/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011 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 "data/dictionary.h"
+#include <stdint.h>
#include <stdlib.h>
#include <ctype.h>
+#include <unistr.h>
#include "data/attributes.h"
#include "data/case.h"
-#include "data/category.h"
#include "data/identifier.h"
+#include "data/mrset.h"
#include "data/settings.h"
#include "data/value-labels.h"
#include "data/vardict.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. */
char *encoding; /* Character encoding of string data */
void *changed_data;
};
+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)
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 id_is_valid (id, dict->encoding, issue_error);
+}
void
dict_set_change_callback (struct dictionary *d,
dict_set_attributes (d, dict_get_attributes (s));
+ for (i = 0; i < s->n_mrsets; i++)
+ {
+ const struct mrset *old = s->mrsets[i];
+ struct mrset *new;
+ size_t j;
+
+ /* Clone old mrset, then replace vars from D by vars from S. */
+ new = mrset_clone (old);
+ for (j = 0; j < new->n_vars; j++)
+ new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
+
+ dict_add_mrset (d, new);
+ }
+
return d;
}
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);
}
dict_clear (d);
hmap_destroy (&d->name_map);
attrset_destroy (&d->attributes);
+ dict_clear_mrsets (d);
+ free (d->encoding);
free (d);
}
}
/* 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
var_clear_aux (v);
dict_unset_split_var (d, v);
+ dict_unset_mrset_var (d, v);
if (d->weight == v)
dict_set_weight (d, NULL);
&& lex_id_to_token (ss_cstr (name)) == T_ID);
}
-static bool
-make_hinted_name (const struct dictionary *dict, const char *hint,
- char name[VAR_NAME_LEN + 1])
+static char *
+make_hinted_name (const struct dictionary *dict, const char *hint)
{
+ 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 true;
+ 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))
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 true;
+ {
+ free (root);
+ return name;
+ }
+ free (name);
}
}
- return false;
+ free (root);
+
+ return NULL;
}
-static bool
-make_numeric_name (const struct dictionary *dict, unsigned long int *num_start,
- char name[VAR_NAME_LEN + 1])
+static char *
+make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
{
unsigned long int number;
number < ULONG_MAX;
number++)
{
+ char name[3 + INT_STRLEN_BOUND (number) + 1];
+
sprintf (name, "VAR%03lu", number);
if (dict_lookup_var (dict, name) == NULL)
{
if (num_start != NULL)
*num_start = number + 1;
- return true;
+ return xstrdup (name);
}
}
- if (num_start != NULL)
- *num_start = ULONG_MAX;
- return false;
+ NOT_REACHED ();
}
-/* Attempts to devise a variable name unique within DICT.
- Returns true if successful, in which case the new variable
- name is stored into NAME. Returns false if all names that can
- be generated have already been taken. (Returning false is
- quite unlikely: at least ULONG_MAX unique names can be
- generated.)
+/* Devises and returns a variable name unique within DICT. The variable name
+ is owned by the caller, which must free it with free() when it is no longer
+ needed.
HINT, if it is non-null, is used as a suggestion that will be
modified for suitability as a variable name and for
value is used. If NUM_START is non-null, then its value is
used as the minimum numeric value to check, and it is updated
to the next value to be checked.
- */
-bool
+*/
+char *
dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
- unsigned long int *num_start,
- char name[VAR_NAME_LEN + 1])
+ unsigned long int *num_start)
{
- return ((hint != NULL && make_hinted_name (dict, hint, name))
- || make_numeric_name (dict, num_start, name));
+ if (hint != NULL)
+ {
+ char *hinted_name = make_hinted_name (dict, hint);
+ if (hinted_name != NULL)
+ return hinted_name;
+ }
+ return make_numeric_name (dict, num_start);
}
/* Returns the weighting variable in dictionary D, or a null
/* Removes variable V, which must be in D, from D's set of split
variables. */
-void
+static void
dict_unset_split_var (struct dictionary *d, struct variable *v)
{
int orig_count;
}
/* Sets D's file label to LABEL, truncating it to a maximum of 60
- characters. */
+ characters.
