/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
#include <stdint.h>
#include <stdlib.h>
#include <ctype.h>
+#include <unistr.h>
#include "data/attributes.h"
#include "data/case.h"
#include "libpspp/compiler.h"
#include "libpspp/hash-functions.h"
#include "libpspp/hmap.h"
+#include "libpspp/i18n.h"
#include "libpspp/message.h"
#include "libpspp/misc.h"
#include "libpspp/pool.h"
#include "libpspp/str.h"
+#include "libpspp/string-array.h"
#include "gl/intprops.h"
#include "gl/minmax.h"
#include "gl/xalloc.h"
+#include "gl/xmemdup0.h"
#include "gettext.h"
#define _(msgid) gettext (msgid)
struct variable *filter; /* FILTER variable. */
casenumber case_limit; /* Current case limit (N command). */
char *label; /* File label. */
- struct string documents; /* Documents, as a string. */
+ struct string_array documents; /* Documents. */
struct vector **vector; /* Vectors of variables. */
size_t vector_cnt; /* Number of vectors. */
struct attrset attributes; /* Custom attributes. */
struct mrset **mrsets; /* Multiple response sets. */
size_t n_mrsets; /* Number of multiple response sets. */
+ /* Whether variable names must be valid identifiers. Normally, this is
+ true, but sometimes a dictionary is prepared for external use
+ (e.g. output to a CSV file) where names don't have to be valid. */
+ bool names_must_be_ids;
+
char *encoding; /* Character encoding of string data */
const struct dict_callbacks *callbacks; /* Callbacks on dictionary
static void dict_unset_split_var (struct dictionary *, struct variable *);
static void dict_unset_mrset_var (struct dictionary *, struct variable *);
-void
-dict_set_encoding (struct dictionary *d, const char *enc)
-{
- if (enc)
- {
- free (d->encoding);
- d->encoding = xstrdup (enc);
- }
-}
-
+/* Returns the encoding for data in dictionary D. The return value is a
+ nonnull string that contains an IANA character set name. */
const char *
dict_get_encoding (const struct dictionary *d)
{
return d->encoding ;
}
+/* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
+ encoding, false otherwise. If ISSUE_ERROR is true, issues an explanatory
+ error message on failure. */
+bool
+dict_id_is_valid (const struct dictionary *dict, const char *id,
+ bool issue_error)
+{
+ return (!dict->names_must_be_ids
+ || id_is_valid (id, dict->encoding, issue_error));
+}
void
dict_set_change_callback (struct dictionary *d,
dest->cb_data = src->cb_data;
}
-/* Creates and returns a new dictionary. */
+/* Creates and returns a new dictionary with the specified ENCODING. */
struct dictionary *
-dict_create (void)
+dict_create (const char *encoding)
{
struct dictionary *d = xzalloc (sizeof *d);
+ d->encoding = xstrdup (encoding);
+ d->names_must_be_ids = true;
hmap_init (&d->name_map);
attrset_init (&d->attributes);
+
return d;
}
dictionary. If the new dictionary won't be used to access
cases produced with the old dictionary, then the new
dictionary's case indexes should be compacted with
- dict_compact_values to save space. */
+ dict_compact_values to save space.
