp[i].neg);
}
output_frequency_table (t2s, stp, dict);
output_statistics_table (t2s, stp);
free (stp);
}
