1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000 Free Software Foundation, Inc.
3 Written by Ben Pfaff <blp@gnu.org>.
5 This program is free software; you can redistribute it and/or
6 modify it under the terms of the GNU General Public License as
7 published by the Free Software Foundation; either version 2 of the
8 License, or (at your option) any later version.
10 This program is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
26 #include <data/case-source.h>
27 #include <data/case.h>
28 #include <data/data-in.h>
29 #include <data/dictionary.h>
30 #include <data/procedure.h>
31 #include <data/variable.h>
32 #include <language/command.h>
33 #include <language/data-io/data-reader.h>
34 #include <language/data-io/file-handle.h>
35 #include <language/lexer/lexer.h>
36 #include <language/lexer/variable-parser.h>
37 #include <libpspp/alloc.h>
38 #include <libpspp/array.h>
39 #include <libpspp/compiler.h>
40 #include <libpspp/message.h>
41 #include <libpspp/message.h>
42 #include <libpspp/misc.h>
43 #include <libpspp/pool.h>
44 #include <libpspp/str.h>
49 #define _(msgid) gettext (msgid)
51 /* FIXME: /N subcommand not implemented. It should be pretty simple,
54 /* Different types of variables for MATRIX DATA procedure. Order is
55 important: these are used for sort keys. */
58 MXD_SPLIT, /* SPLIT FILE variables. */
59 MXD_ROWTYPE, /* ROWTYPE_. */
60 MXD_FACTOR, /* Factor variables. */
61 MXD_VARNAME, /* VARNAME_. */
62 MXD_CONTINUOUS, /* Continuous variables. */
67 /* Format type enums. */
74 /* Matrix section enums. */
82 /* Diagonal inclusion enums. */
110 /* 0=vector, 1=matrix, 2=scalar. */
111 static const int content_type[PROX + 1] =
113 0, 2, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1,
116 /* Name of each content type. */
117 static const char *content_names[PROX + 1] =
119 "N", "N", "N_MATRIX", "MEAN", "STDDEV", "COUNT", "MSE",
120 "DFE", "MAT", "COV", "CORR", "PROX",
123 /* A MATRIX DATA input program. */
124 struct matrix_data_pgm
126 struct pool *container; /* Arena used for all allocations. */
127 struct dfm_reader *reader; /* Data file to read. */
130 enum format_type fmt; /* LIST or FREE. */
131 enum matrix_section section;/* LOWER or UPPER or FULL. */
132 enum include_diagonal diag; /* DIAGONAL or NODIAGONAL. */
134 int explicit_rowtype; /* ROWTYPE_ specified explicitly in data? */
135 struct variable *rowtype_, *varname_; /* ROWTYPE_, VARNAME_ variables. */
137 struct variable *single_split; /* Single SPLIT FILE variable. */
139 /* Factor variables. */
140 size_t n_factors; /* Number of factor variables. */
141 struct variable **factors; /* Factor variables. */
142 int is_per_factor[PROX + 1]; /* Is there per-factor data? */
144 int cells; /* Number of cells, or -1 if none. */
146 int pop_n; /* Population N specified by user. */
148 /* CONTENTS subcommand. */
149 int contents[EOC * 3 + 1]; /* Contents. */
150 int n_contents; /* Number of entries. */
152 /* Continuous variables. */
153 int n_continuous; /* Number of continuous variables. */
154 int first_continuous; /* Index into default_dict.var of
155 first continuous variable. */
158 /* Auxiliary data attached to MATRIX DATA variables. */
161 int var_type; /* Variable type. */
162 int sub_type; /* Subtype. */
165 static const struct case_source_class matrix_data_with_rowtype_source_class;
166 static const struct case_source_class matrix_data_without_rowtype_source_class;
168 static int compare_variables_by_mxd_var_type (const void *pa,
170 static bool read_matrices_without_rowtype (struct matrix_data_pgm *);
171 static bool read_matrices_with_rowtype (struct matrix_data_pgm *);
172 static int string_to_content_type (char *, int *);
173 static void attach_mxd_aux (struct variable *, int var_type, int sub_type);
176 cmd_matrix_data (void)
179 struct matrix_data_pgm *mx;
180 struct file_handle *fh = fh_inline_file ();
185 discard_variables ();
187 pool = pool_create ();
188 mx = pool_alloc (pool, sizeof *mx);
189 mx->container = pool;
194 mx->explicit_rowtype = 0;
197 mx->single_split = NULL;
200 memset (mx->is_per_factor, 0, sizeof mx->is_per_factor);