+
+ Callbacks are not cloned. */
struct dictionary *
dict_clone (const struct dictionary *s)
{
struct dictionary *d;
size_t i;
- d = dict_create ();
+ d = dict_create (s->encoding);
+ dict_set_names_must_be_ids (d, dict_get_names_must_be_ids (s));
for (i = 0; i < s->var_cnt; i++)
{
for (i = 0; i < s->vector_cnt; i++)
d->vector[i] = vector_clone (s->vector[i], s, d);
- if ( s->encoding)
- d->encoding = xstrdup (s->encoding);
-
dict_set_attributes (d, dict_get_attributes (s));
for (i = 0; i < s->n_mrsets; i++)
d->case_limit = 0;
free (d->label);
d->label = NULL;
- ds_destroy (&d->documents);
+ string_array_clear (&d->documents);
dict_clear_vectors (d);
attrset_clear (&d->attributes);
}
-/* Destroys the aux data for every variable in D, by calling
- var_clear_aux() for each variable. */
-void
-dict_clear_aux (struct dictionary *d)
-{
- int i;
-
- for (i = 0; i < d->var_cnt; i++)
- var_clear_aux (d->var[i].var);
-}
-
/* Clears a dictionary and destroys it. */
void
dict_destroy (struct dictionary *d)
{
if (d != NULL)
{
- /* In general, we don't want callbacks occuring, if the dictionary
+ /* In general, we don't want callbacks occurring, if the dictionary
is being destroyed */
d->callbacks = NULL ;
dict_clear (d);
+ string_array_destroy (&d->documents);
hmap_destroy (&d->name_map);
attrset_destroy (&d->attributes);
dict_clear_mrsets (d);
}
static struct variable *
-add_var (struct dictionary *d, struct variable *v)
+add_var_with_case_index (struct dictionary *d, struct variable *v,
+ int case_index)
{
struct vardict_info *vardict;
+ assert (case_index >= d->next_value_idx);
+
/* Update dictionary. */
if (d->var_cnt >= d->var_cap)
{
vardict->dict = d;
vardict->var = v;
hmap_insert (&d->name_map, &vardict->name_node,
- hash_case_string (var_get_name (v), 0));
- vardict->case_index = d->next_value_idx;
+ utf8_hash_case_string (var_get_name (v), 0));
+ vardict->case_index = case_index;
var_set_vardict (v, vardict);
if ( d->changed ) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->var_added )
d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
- d->next_value_idx++;
invalidate_proto (d);
+ d->next_value_idx = case_index + 1;
return v;
}
+static struct variable *
+add_var (struct dictionary *d, struct variable *v)
+{
+ return add_var_with_case_index (d, v, d->next_value_idx);
+}
+
/* Creates and returns a new variable in D with the given NAME
and WIDTH. Returns a null pointer if the given NAME would
duplicate that of an existing variable in the dictionary. */
return add_var (d, new_var);
}
+struct variable *
+dict_clone_var_in_place_assert (struct dictionary *d,
+ const struct variable *old_var)
+{
+ assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
+ return add_var_with_case_index (d, var_clone (old_var),
+ var_get_case_index (old_var));
+}
+
/* Returns the variable named NAME in D, or a null pointer if no
variable has that name. */
struct variable *
struct vardict_info *vardict;
HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
- hash_case_string (name, 0), &d->name_map)
+ utf8_hash_case_string (name, 0), &d->name_map)
{
struct variable *var = vardict->var;
- if (!strcasecmp (var_get_name (var), name))
+ if (!utf8_strcasecmp (var_get_name (var), name))
return var;
}
static void
reindex_var (struct dictionary *d, struct vardict_info *vardict)
{
- struct variable *var = vardict->var;
+ struct variable *old = (d->callbacks && d->callbacks->var_changed
+ ? var_clone (vardict->var)
+ : NULL);
+ struct variable *var = vardict->var;
var_set_vardict (var, vardict);
hmap_insert_fast (&d->name_map, &vardict->name_node,
vardict->name_node.hash);
if ( d->changed ) d->changed (d, d->changed_data);
- if ( d->callbacks && d->callbacks->var_changed )
- d->callbacks->var_changed (d, var_get_dict_index (var), d->cb_data);
+ if (old)
+ {
+ d->callbacks->var_changed (d, var_get_dict_index (var), VAR_TRAIT_POSITION, old, d->cb_data);
+ var_destroy (old);
+ }
}
/* Sets the case_index in V's vardict to CASE_INDEX. */
/* Deletes variable V from dictionary D and frees V.