+
+ 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 ? xstrndup (label, 60) : NULL;
+ d->label = label != NULL && label[0] != '\0' ? xstrndup (label, 60) : NULL;
}
-/* 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;
- ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
+ dict_clear_documents (d);
- /* 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_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
+ 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;
+
+ 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
d->vector = NULL;
d->vector_cnt = 0;
}
+\f
+/* Multiple response sets. */
+
+/* Returns the multiple response set in DICT with index IDX, which must be
+ between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
+const struct mrset *
+dict_get_mrset (const struct dictionary *dict, size_t idx)
+{
+ assert (idx < dict->n_mrsets);
+ return dict->mrsets[idx];
+}
+
+/* Returns the number of multiple response sets in DICT. */
+size_t
+dict_get_n_mrsets (const struct dictionary *dict)
+{
+ return dict->n_mrsets;
+}
+/* Looks for a multiple response set named NAME in DICT. If it finds one,
+ returns its index; otherwise, returns SIZE_MAX. */
+static size_t
+dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
+{
+ size_t i;
+
+ for (i = 0; i < dict->n_mrsets; i++)
+ if (!strcasecmp (name, dict->mrsets[i]->name))
+ return i;
+
+ return SIZE_MAX;
+}
+
+/* Looks for a multiple response set named NAME in DICT. If it finds one,
+ returns it; otherwise, returns NULL. */
+const struct mrset *
+dict_lookup_mrset (const struct dictionary *dict, const char *name)
+{
+ size_t idx = dict_lookup_mrset_idx (dict, name);
+ return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
+}
+
+/* Adds MRSET to DICT, replacing any existing set with the same name. Returns
+ true if a set was replaced, false if none existed with the specified name.
+
+ Ownership of MRSET is transferred to DICT. */
+bool
+dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
+{
+ size_t idx;
+
+ assert (mrset_ok (mrset, dict));
+
+ idx = dict_lookup_mrset_idx (dict, mrset->name);
+ if (idx == SIZE_MAX)
+ {
+ dict->mrsets = xrealloc (dict->mrsets,
+ (dict->n_mrsets + 1) * sizeof *dict->mrsets);
+ dict->mrsets[dict->n_mrsets++] = mrset;
+ return true;
+ }
+ else
+ {
+ mrset_destroy (dict->mrsets[idx]);
+ dict->mrsets[idx] = mrset;
+ return false;
+ }
+}
+
+/* Looks for a multiple response set in DICT named NAME. If found, removes it
+ from DICT and returns true. If none is found, returns false without
+ modifying DICT.
+
+ Deleting one multiple response set causes the indexes of other sets within
+ DICT to change. */
+bool
+dict_delete_mrset (struct dictionary *dict, const char *name)
+{
+ size_t idx = dict_lookup_mrset_idx (dict, name);
+ if (idx != SIZE_MAX)
+ {
+ mrset_destroy (dict->mrsets[idx]);
+ dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
+ return true;
+ }
+ else
+ return false;
+}
+
+/* Deletes all multiple response sets from DICT. */
+void
+dict_clear_mrsets (struct dictionary *dict)
+{
+ size_t i;
+
+ for (i = 0; i < dict->n_mrsets; i++)
+ mrset_destroy (dict->mrsets[i]);
+ free (dict->mrsets);
+ dict->mrsets = NULL;
+ dict->n_mrsets = 0;
+}
+
+/* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
+static void
+dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
+{
+ size_t i;
+
+ assert (dict_contains_var (dict, var));
+
+ for (i = 0; i < dict->n_mrsets; )
+ {
+ struct mrset *mrset = dict->mrsets[i];
+ size_t j;
+
+ for (j = 0; j < mrset->n_vars; )
+ if (mrset->vars[j] == var)
+ remove_element (mrset->vars, mrset->n_vars--,
+ sizeof *mrset->vars, j);
+ else
+ j++;
+
+ if (mrset->n_vars < 2)
+ {
+ mrset_destroy (mrset);
+ dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
+ }
+ else
+ i++;
+ }
+}
+\f
/* Returns D's attribute set. The caller may examine or modify
the attribute set, but must not destroy it. Destroying D or
calling dict_set_attributes for D will also destroy D's
projects / pspp-builds.git / blobdiff