204 mx->n_continuous = 0;
205 mx->first_continuous = 0;
210 if (lex_match_id ("VARIABLES"))
217 msg (SE, _("VARIABLES subcommand multiply specified."));
223 if (!parse_DATA_LIST_vars (&v, &nv, PV_NO_DUPLICATE))
229 for (i = 0; i < nv; i++)
230 if (!strcasecmp (v[i], "VARNAME_"))
232 msg (SE, _("VARNAME_ cannot be explicitly specified on "
234 for (i = 0; i < nv; i++)
244 for (i = 0; i < nv; i++)
246 struct variable *new_var;
248 if (strcasecmp (v[i], "ROWTYPE_"))
250 new_var = dict_create_var_assert (default_dict, v[i], 0);
251 attach_mxd_aux (new_var, MXD_CONTINUOUS, i);
254 mx->explicit_rowtype = 1;
260 mx->rowtype_ = dict_create_var_assert (default_dict,
262 attach_mxd_aux (mx->rowtype_, MXD_ROWTYPE, 0);
264 else if (lex_match_id ("FILE"))
267 fh = fh_parse (FH_REF_FILE | FH_REF_INLINE);
271 else if (lex_match_id ("FORMAT"))
275 while (token == T_ID)
277 if (lex_match_id ("LIST"))
279 else if (lex_match_id ("FREE"))
281 else if (lex_match_id ("LOWER"))
283 else if (lex_match_id ("UPPER"))
285 else if (lex_match_id ("FULL"))
287 else if (lex_match_id ("DIAGONAL"))
289 else if (lex_match_id ("NODIAGONAL"))
290 mx->diag = NODIAGONAL;
293 lex_error (_("in FORMAT subcommand"));
298 else if (lex_match_id ("SPLIT"))
304 msg (SE, _("SPLIT subcommand multiply specified."));
311 lex_error (_("in SPLIT subcommand"));
315 if (dict_lookup_var (default_dict, tokid) == NULL
316 && (lex_look_ahead () == '.' || lex_look_ahead () == '/'))
318 if (!strcasecmp (tokid, "ROWTYPE_")
319 || !strcasecmp (tokid, "VARNAME_"))
321 msg (SE, _("Split variable may not be named ROWTYPE_ "
326 mx->single_split = dict_create_var_assert (default_dict,
328 attach_mxd_aux (mx->single_split, MXD_CONTINUOUS, 0);
331 dict_set_split_vars (default_dict, &mx->single_split, 1);
335 struct variable **split;
338 if (!parse_variables (default_dict, &split, &n, PV_NO_DUPLICATE))
341 dict_set_split_vars (default_dict, split, n);
345 struct variable *const *split = dict_get_split_vars (default_dict);
346 size_t split_cnt = dict_get_split_cnt (default_dict);
349 for (i = 0; i < split_cnt; i++)
351 struct mxd_var *mv = split[i]->aux;
353 if (mv->var_type != MXD_CONTINUOUS)
355 msg (SE, _("Split variable %s is already another type."),
359 var_clear_aux (split[i]);
360 attach_mxd_aux (split[i], MXD_SPLIT, i);
364 else if (lex_match_id ("FACTORS"))
370 msg (SE, _("FACTORS subcommand multiply specified."));
375 if (!parse_variables (default_dict, &mx->factors, &mx->n_factors,
382 for (i = 0; i < mx->n_factors; i++)
384 struct variable *v = mx->factors[i];
385 struct mxd_var *mv = v->aux;
387 if (mv->var_type != MXD_CONTINUOUS)
389 msg (SE, _("Factor variable %s is already another type."),
394 attach_mxd_aux (v, MXD_FACTOR, i);
398 else if (lex_match_id ("CELLS"))
404 msg (SE, _("CELLS subcommand multiply specified."));
408 if (!lex_is_integer () || lex_integer () < 1)
410 lex_error (_("expecting positive integer"));
414 mx->cells = lex_integer ();
417 else if (lex_match_id ("N"))
423 msg (SE, _("N subcommand multiply specified."));
427 if (!lex_is_integer () || lex_integer () < 1)
429 lex_error (_("expecting positive integer"));
433 mx->pop_n = lex_integer ();
436 else if (lex_match_id ("CONTENTS"))
438 int inside_parens = 0;
439 unsigned collide = 0;
444 msg (SE, _("CONTENTS subcommand multiply specified."));
454 for (i = 0; i <= PROX; i++)
455 mx->is_per_factor[i] = 0;
464 msg (SE, _("Nested parentheses not allowed."));
470 else if (lex_match (')'))
474 msg (SE, _("Mismatched right parenthesis (`(')."));
477 if (mx->contents[mx->n_contents - 1] == LPAREN)
479 msg (SE, _("Empty parentheses not allowed."));
492 lex_error (_("in CONTENTS subcommand"));
496 content_type = string_to_content_type (tokid,
498 if (content_type == -1)
500 lex_error (_("in CONTENTS subcommand"));
505 if (collide & (1 << collide_index))
507 msg (SE, _("Content multiply specified for %s."),
508 content_names[content_type]);
511 collide |= (1 << collide_index);
514 mx->is_per_factor[item] = inside_parens;
516 mx->contents[mx->n_contents++] = item;
518 if (token == '/' || token == '.')
