/* Parses a list of sort fields and appends them to ORDERING,
which the caller must already have initialized.
Returns true if successful, false on error.
If SAW_DIRECTION is nonnull, sets *SAW_DIRECTION to true if at
least one parenthesized sort direction was specified, false
otherwise. */
bool
parse_sort_criteria (struct lexer *lexer, const struct dictionary *dict,
struct subcase *ordering,
const struct variable ***vars, bool *saw_direction)
{
const struct variable **local_vars = NULL;
size_t var_cnt = 0;
if (vars == NULL)
vars = &local_vars;
*vars = NULL;
if (saw_direction != NULL)
*saw_direction = false;
do
{
size_t prev_var_cnt = var_cnt;
enum subcase_direction direction;
size_t i;
/* Variables. */
if (!parse_variables_const (lexer, dict, vars, &var_cnt,
PV_APPEND | PV_NO_SCRATCH))
goto error;
/* Sort direction. */
if (lex_match (lexer, T_LPAREN))
{
if (lex_match_id (lexer, "D") || lex_match_id (lexer, "DOWN"))
direction = SC_DESCEND;
else if (lex_match_id (lexer, "A") || lex_match_id (lexer, "UP"))
direction = SC_ASCEND;
else
{
lex_error_expecting (lexer, "A", "D", NULL_SENTINEL);
goto error;
}
if (!lex_force_match (lexer, T_RPAREN))
goto error;
if (saw_direction != NULL)
*saw_direction = true;
}
else
direction = SC_ASCEND;
for (i = prev_var_cnt; i < var_cnt; i++)
{
const struct variable *var = (*vars)[i];
if (!subcase_add_var (ordering, var, direction))
msg (SW, _("Variable %s specified twice in sort criteria."),
var_get_name (var));
}
}
while (lex_token (lexer) == T_ID
&& dict_lookup_var (dict, lex_tokcstr (lexer)) != NULL);
free (local_vars);
return true;
error:
free (local_vars);
if (vars)
*vars = NULL;
return false;
}
/* Parse all the aggregate functions. */
static bool
parse_aggregate_functions (struct lexer *lexer, const struct dictionary *dict,
struct agr_proc *agr)
{
struct agr_var *tail; /* Tail of linked list starting at agr->vars. */
/* Parse everything. */
tail = NULL;
for (;;)
{
char **dest;
char **dest_label;
size_t n_dest;
struct string function_name;
enum mv_class exclude;
const struct agr_func *function;
int func_index;
union agr_argument arg[2];
const struct variable **src;
size_t n_src;
size_t i;
dest = NULL;
dest_label = NULL;
n_dest = 0;
src = NULL;
function = NULL;
n_src = 0;
arg[0].c = NULL;
arg[1].c = NULL;
ds_init_empty (&function_name);
/* Parse the list of target variables. */
while (!lex_match (lexer, T_EQUALS))
{
size_t n_dest_prev = n_dest;
if (!parse_DATA_LIST_vars (lexer, dict, &dest, &n_dest,
(PV_APPEND | PV_SINGLE | PV_NO_SCRATCH
| PV_NO_DUPLICATE)))
goto error;
/* Assign empty labels. */
{
int j;
dest_label = xnrealloc (dest_label, n_dest, sizeof *dest_label);
for (j = n_dest_prev; j < n_dest; j++)
dest_label[j] = NULL;
}
if (lex_is_string (lexer))
{
dest_label[n_dest - 1] = xstrdup (lex_tokcstr (lexer));
lex_get (lexer);
}
}
/* Get the name of the aggregation function. */
if (lex_token (lexer) != T_ID)
{
lex_error (lexer, _("expecting aggregation function"));
goto error;
}
ds_assign_substring (&function_name, lex_tokss (lexer));