This is a very bad idea if there might be any pointers to V
- from outside D. In general, no variable in the active file's
+ from outside D. In general, no variable in the active dataset's
dictionary should be deleted when any transformations are
active on the dictionary's dataset, because those
transformations might reference the deleted variable. The
{
int dict_index = var_get_dict_index (v);
const int case_index = var_get_case_index (v);
- const int width = var_get_width (v);
assert (dict_contains_var (d, v));
- /* Delete aux data. */
- var_clear_aux (v);
-
dict_unset_split_var (d, v);
dict_unset_mrset_var (d, v);
/* Free memory. */
var_clear_vardict (v);
- var_destroy (v);
if ( d->changed ) d->changed (d, d->changed_data);
invalidate_proto (d);
if (d->callbacks && d->callbacks->var_deleted )
- d->callbacks->var_deleted (d, dict_index, case_index, width, d->cb_data);
+ d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
+
+ var_destroy (v);
}
/* Deletes the COUNT variables listed in VARS from D. This is
struct vardict_info *vardict = var_get_vardict (v);
var_clear_vardict (v);
var_set_name (v, new_name);
- vardict->name_node.hash = hash_case_string (new_name, 0);
+ vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
var_set_vardict (v, vardict);
}
-/* Changes the name of V in D to name NEW_NAME. Assert-fails if
- a variable named NEW_NAME is already in D, except that
- NEW_NAME may be the same as V's existing name. */
-void
-dict_rename_var (struct dictionary *d, struct variable *v,
- const char *new_name)
+/* Tries to changes the name of V in D to name NEW_NAME. Returns true if
+ successful, false if a variable (other than V) with the given name already
+ exists in D. */
+bool
+dict_try_rename_var (struct dictionary *d, struct variable *v,
+ const char *new_name)
{
- assert (!strcasecmp (var_get_name (v), new_name)
- || dict_lookup_var (d, new_name) == NULL);
+ struct variable *conflict = dict_lookup_var (d, new_name);
+ if (conflict && v != conflict)
+ return false;
+ struct variable *old = var_clone (v);
unindex_var (d, var_get_vardict (v));
rename_var (v, new_name);
reindex_var (d, var_get_vardict (v));
if ( d->changed ) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->var_changed )
- d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
+ d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
+
+ var_destroy (old);
+
+ return true;
+}
+
+/* Changes the name of V in D to name NEW_NAME. Assert-fails if
+ a variable named NEW_NAME is already in D, except that
+ NEW_NAME may be the same as V's existing name. */
+void
+dict_rename_var (struct dictionary *d, struct variable *v,
+ const char *new_name)
+{
+ bool ok UNUSED = dict_try_rename_var (d, v, new_name);
+ assert (ok);
}
/* Renames COUNT variables specified in VARS to the names given
static char *
make_hinted_name (const struct dictionary *dict, const char *hint)
{
- char name[VAR_NAME_LEN + 1];
+ size_t hint_len = strlen (hint);
bool dropped = false;
- char *cp;
-
- for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++)
+ char *root, *rp;
+ size_t ofs;
+ int mblen;
+
+ /* The allocation size here is OK: characters that are copied directly fit
+ OK, and characters that are not copied directly are replaced by a single
+ '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
+ replaced by '_' too. */
+ root = rp = xmalloc (hint_len + 1);
+ for (ofs = 0; ofs < hint_len; ofs += mblen)
{
- if (cp == name
- ? lex_is_id1 (*hint) && *hint != '$'
- : lex_is_idn (*hint))
+ ucs4_t uc;
+
+ mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
+ hint_len - ofs);
+ if (rp == root
+ ? lex_uc_is_id1 (uc) && uc != '$'
+ : lex_uc_is_idn (uc))
{
if (dropped)
{
- *cp++ = '_';
+ *rp++ = '_';
dropped = false;
}
- if (cp < name + VAR_NAME_LEN)
- *cp++ = *hint;
+ rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
}
- else if (cp > name)
+ else if (rp != root)
dropped = true;
}
- *cp = '\0';
+ *rp = '\0';
- if (name[0] != '\0')
+ if (root[0] != '\0')
{
- size_t len = strlen (name);
unsigned long int i;
- if (var_name_is_insertable (dict, name))
- return xstrdup (name);
+ if (var_name_is_insertable (dict, root))
+ return root;
for (i = 0; i < ULONG_MAX; i++)
{
char suffix[INT_BUFSIZE_BOUND (i) + 1];
- int ofs;
+ char *name;
suffix[0] = '_';
- if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
+ if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
NOT_REACHED ();
- ofs = MIN (VAR_NAME_LEN - strlen (suffix), len);
- strcpy (&name[ofs], suffix);
-
+ name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
if (var_name_is_insertable (dict, name))
- return xstrdup (name);
+ {
+ free (root);
+ return name;
+ }
+ free (name);
}
}
+ free (root);
+
return NULL;
}
return make_numeric_name (dict, num_start);
}
+/* Returns whether variable names must be valid identifiers. Normally, this is
+ true, but sometimes a dictionary is prepared for external use (e.g. output
+ to a CSV file) where names don't have to be valid. */
+bool
+dict_get_names_must_be_ids (const struct dictionary *d)
+{
+ return d->names_must_be_ids;
+}
+
+/* Sets whether variable names must be valid identifiers. Normally, this is
+ true, but sometimes a dictionary is prepared for external use (e.g. output
+ to a CSV file) where names don't have to be valid.
