struct agr_proc agr;
struct file_handle *out_file = NULL;
- /* Have we seen these subcommands? */
- unsigned seen = 0;
+ bool copy_documents = false;
+ bool presorted = false;
+ bool saw_direction;
- agr.writer = NULL;
- agr.sink = NULL;
+ memset(&agr, 0 , sizeof (agr));
agr.missing = ITEMWISE;
- agr.sort = NULL;
- agr.break_vars = NULL;
- agr.agr_vars = NULL;
- agr.dict = NULL;
- agr.case_cnt = 0;
- agr.prev_break = NULL;
agr.dict = dict_create ();
dict_set_label (agr.dict, dict_get_label (default_dict));
dict_set_documents (agr.dict, dict_get_documents (default_dict));
+
+ /* OUTFILE subcommand must be first. */
+ if (!lex_force_match_id ("OUTFILE"))
+ goto error;
+ lex_match ('=');
+ if (!lex_match ('*'))
+ {
+ out_file = fh_parse ();
+ if (out_file == NULL)
+ goto error;
+ }
/* Read most of the subcommands. */
for (;;)
{
lex_match ('/');
- if (lex_match_id ("OUTFILE"))
- {
- if (seen & 1)
- {
- msg (SE, _("%s subcommand given multiple times."),"OUTFILE");
- goto error;
- }
- seen |= 1;
-
- lex_match ('=');
- if (!lex_match ('*'))
- {
- out_file = fh_parse ();
- if (out_file == NULL)
- goto error;
- }
- }
- else if (lex_match_id ("MISSING"))
+ if (lex_match_id ("MISSING"))
{
lex_match ('=');
if (!lex_match_id ("COLUMNWISE"))
agr.missing = COLUMNWISE;
}
else if (lex_match_id ("DOCUMENT"))
- seen |= 2;
+ copy_documents = true;
else if (lex_match_id ("PRESORTED"))
- seen |= 4;
+ presorted = true;
else if (lex_match_id ("BREAK"))
{
int i;
- if (seen & 8)
- {
- msg (SE, _("%s subcommand given multiple times."),"BREAK");
- goto error;
- }
- seen |= 8;
-
lex_match ('=');
agr.sort = sort_parse_criteria (default_dict,
- &agr.break_vars, &agr.break_var_cnt);
+ &agr.break_vars, &agr.break_var_cnt,
+ &saw_direction);
if (agr.sort == NULL)
goto error;
for (i = 0; i < agr.break_var_cnt; i++)
{
struct variable *v = dict_clone_var (agr.dict, agr.break_vars[i],
- agr.break_vars[i]->name);
+ agr.break_vars[i]->name,
+ agr.break_vars[i]->longname
+ );
assert (v != NULL);
}
+
+ /* BREAK must follow the options. */
+ break;
}
- else break;
+ else
+ {
+ lex_error (_("expecting BREAK"));
+ goto error;
+ }
}
-
- /* Check for proper syntax. */
- if (!(seen & 8))
- msg (SW, _("BREAK subcommand not specified."));
+ 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 ('/');
if (!parse_aggregate_functions (&agr))
goto error;
/* Delete documents. */
- if (!(seen & 2))
+ if (!copy_documents)
dict_set_documents (agr.dict, NULL);
/* Cancel SPLIT FILE. */
so TEMPORARY is moot. */
cancel_temporary ();
- if (agr.sort != NULL && (seen & 4) == 0)
+ if (agr.sort != NULL && !presorted)
sort_active_file_in_place (agr.sort);
agr.sink = create_case_sink (&storage_sink_class, agr.dict, NULL);
}
else
{
- agr.writer = sfm_open_writer (out_file, agr.dict, get_scompression ());
+ agr.writer = sfm_open_writer (out_file, agr.dict, get_scompression (), 0);
if (agr.writer == NULL)
goto error;
- if (agr.sort != NULL && (seen & 4) == 0)
+ if (agr.sort != NULL && !presorted)
{
/* Sorting is needed. */
struct casefile *dst;
}
for (function = agr_func_tab; function->name; function++)
- if (!strcmp (function->name, tokid))
+ if (!strcasecmp (function->name, tokid))
break;
if (NULL == function->name)
{
lex_error (_("expecting `('"));
goto error;
}
- } else {
+ }
+ else
+ {
/* Parse list of source variables. */
{
int pv_opts = PV_NO_SCRATCH;
arg[i].c = xstrdup (ds_c_str (&tokstr));
type = ALPHA;
}
- else if (token == T_NUM)
+ else if (lex_is_number ())
{
arg[i].f = tokval;
type = NUMERIC;
goto error;
}
- /* Now check that the number of source variables match the
- number of target variables. Do this here because if we
- do it earlier then the user can get very misleading error
- messages; i.e., `AGGREGATE x=SUM(y t).' will get this
- error message when a proper message would be more like
- `unknown variable t'. */
+ /* 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 (%d) does not match "
n_src, n_dest);
goto error;
}
+
+ if ((func_index == PIN || func_index == POUT
+ || func_index == FIN || func_index == FOUT)
+ && ((src[0]->type == NUMERIC && arg[0].f > arg[1].f)
+ || (src[0]->type == ALPHA
+ && st_compare_pad (arg[0].c, strlen (arg[0].c),
+ arg[1].c, strlen (arg[1].c)) > 0)))
+ {
+ union value 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
/* Create the target variable in the aggregate
dictionary. */
{
+ static const struct fmt_spec f8_2 = {FMT_F, 8, 2};