exclude = ds_chomp_byte (&function_name, '.') ? MV_SYSTEM : MV_ANY;
for (function = agr_func_tab; function->name; function++)
if (!c_strcasecmp (function->name, ds_cstr (&function_name)))
break;
if (NULL == function->name)
{
msg (SE, _("Unknown aggregation function %s."),
ds_cstr (&function_name));
goto error;
}
ds_destroy (&function_name);
func_index = function - agr_func_tab;
lex_get (lexer);
/* Check for leading lparen. */
if (!lex_match (lexer, T_LPAREN))
{
if (function->src_vars == AGR_SV_YES)
{
lex_force_match (lexer, T_LPAREN);
goto error;
}
}
else
{
/* Parse list of source variables. */
{
int pv_opts = PV_NO_SCRATCH;
if (func_index == SUM || func_index == MEAN || func_index == SD)
pv_opts |= PV_NUMERIC;
else if (function->n_args)
pv_opts |= PV_SAME_TYPE;
if (!parse_variables_const (lexer, dict, &src, &n_src, pv_opts))
goto error;
}
/* Parse function arguments, for those functions that
require arguments. */
if (function->n_args != 0)
for (i = 0; i < function->n_args; i++)
{
int type;
lex_match (lexer, T_COMMA);
if (lex_is_string (lexer))
{
arg[i].c = recode_string (dict_get_encoding (agr->dict),
"UTF-8", lex_tokcstr (lexer),
-1);
type = VAL_STRING;
}
else if (lex_is_number (lexer))
{
arg[i].f = lex_tokval (lexer);
type = VAL_NUMERIC;
}
else
{
msg (SE, _("Missing argument %zu to %s."),
i + 1, function->name);
goto error;
}
lex_get (lexer);
if (type != var_get_type (src[0]))
{
msg (SE, _("Arguments to %s must be of same type as "
"source variables."),
function->name);
goto error;
}
}
/* Trailing rparen. */
if (!lex_force_match (lexer, T_RPAREN))
goto error;
/* Now check that the number of source variables match
the number of target variables. If we check earlier
than this, the user can get very misleading error
message, i.e. `AGGREGATE x=SUM(y t).' will get this
error message when a proper message would be more
like `unknown variable t'. */
if (n_src != n_dest)
{
msg (SE, _("Number of source variables (%zu) does not match "
"number of target variables (%zu)."),
n_src, n_dest);
goto error;
}
if ((func_index == PIN || func_index == POUT
|| func_index == FIN || func_index == FOUT)
&& (var_is_numeric (src[0])
? arg[0].f > arg[1].f
: str_compare_rpad (arg[0].c, arg[1].c) > 0))
{
union agr_argument t = arg[0];
arg[0] = arg[1];
arg[1] = t;
msg (SW, _("The value arguments passed to the %s function "
"are out-of-order. They will be treated as if "
"they had been specified in the correct order."),
function->name);
}
}
/* Finally add these to the linked list of aggregation
variables. */
for (i = 0; i < n_dest; i++)
{
struct agr_var *v = xzalloc (sizeof *v);
/* Add variable to chain. */
if (agr->agr_vars != NULL)
tail->next = v;
else
agr->agr_vars = v;
tail = v;
tail->next = NULL;
v->moments = NULL;
/* Create the target variable in the aggregate
dictionary. */
{
struct variable *destvar;
v->function = func_index;
if (src)
{
v->src = src[i];
if (var_is_alpha (src[i]))
{
v->function |= FSTRING;
v->string = xmalloc (var_get_width (src[i]));
}
if (function->alpha_type == VAL_STRING)
