X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fsfm-read.c;h=bbf5762c11a57e29876f5516f40c90a58a685f12;hb=027eba771a7ef9d25e97f0026d6469ea34c06523;hp=fc51472020af32853156365f9fdbc35088632231;hpb=fcb9e49b2a2d57af7c001ae5d2eda9ac443ba36b;p=pspp diff --git a/src/sfm-read.c b/src/sfm-read.c index fc51472020..bbf5762c11 100644 --- a/src/sfm-read.c +++ b/src/sfm-read.c @@ -17,40 +17,24 @@ Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ -/* AIX requires this to be the first thing in the file. */ #include -#if __GNUC__ -#define alloca __builtin_alloca -#else -#if HAVE_ALLOCA_H -#include -#else -#ifdef _AIX -#pragma alloca -#else -#ifndef alloca /* predefined by HP cc +Olibcalls */ -char *alloca (); -#endif -#endif -#endif -#endif - +#include "sfm.h" +#include "sfmP.h" #include #include #include #include #include #include "alloc.h" -#include "avl.h" #include "error.h" #include "file-handle.h" #include "filename.h" #include "format.h" #include "getline.h" +#include "hash.h" #include "magic.h" #include "misc.h" -#include "sfm.h" -#include "sfmP.h" +#include "value-labels.h" #include "str.h" #include "var.h" @@ -101,65 +85,43 @@ void dump_dictionary (struct dictionary * dict); /* Utilities. */ /* bswap_int32(): Reverse the byte order of 32-bit integer *X. */ -#if __linux__ -#include -#include static inline void -bswap_int32 (int32 * x) +bswap_int32 (int32 *x) { - *x = ntohl (*x); -} -#else /* not Linux */ -static inline void -bswap_int32 (int32 * x) -{ - unsigned char *y = (char *) x; + unsigned char *y = (unsigned char *) x; unsigned char t; + t = y[0]; y[0] = y[3]; y[3] = t; + t = y[1]; y[1] = y[2]; y[2] = t; } -#endif /* not Linux */ /* Reverse the byte order of 64-bit floating point *X. */ static inline void -bswap_flt64 (flt64 * x) +bswap_flt64 (flt64 *x) { - /* Note that under compilers of any quality, half of this function - should optimize out as dead code. */ - unsigned char *y = (char *) x; + unsigned char *y = (unsigned char *) x; + unsigned char t; - if (sizeof (flt64) == 8) - { - unsigned char t; - t = y[0]; - y[0] = y[7]; - y[7] = t; - t = y[1]; - y[1] = y[6]; - y[6] = t; - t = y[2]; - y[2] = y[5]; - y[5] = t; - t = y[3]; - y[3] = y[4]; - y[4] = t; - } - else - { - unsigned char t; - size_t x; + t = y[0]; + y[0] = y[7]; + y[7] = t; - for (x = 0; x < sizeof (flt64) / 2; x++) - { - t = y[x]; - y[x] = y[sizeof (flt64) - x]; - y[sizeof (flt64) - x] = t; - } - } + t = y[1]; + y[1] = y[6]; + y[6] = t; + + t = y[2]; + y[2] = y[5]; + y[5] = t; + + t = y[3]; + y[3] = y[4]; + y[4] = t; } static void @@ -327,10 +289,10 @@ sfm_read_dictionary (struct file_handle * h, struct sfm_read_info * inf) lose ((ME, _("%s: Weighting variable may not be a string variable."), h->fn)); - strcpy (ext->dict->weight_var, wv->name); + dict_set_weight (ext->dict, wv); } else - ext->dict->weight_var[0] = 0; + dict_set_weight (ext->dict, NULL); /* Read records of types 3, 4, 6, and 7. */ for (;;) @@ -449,7 +411,7 @@ lossage: free (var_by_index); fn_close (h->fn, ext->file); if (ext && ext->dict) - free_dictionary (ext->dict); + dict_destroy (ext->dict); free (ext); h->class = NULL; h->ext = NULL; @@ -565,20 +527,7 @@ read_header (struct file_handle * h, struct sfm_read_info * inf) int i; /* Create the dictionary. */ - dict = ext->dict = xmalloc (sizeof *dict); - dict->var = NULL; - dict->var_by_name = NULL; - dict->nvar = 0; - dict->N = 0; - dict->nval = -1; /* Unknown. */ - dict->n_splits = 0; - dict->splits = NULL; - dict->weight_var[0] = 0; - dict->weight_index = -1; - dict->filter_var[0] = 0; - dict->label = NULL; - dict->n_documents = 0; - dict->documents = NULL; + dict = ext->dict = dict_create (); /* Read header, check magic. */ assertive_bufread (h, &hdr, sizeof hdr, 0); @@ -668,14 +617,15 @@ read_header (struct file_handle * h, struct sfm_read_info * inf) { int i; - dict->label = NULL; for (i = sizeof hdr.file_label - 1; i >= 0; i--) if (!isspace ((unsigned char) hdr.file_label[i]) && hdr.file_label[i] != 0) { - dict->label = xmalloc (i + 2); - memcpy (dict->label, hdr.file_label, i + 1); - dict->label[i + 1] = 0; + char *label = xmalloc (i + 2); + memcpy (label, hdr.file_label, i + 1); + label[i + 1] = 0; + dict_set_label (dict, label); + free (label); break; } } @@ -713,14 +663,9 @@ lossage: } /* Reads most of the dictionary from file H; also fills in the - associated VAR_BY_INDEX array. - - Note: the dictionary returned by this function has an invalid NVAL - element, also the VAR[] array does not have the FV and LV elements - set, however the NV elements *are* set. This is because the caller - will probably modify the dictionary before reading it in from the - file. Also, the get.* elements are set to appropriate values to - allow the file to be read. */ + associated VAR_BY_INDEX array. The get.* elements in the + created dictionary are set to appropriate values to allow the + file to be read. */ static int read_variables (struct file_handle * h, struct variable *** var_by_index) { @@ -734,7 +679,6 @@ read_variables (struct file_handle * h, struct variable *** var_by_index) int next_value = 0; /* Index to next `value' structure. */ /* Allocate variables. */ - dict->var = xmalloc (sizeof *dict->var * ext->case_size); *var_by_index = xmalloc (sizeof **var_by_index * ext->case_size); /* Read in the entry for each variable and use the info to @@ -742,6 +686,7 @@ read_variables (struct file_handle * h, struct variable *** var_by_index) for (i = 0; i < ext->case_size; i++) { struct variable *vv; + char name[9]; int j; assertive_bufread (h, &sv, sizeof sv, 0); @@ -789,13 +734,6 @@ read_variables (struct file_handle * h, struct variable *** var_by_index) lose ((ME, _("%s: position %d: Missing value indicator field is not " "-3, -2, 0, 1, 2, or 3."), h->fn, i)); - /* Construct internal variable structure, initialize critical bits. */ - vv = (*var_by_index)[i] = dict->var[dict->nvar++] = xmalloc (sizeof *vv); - vv->index = dict->nvar - 1; - vv->foo = -1; - vv->label = NULL; - vv->val_lab = NULL; - /* Copy first character of variable name. */ if (!isalpha ((unsigned char) sv.name[0]) && sv.name[0] != '@' && sv.name[0] != '#') @@ -808,7 +746,7 @@ read_variables (struct file_handle * h, struct variable *** var_by_index) msg (MW, _("%s: position %d: Variable name begins with octothorpe " "(`#'). Scratch variables should not appear in system " "files."), h->fn, i); - vv->name[0] = toupper ((unsigned char) (sv.name[0])); + name[0] = toupper ((unsigned char) (sv.name[0])); /* Copy remaining characters of variable name. */ for (j = 1; j < 8; j++) @@ -821,38 +759,31 @@ read_variables (struct file_handle * h, struct variable *** var_by_index) { msg (MW, _("%s: position %d: Variable name character %d is " "lowercase letter %c."), h->fn, i, j + 1, sv.name[j]); - vv->name[j] = toupper ((unsigned char) (c)); + name[j] = toupper ((unsigned char) (c)); } else if (isalnum (c) || c == '.' || c == '@' || c == '#' || c == '$' || c == '_') - vv->name[j] = c; + name[j] = c; else lose ((ME, _("%s: position %d: character `\\%03o' (%c) is not valid in a " "variable name."), h->fn, i, c, c)); } - vv->name[j] = 0; - - /* Set type, width, and `left' fields and allocate `value' - indices. */ - if (sv.type == 0) - { - vv->type = NUMERIC; - vv->width = 