+
+ Changing this setting from false to true doesn't make the dictionary check
+ all the existing variable names, so it can cause an invariant violation. */
+void
+dict_set_names_must_be_ids (struct dictionary *d, bool names_must_be_ids)
+{
+ d->names_must_be_ids = names_must_be_ids;
+}
+
/* Returns the weighting variable in dictionary D, or a null
pointer if the dictionary is unweighted. */
struct variable *
else
{
double w = case_num (c, d->weight);
- if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
- w = 0.0;
- if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
- *warn_on_invalid = false;
- msg (SW, _("At least one case in the data file had a weight value "
- "that was user-missing, system-missing, zero, or "
- "negative. These case(s) were ignored."));
- }
- return w;
+
+ return var_force_valid_weight (d->weight, w, warn_on_invalid);
}
}
return d->label;
}
-/* Sets D's file label to LABEL, truncating it to a maximum of 60
- characters.
+/* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
+ encoding.
Removes D's label if LABEL is null or the empty string. */
void
dict_set_label (struct dictionary *d, const char *label)
{
free (d->label);
- d->label = label != NULL && label[0] != '\0' ? xstrndup (label, 60) : NULL;
+ if (label == NULL || label[0] == '\0')
+ d->label = NULL;
+ else
+ d->label = utf8_encoding_trunc (label, d->encoding, 60);
}
-/* Returns the documents for D, or a null pointer if D has no
- documents. If the return value is nonnull, then the string
- will be an exact multiple of DOC_LINE_LENGTH bytes in length,
- with each segment corresponding to one line. */
-const char *
+/* Returns the documents for D, as an UTF-8 encoded string_array. The
+ return value is always nonnull; if there are no documents then the
+ string_arary is empty.*/
+const struct string_array *
dict_get_documents (const struct dictionary *d)
{
- return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
+ return &d->documents;
}
-/* Sets the documents for D to DOCUMENTS, or removes D's
- documents if DOCUMENT is a null pointer. If DOCUMENTS is
- nonnull, then it should be an exact multiple of
- DOC_LINE_LENGTH bytes in length, with each segment
- corresponding to one line. */
+/* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
void
-dict_set_documents (struct dictionary *d, const char *documents)
+dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
{
- size_t remainder;
+ size_t i;
+
+ dict_clear_documents (d);
- ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
+ for (i = 0; i < new_docs->n; i++)
+ dict_add_document_line (d, new_docs->strings[i], false);
+}
+
+/* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
+ into individual lines at new-line characters. Each line is truncated to at
+ most DOC_LINE_LENGTH bytes in D's encoding. */
+void
+dict_set_documents_string (struct dictionary *d, const char *new_docs)
+{
+ const char *s;
- /* In case the caller didn't get it quite right, pad out the
- final line with spaces. */
- remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
- if (remainder != 0)
- ds_put_byte_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
+ dict_clear_documents (d);
+ for (s = new_docs; *s != '\0'; )
+ {
+ size_t len = strcspn (s, "\n");
+ char *line = xmemdup0 (s, len);
+ dict_add_document_line (d, line, false);
+ free (line);
+
+ s += len;
+ if (*s == '\n')
+ s++;
+ }
}
/* Drops the documents from dictionary D. */
void
dict_clear_documents (struct dictionary *d)
{
- ds_clear (&d->documents);
+ string_array_clear (&d->documents);
}
-/* Appends LINE to the documents in D. LINE will be truncated or
- padded on the right with spaces to make it exactly
- DOC_LINE_LENGTH bytes long. */
-void
-dict_add_document_line (struct dictionary *d, const char *line)
+/* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
+ truncated so that it is no more than 80 bytes in the dictionary's
+ encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
+ then a warning will be issued. */
+bool
+dict_add_document_line (struct dictionary *d, const char *line,
+ bool issue_warning)
{
- if (strlen (line) > DOC_LINE_LENGTH)
+ size_t trunc_len;
+ bool truncated;
+
+ trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
+ truncated = line[trunc_len] != '\0';
+ if (truncated && issue_warning)
{
/* Note to translators: "bytes" is correct, not characters */
msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
}
- buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
- DOC_LINE_LENGTH, line, ' ');
+
+ string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
+
+ return !truncated;
}
/* Returns the number of document lines in dictionary D. */
size_t
dict_get_document_line_cnt (const struct dictionary *d)
{
- return ds_length (&d->documents) / DOC_LINE_LENGTH;
+ return d->documents.n;
}
-/* Copies document line number IDX from dictionary D into
- LINE, trimming off any trailing white space. */
-void
-dict_get_document_line (const struct dictionary *d,
- size_t idx, struct string *line)
+/* Returns document line number IDX in dictionary D. The caller must not
+ modify or free the returned string. */
+const char *
+dict_get_document_line (const struct dictionary *d, size_t idx)
{
- assert (idx < dict_get_document_line_cnt (d));
- ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
- DOC_LINE_LENGTH));
- ds_rtrim (line, ss_cstr (CC_SPACES));
+ assert (idx < d->documents.n);
+ return d->documents.strings[idx];
}
/* Creates in D a vector named NAME that contains the CNT
{
size_t i;
for (i = 0; i < d->vector_cnt; i++)
- if (!strcasecmp (vector_get_name (d->vector[i]), name))
+ if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
return d->vector[i];
return NULL;
}
size_t i;
for (i = 0; i < dict->n_mrsets; i++)
- if (!strcasecmp (name, dict->mrsets[i]->name))
+ if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
return i;
return SIZE_MAX;
calling dict_set_attributes for D will also destroy D's
attribute set. */
struct attrset *
-dict_get_attributes (const struct dictionary *d)
+dict_get_attributes (const struct dictionary *d)
{
return CONST_CAST (struct attrset *, &d->attributes);
}
/* Returns true if D has at least one attribute in its attribute
set, false if D's attribute set is empty. */
bool
-dict_has_attributes (const struct dictionary *d)
+dict_has_attributes (const struct dictionary *d)
{
return attrset_count (&d->attributes) > 0;
}
-/* Called from variable.c to notify the dictionary that some property of
- the variable has changed */
+/* Called from variable.c to notify the dictionary that some property (indicated
+ by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
+ prior to the change. OLDVAR is destroyed by this function.
+*/
void
-dict_var_changed (const struct variable *v)
+dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
{
if ( var_has_vardict (v))
{
if (d->changed ) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->var_changed )
- d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
+ d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
}
+ var_destroy (oldvar);
}
-/* Called from variable.c to notify the dictionary that the variable's width
- has changed */
-void
-dict_var_resized (const struct variable *v, int old_width)
-{
- if ( var_has_vardict (v))
- {
- const struct vardict_info *vardict = var_get_vardict (v);
- struct dictionary *d;
-
- d = vardict->dict;
-
- if (d->changed) d->changed (d, d->changed_data);
-
- invalidate_proto (d);
- if ( d->callbacks && d->callbacks->var_resized )
- d->callbacks->var_resized (d, var_get_dict_index (v), old_width,
- d->cb_data);
- }
-}
-
-/* Called from variable.c to notify the dictionary that the variable's display width
- has changed */
-void
-dict_var_display_width_changed (const struct variable *v)
-{
- if ( var_has_vardict (v))
- {
- const struct vardict_info *vardict = var_get_vardict (v);
- struct dictionary *d;
-
- d = vardict->dict;
-
- if (d->changed) d->changed (d, d->changed_data);
- if ( d->callbacks && d->callbacks->var_display_width_changed )
- d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);
- }
-}
\f
/* Dictionary used to contain "internal variables". */
static struct dictionary *internal_dict;
dict_create_internal_var (int case_idx, int width)
{
if (internal_dict == NULL)
- internal_dict = dict_create ();
+ internal_dict = dict_create ("UTF-8");
for (;;)
{
projects / pspp / blobdiff