destvar = dict_clone_var_as (agr->dict, v->src, dest[i]);
else
{
assert (var_is_numeric (v->src)
|| function->alpha_type == VAL_NUMERIC);
destvar = dict_create_var (agr->dict, dest[i], 0);
if (destvar != NULL)
{
struct fmt_spec f;
if ((func_index == N || func_index == NMISS)
&& dict_get_weight (dict) != NULL)
f = fmt_for_output (FMT_F, 8, 2);
else
f = function->format;
var_set_both_formats (destvar, &f);
}
}
} else {
struct fmt_spec f;
v->src = NULL;
destvar = dict_create_var (agr->dict, dest[i], 0);
if (destvar != NULL)
{
if ((func_index == N || func_index == NMISS)
&& dict_get_weight (dict) != NULL)
f = fmt_for_output (FMT_F, 8, 2);
else
f = function->format;
var_set_both_formats (destvar, &f);
}
}
if (!destvar)
{
msg (SE, _("Variable name %s is not unique within the "
"aggregate file dictionary, which contains "
"the aggregate variables and the break "
"variables."),
dest[i]);
goto error;
}
free (dest[i]);
if (dest_label[i])
var_set_label (destvar, dest_label[i]);
v->dest = destvar;
}
v->exclude = exclude;
if (v->src != NULL)
{
int j;
if (var_is_numeric (v->src))
for (j = 0; j < function->n_args; j++)
v->arg[j].f = arg[j].f;
else
for (j = 0; j < function->n_args; j++)
v->arg[j].c = xstrdup (arg[j].c);
}
}
if (src != NULL && var_is_alpha (src[0]))
for (i = 0; i < function->n_args; i++)
{
free (arg[i].c);
arg[i].c = NULL;
}
free (src);
free (dest);
free (dest_label);
if (!lex_match (lexer, T_SLASH))
{
if (lex_token (lexer) == T_ENDCMD)
return true;
lex_error (lexer, "expecting end of command");
return false;
}
continue;
error:
ds_destroy (&function_name);
for (i = 0; i < n_dest; i++)
{
free (dest[i]);
free (dest_label[i]);
}
free (dest);
free (dest_label);
free (arg[0].c);
free (arg[1].c);
if (src && n_src && var_is_alpha (src[0]))
for (i = 0; i < function->n_args; i++)
{
free (arg[i].c);
arg[i].c = NULL;
}
free (src);
return false;
}
}
int
cmd_t_test (struct lexer *lexer, struct dataset *ds)
{
bool ok;
const struct dictionary *dict = dataset_dict (ds);
struct tt tt;
int mode_count = 0;
/* Variables pertaining to the paired mode */
const struct variable **v1 = NULL;
size_t n_v1;
const struct variable **v2 = NULL;
size_t n_v2;
size_t n_pairs = 0;
vp *pairs = NULL;
/* One sample mode */
double testval = SYSMIS;
/* Independent samples mode */
const struct variable *gvar;
union value gval0;
union value gval1;
bool cut = false;
tt.wv = dict_get_weight (dict);
tt.dict = dict;
tt.confidence = 0.95;
tt.exclude = MV_ANY;
tt.missing_type = MISS_ANALYSIS;
tt.n_vars = 0;
tt.vars = NULL;
tt.mode = MODE_undef;
lex_match (lexer, T_EQUALS);
for (; lex_token (lexer) != T_ENDCMD; )
{
lex_match (lexer, T_SLASH);
if (lex_match_id (lexer, "TESTVAL"))
{
mode_count++;
tt.mode = MODE_SINGLE;
lex_match (lexer, T_EQUALS);
lex_force_num (lexer);
testval = lex_number (lexer);
lex_get (lexer);
}
else if (lex_match_id (lexer, "GROUPS"))
{
mode_count++;
cut = false;
tt.mode = MODE_INDEP;
lex_match (lexer, T_EQUALS);
if (NULL == (gvar = parse_variable (lexer, dict)))
goto parse_failed;