0; - vv->get.nv = 1; - vv->get.fv = next_value++; - vv->nv = 1; - } + name[j] = 0; + + /* Create variable. */ + vv = (*var_by_index)[i] = dict_create_var (dict, name, sv.type); + if (vv == NULL) + lose ((ME, _("%s: Duplicate variable name `%s' within system file."), + h->fn, name)); + + /* Case reading data. */ + vv->get.fv = next_value; + if (sv.type == 0) + vv->get.nv = 1; else - { - vv->type = ALPHA; - vv->width = sv.type; - vv->nv = DIV_RND_UP (vv->width, MAX_SHORT_STRING); - vv->get.nv = DIV_RND_UP (vv->width, sizeof (flt64)); - vv->get.fv = next_value; - next_value += vv->get.nv; - long_string_count = vv->get.nv - 1; - } - vv->left = (vv->name[0] == '#'); + vv->get.nv = DIV_RND_UP (sv.type, sizeof (flt64)); + long_string_count = vv->get.nv - 1; + next_value += vv->get.nv; /* Get variable label, if any. */ if (sv.has_var_label == 1) @@ -949,28 +880,11 @@ read_variables (struct file_handle * h, struct variable *** var_by_index) if (next_value != ext->case_size) lose ((ME, _("%s: System file header indicates %d variable positions but " "%d were read from file."), h->fn, ext->case_size, next_value)); - dict->var = xrealloc (dict->var, sizeof *dict->var * dict->nvar); - - /* Construct AVL tree of dictionary in order to speed up later - processing and to check for duplicate varnames. */ - dict->var_by_name = avl_create (NULL, cmp_variable, NULL); - for (i = 0; i < dict->nvar; i++) - if (NULL != avl_insert (dict->var_by_name, dict->var[i])) - lose ((ME, _("%s: Duplicate variable name `%s' within system file."), - h->fn, dict->var[i]->name)); return 1; lossage: - for (i = 0; i < dict->nvar; i++) - { - free (dict->var[i]->label); - free (dict->var[i]); - } - free (dict->var); - if (dict->var_by_name) - avl_destroy (dict->var_by_name, NULL); - free (dict); + dict_destroy (dict); ext->dict = NULL; return 0; @@ -1014,8 +928,14 @@ read_value_labels (struct file_handle * h, struct variable ** var_by_index) { struct sfm_fhuser_ext *ext = h->ext; /* File extension record. */ - flt64 *raw_label = NULL; /* Array of raw label values. */ - struct value_label **cooked_label = NULL; /* Array of cooked labels. */ + struct label + { + unsigned char raw_value[8]; /* Value as uninterpreted bytes. */ + union value value; /* Value. */ + char *label; /* Null-terminated label string. */ + }; + + struct label *labels = NULL; int32 n_labels; /* Number of labels. */ struct variable **var = NULL; /* Associated variables. */ @@ -1033,34 +953,28 @@ read_value_labels (struct file_handle * h, struct variable ** var_by_index) bswap_int32 (&n_labels); /* Allocate memory. */ - raw_label = xmalloc (sizeof *raw_label * n_labels); - cooked_label = xmalloc (sizeof *cooked_label * n_labels); + labels = xmalloc (n_labels * sizeof *labels); for (i = 0; i < n_labels; i++) - cooked_label[i] = NULL; + labels[i].label = NULL; - /* Read each value/label tuple. */ + /* Read each value/label tuple into labels[]. */ for (i = 0; i < n_labels; i++) { - flt64 value; + struct label *label = labels + i; unsigned char label_len; + size_t padded_len; - int rem; - - /* Read value, label length. */ - assertive_bufread (h, &value, sizeof value, 0); - assertive_bufread (h, &label_len, 1, 0); - memcpy (&raw_label[i], &value, sizeof value); + /* Read value. */ + assertive_bufread (h, label->raw_value, sizeof label->raw_value, 0); - /* Read label. */ - cooked_label[i] = xmalloc (sizeof **cooked_label); - cooked_label[i]->s = xmalloc (label_len + 1); - assertive_bufread (h, cooked_label[i]->s, label_len, 0); - cooked_label[i]->s[label_len] = 0; + /* Read label length. */ + assertive_bufread (h, &label_len, sizeof label_len, 