int
cmd_variable_labels (struct lexer *lexer, struct dataset *ds)
{
struct dictionary *dict = dataset_dict (ds);
do
{
struct variable **v;
size_t nv;
size_t i;
if (!parse_variables (lexer, dict, &v, &nv, PV_NONE))
return CMD_FAILURE;
if (!lex_force_string (lexer))
{
free (v);
return CMD_FAILURE;
}
for (i = 0; i < nv; i++)
var_set_label (v[i], lex_tokcstr (lexer));
lex_get (lexer);
while (lex_token (lexer) == T_SLASH)
lex_get (lexer);
free (v);
}
while (lex_token (lexer) != T_ENDCMD);
return CMD_SUCCESS;
}
/* 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;
}
int
cmd_missing_values (struct lexer *lexer, struct dataset *ds)
{
struct dictionary *dict = dataset_dict (ds);
struct variable **v = NULL;
size_t nv;
bool ok = true;
while (lex_token (lexer) != T_ENDCMD)
{
size_t i;
if (!parse_variables (lexer, dict, &v, &nv, PV_NONE))
goto error;
if (!lex_force_match (lexer, T_LPAREN))
goto error;
for (i = 0; i < nv; i++)
var_clear_missing_values (v[i]);
if (!lex_match (lexer, T_RPAREN))
{
struct missing_values mv;
for (i = 0; i < nv; i++)
if (var_get_type (v[i]) != var_get_type (v[0]))
{
const struct variable *n = var_is_numeric (v[0]) ? v[0] : v[i];
const struct variable *s = var_is_numeric (v[0]) ? v[i] : v[0];
msg (SE, _("Cannot mix numeric variables (e.g. %s) and "
"string variables (e.g. %s) within a single list."),
var_get_name (n), var_get_name (s));
goto error;
}
if (var_is_numeric (v[0]))
{
mv_init (&mv, 0);
while (!lex_match (lexer, T_RPAREN))
{
enum fmt_type type = var_get_print_format (v[0])->type;
double x, y;
bool ok;
if (!parse_num_range (lexer, &x, &y, &type))
goto error;
ok = (x == y
? mv_add_num (&mv, x)
: mv_add_range (&mv, x, y));
if (!ok)
ok = false;
lex_match (lexer, T_COMMA);
}
}
else
{
mv_init (&mv, MV_MAX_STRING);
while (!lex_match (lexer, T_RPAREN))
{
uint8_t value[MV_MAX_STRING];
char *dict_mv;
size_t length;
if (!lex_force_string (lexer))
{
ok = false;
break;
}
dict_mv = recode_string (dict_get_encoding (dict), "UTF-8",
lex_tokcstr (lexer),
ss_length (lex_tokss (lexer)));
length = strlen (dict_mv);
if (length > MV_MAX_STRING)
{
/* XXX truncate graphemes not bytes */
msg (SE, _("Truncating missing value to maximum "
"acceptable length (%d bytes)."),
MV_MAX_STRING);
length = MV_MAX_STRING;
}
memset (value, ' ', MV_MAX_STRING);
memcpy (value, dict_mv, length);
free (dict_mv);
if (!mv_add_str (&mv, value))
ok = false;
lex_get (lexer);
lex_match (lexer, T_COMMA);
}
}
for (i = 0; i < nv; i++)
{
if (mv_is_resizable (&mv, var_get_width (v[i])))
var_set_missing_values (v[i], &mv);
else
{
msg (SE, _("Missing values provided are too long to assign "
"to variable of width %d."),
var_get_width (v[i]));
ok = false;
}
}
mv_destroy (&mv);
}
lex_match (lexer, T_SLASH);
free (v);
v = NULL;
}
free (v);
return ok ? CMD_SUCCESS : CMD_FAILURE;
error:
free (v);
return CMD_FAILURE;
}
/* Parses the whole DO REPEAT command specification.
Returns success. */
static bool
parse_specification (struct lexer *lexer, struct dictionary *dict,
struct hmap *dummies)
{
struct dummy_var *first_dv = NULL;
hmap_init (dummies);
do
{
struct dummy_var *dv;
const char *name;
bool ok;
/* Get a stand-in variable name and make sure it's unique. */
if (!lex_force_id (lexer))
goto error;
name = lex_tokcstr (lexer);
if (dict_lookup_var (dict, name))
msg (SW, _("Dummy variable name `%s' hides dictionary variable `%s'."),
name, name);
if (find_dummy_var (dummies, name, strlen (name)))
{
msg (SE, _("Dummy variable name `%s' is given twice."), name);
goto error;
}
/* Make a new macro. */
dv = xmalloc (sizeof *dv);
dv->name = xstrdup (name);
dv->values = NULL;
dv->n_values = 0;
hmap_insert (dummies, &dv->hmap_node, hash_dummy (name, strlen (name)));
/* Skip equals sign. */
lex_get (lexer);
if (!lex_force_match (lexer, T_EQUALS))
goto error;
/* Get the details of the variable's possible values. */
if (lex_token (lexer) == T_ID || lex_token (lexer) == T_ALL)
ok = parse_ids (lexer, dict, dv);
else if (lex_is_number (lexer))
ok = parse_numbers (lexer, dv);
else if (lex_is_string (lexer))
ok = parse_strings (lexer, dv);
else
{
lex_error (lexer, NULL);
goto error;
}
if (!ok)
goto error;
assert (dv->n_values > 0);
if (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
{
lex_error (lexer, NULL);
goto error;
}
/* If this is the first variable then it defines how many replacements
there must be; otherwise enforce this number of replacements. */
if (first_dv == NULL)
first_dv = dv;
else if (first_dv->n_values != dv->n_values)
{
msg (SE, _("Dummy variable `%s' had %zu substitutions, so `%s' must "
"also, but %zu were specified."),
first_dv->name, first_dv->n_values,
dv->name, dv->n_values);
goto error;
}
lex_match (lexer, T_SLASH);
}
while (!lex_match (lexer, T_ENDCMD));
while (lex_match (lexer, T_ENDCMD))
continue;
return true;
error:
destroy_dummies (dummies);
return false;
}
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;
}
/* 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;
}
}