if (lex_match (lexer, T_LPAREN))
{
value_init (&gval0, var_get_width (gvar));
parse_value (lexer, &gval0, gvar);
cut = true;
if (lex_match (lexer, T_COMMA))
{
value_init (&gval1, var_get_width (gvar));
parse_value (lexer, &gval1, gvar);
cut = false;
}
lex_force_match (lexer, T_RPAREN);
}
else
{
value_init (&gval0, 0);
value_init (&gval1, 0);
gval0.f = 1.0;
gval1.f = 2.0;
cut = false;
}
if ( cut == true && var_is_alpha (gvar))
{
msg (SE, _("When applying GROUPS to a string variable, two "
"values must be specified."));
goto parse_failed;
}
}
else if (lex_match_id (lexer, "PAIRS"))
{
bool with = false;
bool paired = false;
if (tt.n_vars > 0)
{
msg (SE, _("VARIABLES subcommand may not be used with PAIRS."));
goto parse_failed;
}
mode_count++;
tt.mode = MODE_PAIRED;
lex_match (lexer, T_EQUALS);
if (!parse_variables_const (lexer, dict,
&v1, &n_v1,
PV_NO_DUPLICATE | PV_NUMERIC))
goto parse_failed;
if ( lex_match (lexer, T_WITH))
{
with = true;
if (!parse_variables_const (lexer, dict,
&v2, &n_v2,
PV_NO_DUPLICATE | PV_NUMERIC))
goto parse_failed;
if (lex_match (lexer, T_LPAREN)
&& lex_match_id (lexer, "PAIRED")
&& lex_match (lexer, T_RPAREN))
{
paired = true;
if (n_v1 != n_v2)
{
msg (SE, _("PAIRED was specified but the number of variables "
"preceding WITH (%zu) did not match the number "
"following (%zu)."),
n_v1, n_v2);
goto parse_failed;
}
}
}
{
int i;
if ( !with )
n_pairs = (n_v1 * (n_v1 - 1)) / 2.0;
else if ( paired )
n_pairs = n_v1;
else
n_pairs = n_v1 * n_v2;
pairs = xcalloc (n_pairs, sizeof *pairs);
if ( with)
{
int x = 0;
if (paired)
{
for (i = 0 ; i < n_v1; ++i)
{
vp *pair = &pairs[i];
(*pair)[0] = v1[i];
(*pair)[1] = v2[i];
}
}
else
{
for (i = 0 ; i < n_v1; ++i)
{
int j;
for (j = 0 ; j < n_v2; ++j)
{
vp *pair = &pairs[x++];
(*pair)[0] = v1[i];
(*pair)[1] = v2[j];
}
}
}
}
else
{
int x = 0;
for (i = 0 ; i < n_v1; ++i)
{
int j;
for (j = i + 1 ; j < n_v1; ++j)
{
vp *pair = &pairs[x++];
(*pair)[0] = v1[i];
(*pair)[1] = v1[j];
}
}
}
}
}
else if (lex_match_id (lexer, "VARIABLES"))
{
if ( tt.mode == MODE_PAIRED)
{
msg (SE, _("VARIABLES subcommand may not be used with PAIRS."));
goto parse_failed;
}
lex_match (lexer, T_EQUALS);
if (!parse_variables_const (lexer, dict,
&tt.vars,
&tt.n_vars,
PV_NO_DUPLICATE | PV_NUMERIC))
goto parse_failed;
}
else if ( lex_match_id (lexer, "MISSING"))
{
lex_match (lexer, T_EQUALS);
while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH)
{
if (lex_match_id (lexer, "INCLUDE"))
{
tt.exclude = MV_SYSTEM;
}
else if (lex_match_id (lexer, "EXCLUDE"))
{
tt.exclude = MV_ANY;
}
else if (lex_match_id (lexer, "LISTWISE"))
{
tt.missing_type = MISS_LISTWISE;
}
else if (lex_match_id (lexer, "ANALYSIS"))
{
tt.missing_type = MISS_ANALYSIS;
}
else
{
lex_error (lexer, NULL);
goto parse_failed;
}
lex_match (lexer, T_COMMA);
}
}
else if (lex_match_id (lexer, "CRITERIA"))
{
lex_match (lexer, T_EQUALS);
if ( lex_force_match_id (lexer, "CIN"))
if ( lex_force_match (lexer, T_LPAREN))
{
lex_force_num (lexer);
tt.confidence = lex_number (lexer);
lex_get (lexer);