0); + padded_len = ROUND_UP (label_len + 1, sizeof (flt64)); - /* Skip padding. */ - rem = REM_RND_UP (label_len + 1, sizeof (flt64)); - if (rem) - assertive_bufread (h, &value, rem, 0); + /* Read label, padding. */ + label->label = xmalloc (padded_len + 1); + assertive_bufread (h, label->label, padded_len - 1, 0); + label->label[label_len] = 0; } /* Second step: Read the type 4 record that has the list of @@ -1076,7 +990,7 @@ read_value_labels (struct file_handle * h, struct variable ** var_by_index) if (rec_type != 4) lose ((ME, _("%s: Variable index record (type 4) does not immediately " - "follow value label record (type 3) as it ought."), h->fn)); + "follow value label record (type 3) as it should."), h->fn)); } /* Read number of variables associated with value label from type 4 @@ -1084,15 +998,13 @@ read_value_labels (struct file_handle * h, struct variable ** var_by_index) assertive_bufread (h, &n_vars, sizeof n_vars, 0); if (ext->reverse_endian) bswap_int32 (&n_vars); - if (n_vars < 1 || n_vars > ext->dict->nvar) + if (n_vars < 1 || n_vars > dict_get_var_cnt (ext->dict)) lose ((ME, _("%s: Number of variables associated with a value label (%d) " "is not between 1 and the number of variables (%d)."), - h->fn, n_vars, ext->dict->nvar)); - - /* Allocate storage. */ - var = xmalloc (sizeof *var * n_vars); + h->fn, n_vars, dict_get_var_cnt (ext->dict))); /* Read the list of variables. */ + var = xmalloc (n_vars * sizeof *var); for (i = 0; i < n_vars; i++) { int32 var_index; @@ -1110,12 +1022,12 @@ read_value_labels (struct file_handle * h, struct variable ** var_by_index) /* Make sure it's a real variable. */ v = var_by_index[var_index - 1]; if (v == NULL) - lose ((ME, _("%s: Variable index associated with value label (%d) refers " - "to a continuation of a string variable, not to an actual " - "variable."), h->fn, var_index)); + lose ((ME, _("%s: Variable index associated with value label (%d) " + "refers to a continuation of a string variable, not to " + "an actual variable."), h->fn, var_index)); if (v->type == ALPHA && v->width > MAX_SHORT_STRING) - lose ((ME, _("%s: Value labels are not allowed on long string variables " - "(%s)."), h->fn, v->name)); + lose ((ME, _("%s: Value labels are not allowed on long string " + "variables (%s)."), h->fn, v->name)); /* Add it to the list of variables. */ var[i] = v; @@ -1130,68 +1042,62 @@ read_value_labels (struct file_handle * h, struct variable ** var_by_index) var[0]->name, var[0]->type == ALPHA ? _("string") : _("numeric"), var[i]->name, var[i]->type == ALPHA ? _("string") : _("numeric"))); - /* Create a value_label for each value/label tuple, now that we know - the desired type. */ - for (i = 0; i < n_labels; i++) + /* Fill in labels[].value, now that we know the desired type. */ + for (i = 0; i < n_labels; i++) { + struct label *label = labels + i; + if (var[0]->type == ALPHA) - { - const int copy_len = min (sizeof (flt64), MAX_SHORT_STRING); - memcpy (cooked_label[i]->v.s, (char *) &raw_label[i], copy_len); - if (MAX_SHORT_STRING > copy_len) - memset (&cooked_label[i]->v.s[copy_len], ' ', - MAX_SHORT_STRING - copy_len); - } else { - cooked_label[i]->v.f = raw_label[i]; - if (ext->reverse_endian) - bswap_flt64 (&cooked_label[i]->v.f); - } - cooked_label[i]->ref_count = n_vars; + { + const int copy_len = min (sizeof (label->raw_value), + sizeof (label->label)); + memcpy (label->value.s, label->raw_value, copy_len); + } else { + flt64 f; + assert (sizeof f == sizeof label->raw_value); + memcpy (&f, label->raw_value, sizeof f); + if (ext->reverse_endian) + bswap_flt64 (&f); + label->value.f = f; + } } - + /* Assign the value_label's to each variable. */ for (i = 0; i < n_vars; i++) { struct variable *v = var[i]; int j; - /* Create AVL tree if necessary. */ - if (!v->val_lab) - v->val_lab = avl_create (NULL, val_lab_cmp, (void *) (v->width)); - /* Add each label to the variable. */ for (j = 0; j < n_labels; j++) { - struct value_label *old = avl_replace (v->val_lab, cooked_label[j]); - if (old == NULL) + struct label *label = labels + j; + if (!val_labs_replace (v->val_labs, label->value, label->label)) continue; if (var[0]->type == NUMERIC) msg (MW, _("%s: File contains duplicate label for value %g for " - "variable %s."), h->fn, cooked_label[j]->v.f, v->name); + "variable %s."), h->fn, label->value.f, v->name); else msg (MW, _("%s: File contains duplicate label for value `%.*s' " - "for variable %s."), h->fn, v->width, - cooked_label[j]->v.s, v->name); - - free_value_label (old); + "for variable %s."), + h->fn, v->width, label->value.s, v->name); } } - free (cooked_label); - free (raw_label); + for (i = 0; i < n_labels; i++) + free (labels[i].label); + free (labels); free (var); return 1; lossage: - if (cooked_label) - for (i = 0; i < n_labels; i++) - if (cooked_label[i]) - { - free (cooked_label[i]->s); - free (cooked_label[i]); - } - free (raw_label); + if (labels) + { + for (i = 0; i < n_labels; i++) + free (labels[i].label); + free (labels); + } free (var); return 0; } @@ -1227,20 +1133,24 @@ read_documents (struct file_handle * h) struct sfm_fhuser_ext *ext = h->ext; struct dictionary *dict = ext->dict; int32 n_lines; + char *documents; - if (dict->documents != NULL) + if (dict_get_documents (dict) != NULL) lose ((ME, _("%s: System file contains multiple type 6 (document) records."), h->fn)); assertive_bufread (h, &n_lines, sizeof n_lines, 0); - dict->n_documents = n_lines; - if (dict->n_documents <= 0) + if (n_lines <= 0) lose ((ME, _("%s: Number of document lines (%ld) must be greater than 0."), - h->fn, (long) dict->n_documents)); + h->fn, (long) n_lines)); - dict->documents = bufread (h, NULL, 80 * n_lines, 0); - if (dict->documents == NULL) + documents = bufread (h, NULL, 80 * n_lines, n_lines * 80 + 1); + /* FIXME? Run through asciify. */ + if (documents == NULL) return 0; + documents[80 * n_lines] = '\0'; + dict_set_documents (dict, documents); + free (documents); return 1; lossage: @@ -1263,12 +1173,10 @@ dump_dictionary (struct dictionary * dict) int n, j; debug_printf ((" var %s", v->name)); - /*debug_printf (("(indices:%d,%d)", v->index, v->foo));*/ debug_printf (("(type:%s,%d)", (v->type == NUMERIC ? _("num") : (v->type == ALPHA ? _("str") : "!!!")), v->width)); debug_printf (("(fv:%d,%d)", v->fv, v->nv)); - /*debug_printf (("(get.fv:%d,%d)", v->get.fv, v->get.nv));*/ debug_printf (("(left:%s)(miss:", v->left ? _("left") : _("right"))); switch (v->miss_type) @@ -1470,10 +1378,6 @@ sfm_read_case (struct file_handle * h, union value * perm, struct dictionary * d int i; - /* Make sure the caller remembered to finish polishing the - dictionary returned by sfm_read_dictionary(). */ - assert (dict->nval > 0); - /* The first concern is to obtain a full case relative to the data file. (Cases in the data file have no particular relationship to cases in the active file.) */ @@ -1498,9 +1402,9 @@ sfm_read_case (struct file_handle * h, union value * perm, struct dictionary * d /* Translate a case in data file format to a case in active file format. */ - for (i = 0; i < dict->nvar; i++) + for (i = 0; i < dict_get_var_cnt (dict); i++) { - struct variable *v = dict->var[i]; + struct variable *v = dict_get_var (dict, i); if (v->get.fv == -1) continue;