struct variable *destvar;
v->function = func_index;
if (src)
{
- int output_width;
-
v->src = src[i];
if (src[i]->type == ALPHA)
v->function |= FSTRING;
v->string = xmalloc (src[i]->width);
}
-
- if (v->src->type == NUMERIC || function->alpha_type == NUMERIC)
- output_width = 0;
- else
- output_width = v->src->width;
if (function->alpha_type == ALPHA)
- destvar = dict_clone_var (agr->dict, v->src, dest[i]);
- else
- {
- destvar = dict_create_var (agr->dict, dest[i], output_width);
- if (output_width == 0)
- destvar->print = destvar->write = function->format;
- if (output_width == 0 && dict_get_weight (default_dict) != NULL
- && (func_index == N || func_index == N_NO_VARS
- || func_index == NU || func_index == NU_NO_VARS))
- {
- struct fmt_spec f = {FMT_F, 8, 2};
-
- destvar->print = destvar->write = f;
- }
- }
+ destvar = dict_clone_var (agr->dict, v->src, 0, dest[i] );
+ else if (v->src->type == NUMERIC
+ || function->alpha_type == NUMERIC)
+ {
+ destvar = dict_create_var (agr->dict, dest[i], 0);
+ if (destvar != NULL)
+ {
+ if ((func_index == N || func_index == NMISS)
+ && dict_get_weight (default_dict) != NULL)
+ destvar->print = destvar->write = f8_2;
+ else
+ destvar->print = destvar->write = function->format;
+ }
+ }
} else {
v->src = NULL;
destvar = dict_create_var (agr->dict, dest[i], 0);
+ if (func_index == N_NO_VARS
+ && dict_get_weight (default_dict) != NULL)
+ destvar->print = destvar->write = f8_2;
+ else
+ destvar->print = destvar->write = function->format;
}
if (!destvar)
"the aggregate variables and the break "
"variables."),
dest[i]);
- free (dest[i]);
goto error;
}
destvar->label = dest_label[i];
dest_label[i] = NULL;
}
- else if (function->alpha_type == ALPHA)
- destvar->print = destvar->write = function->format;
v->dest = destvar;
}
}
if (agr->dict != NULL)
dict_destroy (agr->dict);
+
case_destroy (&agr->agr_case);
}
\f
switch (iter->function)
{
case NMISS:
+ case NMISS | FSTRING:
iter->dbl[0] += weight;
break;
case NUMISS:
+ case NUMISS | FSTRING:
iter->int1++;
break;
}
switch (iter->function)
{
case SUM:
- iter->dbl[0] += v->f;
+ iter->dbl[0] += v->f * weight;
+ iter->int1 = 1;
break;
case MEAN:
iter->dbl[0] += v->f * weight;
case FOUT | FSTRING:
case POUT | FSTRING:
if (memcmp (iter->arg[0].c, v->s, iter->src->width) > 0
- && memcmp (iter->arg[1].c, v->s, iter->src->width) < 0)
+ || memcmp (iter->arg[1].c, v->s, iter->src->width) < 0)
iter->dbl[0] += weight;
iter->dbl[1] += weight;
break;
case N:
+ case N | FSTRING:
iter->dbl[0] += weight;
break;
case NU:
+ case NU | FSTRING:
iter->int1++;
break;
case FIRST:
memcpy (iter->string, v->s, iter->src->width);
iter->int1 = 1;
break;
+ case NMISS:
+ case NMISS | FSTRING:
+ case NUMISS:
+ case NUMISS | FSTRING:
+ /* Our value is not missing or it would have been
+ caught earlier. Nothing to do. */
+ break;
default:
assert (0);
}
&& (i->function & FUNC) != N && (i->function & FUNC) != NU
&& (i->function & FUNC) != NMISS && (i->function & FUNC) != NUMISS)
{
- if (i->function & FSTRING)
+ if (i->dest->type == ALPHA)
memset (v->s, ' ', i->dest->width);
else
v->f = SYSMIS;
switch (i->function)
{
case SUM:
- v->f = i->dbl[0];
+ v->f = i->int1 ? i->dbl[0] : SYSMIS;
break;
case MEAN:
v->f = i->dbl[1] != 0.0 ? i->dbl[0] / i->dbl[1] : SYSMIS;
else
memset (v->s, ' ', i->dest->width);
break;
- case FGT | FSTRING:
- case FLT | FSTRING:
- case FIN | FSTRING:
- case FOUT | FSTRING:
- v->f = i->int2 ? (double) i->int1 / (double) i->int2 : SYSMIS;
- break;
case FGT:
+ case FGT | FSTRING:
case FLT:
+ case FLT | FSTRING:
case FIN:
+ case FIN | FSTRING:
case FOUT:
+ case FOUT | FSTRING:
v->f = i->dbl[1] ? i->dbl[0] / i->dbl[1] : SYSMIS;
break;
case PGT:
v->f = i->dbl[1] ? i->dbl[0] / i->dbl[1] * 100.0 : SYSMIS;
break;
case N:
+ case N | FSTRING:
v->f = i->dbl[0];
break;
case NU:
+ case NU | FSTRING:
v->f = i->int1;
break;
case FIRST:
v->f = i->int1;
break;
case NMISS:
+ case NMISS | FSTRING:
v->f = i->dbl[0];
break;
case NUMISS:
+ case NUMISS | FSTRING:
v->f = i->int1;
break;
default:
for (iter = agr->agr_vars; iter; iter = iter->next)
{
iter->missing = 0;
+ iter->dbl[0] = iter->dbl[1] = iter->dbl[2] = 0.0;
+ iter->int1 = iter->int2 = 0;
switch (iter->function)
{
case MIN:
else
moments1_clear (iter->moments);
break;
- default:
- iter->dbl[0] = iter->dbl[1] = iter->dbl[2] = 0.0;
- iter->int1 = iter->int2 = 0;
- break;
+ default:
+ break;
}
}
}