lex_force_match (lexer, T_RPAREN);
}
}
else
{
lex_error (lexer, NULL);
goto parse_failed;
}
}
if ( mode_count != 1)
{
msg (SE, _("Exactly one of TESTVAL, GROUPS and PAIRS subcommands "
"must be specified."));
goto parse_failed;
}
if (tt.n_vars == 0 && tt.mode != MODE_PAIRED)
{
lex_sbc_missing ("VARIABLES");
goto parse_failed;
}
/* Deal with splits etc */
{
struct casereader *group;
struct casegrouper *grouper = casegrouper_create_splits (proc_open (ds), dict);
while (casegrouper_get_next_group (grouper, &group))
{
if ( tt.mode == MODE_SINGLE)
{
if ( tt.missing_type == MISS_LISTWISE )
group = casereader_create_filter_missing (group,
tt.vars, tt.n_vars,
tt.exclude,
NULL, NULL);
one_sample_run (&tt, testval, group);
}
else if ( tt.mode == MODE_PAIRED)
{
if ( tt.missing_type == MISS_LISTWISE )
{
group = casereader_create_filter_missing (group,
v1, n_v1,
tt.exclude,
NULL, NULL);
group = casereader_create_filter_missing (group,
v2, n_v2,
tt.exclude,
NULL, NULL);
}
paired_run (&tt, n_pairs, pairs, group);
}
else /* tt.mode == MODE_INDEP */
{
if ( tt.missing_type == MISS_LISTWISE )
{
group = casereader_create_filter_missing (group,
tt.vars, tt.n_vars,
tt.exclude,
NULL, NULL);
group = casereader_create_filter_missing (group,
&gvar, 1,
tt.exclude,
NULL, NULL);
}
indep_run (&tt, gvar, cut, &gval0, &gval1, group);
}
}
ok = casegrouper_destroy (grouper);
ok = proc_commit (ds) && ok;
}
free (pairs);
free (v1);
free (v2);
free (tt.vars);
return ok ? CMD_SUCCESS : CMD_FAILURE;
parse_failed:
return CMD_FAILURE;
}
int
cmd_reliability (struct lexer *lexer, struct dataset *ds)
{
const struct dictionary *dict = dataset_dict (ds);
struct reliability reliability;
reliability.n_variables = 0;
reliability.variables = NULL;
reliability.model = MODEL_ALPHA;
reliability.exclude = MV_ANY;
reliability.summary = 0;
reliability.wv = dict_get_weight (dict);
reliability.total_start = 0;
lex_match (lexer, T_SLASH);
if (!lex_force_match_id (lexer, "VARIABLES"))
{
goto error;
}
lex_match (lexer, T_EQUALS);
if (!parse_variables_const (lexer, dict, &reliability.variables, &reliability.n_variables,
PV_NO_DUPLICATE | PV_NUMERIC))
goto error;
if (reliability.n_variables < 2)
msg (MW, _("Reliability on a single variable is not useful."));
{
int i;
struct cronbach *c;
/* Create a default Scale */
reliability.n_sc = 1;
reliability.sc = xzalloc (sizeof (struct cronbach) * reliability.n_sc);
ds_init_cstr (&reliability.scale_name, "ANY");
c = &reliability.sc[0];
c->n_items = reliability.n_variables;
c->items = xzalloc (sizeof (struct variable*) * c->n_items);
for (i = 0 ; i < c->n_items ; ++i)
c->items[i] = reliability.variables[i];
}
while (lex_token (lexer) != T_ENDCMD)
{
lex_match (lexer, T_SLASH);
if (lex_match_id (lexer, "SCALE"))
{
struct const_var_set *vs;
if ( ! lex_force_match (lexer, T_LPAREN))
goto error;
if ( ! lex_force_string (lexer) )
goto error;
ds_init_substring (&reliability.scale_name, lex_tokss (lexer));
lex_get (lexer);
if ( ! lex_force_match (lexer, T_RPAREN))
goto error;
lex_match (lexer, T_EQUALS);
vs = const_var_set_create_from_array (reliability.variables, reliability.n_variables);
if (!parse_const_var_set_vars (lexer, vs, &reliability.sc->items, &reliability.sc->n_items, 0))
{
const_var_set_destroy (vs);
goto error;
}
const_var_set_destroy (vs);
}
else if (lex_match_id (lexer, "MODEL"))
{
lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "ALPHA"))
{
reliability.model = MODEL_ALPHA;
}
else if (lex_match_id (lexer, "SPLIT"))
{
reliability.model = MODEL_SPLIT;
reliability.split_point = -1;
if ( lex_match (lexer, T_LPAREN))
{
lex_force_num (lexer);
reliability.split_point = lex_number (lexer);
lex_get (lexer);
lex_force_match (lexer, T_RPAREN);
}
}
else
goto error;
}
else if (lex_match_id (lexer, "SUMMARY"))
{
lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "TOTAL"))
{
reliability.summary |= SUMMARY_TOTAL;
}
else if (lex_match (lexer, T_ALL))
{
reliability.summary = 0xFFFF;
}
else
goto error;
}
else if (lex_match_id (lexer, "MISSING"))
{
lex_match (lexer, T_EQUALS);
while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH)
{
if (lex_match_id (lexer, "INCLUDE"))
{
reliability.exclude = MV_SYSTEM;
}
else if (lex_match_id (lexer, "EXCLUDE"))
{
reliability.exclude = MV_ANY;
}
else
{
lex_error (lexer, NULL);
goto error;
}
}
}
else
{
lex_error (lexer, NULL);
goto error;
}
}
if ( reliability.model == MODEL_SPLIT)
{
int i;
const struct cronbach *s;
if ( reliability.split_point >= reliability.n_variables)
{
msg (ME, _("The split point must be less than the number of variables"));
goto error;
}
reliability.n_sc += 2 ;
reliability.sc = xrealloc (reliability.sc, sizeof (struct cronbach) * reliability.n_sc);
s = &reliability.sc[0];
reliability.sc[1].n_items =
(reliability.split_point == -1) ? s->n_items / 2 : reliability.split_point;
reliability.sc[2].n_items = s->n_items - reliability.sc[1].n_items;
reliability.sc[1].items = xzalloc (sizeof (struct variable *)
* reliability.sc[1].n_items);
reliability.sc[2].items = xzalloc (sizeof (struct variable *) *
reliability.sc[2].n_items);
for (i = 0; i < reliability.sc[1].n_items ; ++i)
reliability.sc[1].items[i] = s->items[i];
while (i < s->n_items)
{
reliability.sc[2].items[i - reliability.sc[1].n_items] = s->items[i];
i++;
}
}
if ( reliability.summary & SUMMARY_TOTAL)
{
int i;
const int base_sc = reliability.n_sc;
reliability.total_start = base_sc;
reliability.n_sc += reliability.sc[0].n_items ;
reliability.sc = xrealloc (reliability.sc, sizeof (struct cronbach) * reliability.n_sc);
for (i = 0 ; i < reliability.sc[0].n_items; ++i )
{
int v_src;
int v_dest = 0;
struct cronbach *s = &reliability.sc[i + base_sc];
s->n_items = reliability.sc[0].n_items - 1;
s->items = xzalloc (sizeof (struct variable *) * s->n_items);
for (v_src = 0 ; v_src < reliability.sc[0].n_items ; ++v_src)
{
if ( v_src != i)
s->items[v_dest++] = reliability.sc[0].items[v_src];
}
}
}
if ( ! run_reliability (ds, &reliability))
goto error;
free (reliability.variables);
return CMD_SUCCESS;
error:
free (reliability.variables);
return CMD_FAILURE;
}