524 msg (SE, _("Missing right parenthesis."));
527 mx->contents[mx->n_contents] = EOC;
538 lex_error (_("expecting end of command"));
544 msg (SE, _("Missing VARIABLES subcommand."));
548 if (!mx->n_contents && !mx->explicit_rowtype)
550 msg (SW, _("CONTENTS subcommand not specified: assuming file "
551 "contains only CORR matrix."));
553 mx->contents[0] = CORR;
554 mx->contents[1] = EOC;
558 if (mx->n_factors && !mx->explicit_rowtype && mx->cells == -1)
560 msg (SE, _("Missing CELLS subcommand. CELLS is required "
561 "when ROWTYPE_ is not given in the data and "
562 "factors are present."));
566 if (mx->explicit_rowtype && mx->single_split)
568 msg (SE, _("Split file values must be present in the data when "
569 "ROWTYPE_ is present."));
573 /* Create VARNAME_. */
574 mx->varname_ = dict_create_var_assert (default_dict, "VARNAME_", 8);
575 attach_mxd_aux (mx->varname_, MXD_VARNAME, 0);
577 /* Sort the dictionary variables into the desired order for the
578 system file output. */
583 dict_get_vars (default_dict, &v, &nv, 0);
584 qsort (v, nv, sizeof *v, compare_variables_by_mxd_var_type);
585 dict_reorder_vars (default_dict, v, nv);
591 static const struct fmt_spec fmt_tab[MXD_COUNT] =
602 mx->first_continuous = -1;
603 for (i = 0; i < dict_get_var_cnt (default_dict); i++)
605 struct variable *v = dict_get_var (default_dict, i);
606 struct mxd_var *mv = v->aux;
607 int type = mv->var_type;
609 assert (type >= 0 && type < MXD_COUNT);
610 v->print = v->write = fmt_tab[type];
612 if (type == MXD_CONTINUOUS)
614 if (mx->first_continuous == -1 && type == MXD_CONTINUOUS)
615 mx->first_continuous = i;
619 if (mx->n_continuous == 0)
621 msg (SE, _("No continuous variables specified."));
625 mx->reader = dfm_open_reader (fh);
626 if (mx->reader == NULL)
629 if (mx->explicit_rowtype)
630 ok = read_matrices_with_rowtype (mx);
632 ok = read_matrices_without_rowtype (mx);
634 dfm_close_reader (mx->reader);
636 pool_destroy (mx->container);
638 return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
641 discard_variables ();
643 pool_destroy (mx->container);
644 return CMD_CASCADING_FAILURE;
647 /* Look up string S as a content-type name and return the
648 corresponding enumerated value, or -1 if there is no match. If
649 COLLIDE is non-NULL then *COLLIDE returns a value (suitable for use
650 as a bit-index) which can be used for determining whether a related
651 statistic has already been used. */
653 string_to_content_type (char *s, int *collide)
664 {N_VECTOR, 0, "N_VECTOR"},
666 {N_SCALAR, 0, "N_SCALAR"},
667 {N_MATRIX, 1, "N_MATRIX"},
669 {STDDEV, 3, "STDDEV"},
681 for (tp = tab; tp->value != -1; tp++)
682 if (!strcasecmp (s, tp->string))
685 *collide = tp->collide;
692 /* Compare two variables using p.mxd.var_type and p.mxd.sub_type
695 compare_variables_by_mxd_var_type (const void *a_, const void *b_)
697 struct variable *const *pa = a_;
698 struct variable *const *pb = b_;
699 const struct mxd_var *a = (*pa)->aux;
700 const struct mxd_var *b = (*pb)->aux;
702 if (a->var_type != b->var_type)
703 return a->var_type > b->var_type ? 1 : -1;
705 return a->sub_type < b->sub_type ? -1 : a->sub_type > b->sub_type;
708 /* Attaches a struct mxd_var with the specific member values to
711 attach_mxd_aux (struct variable *v, int var_type, int sub_type)
715 assert (v->aux == NULL);
716 mv = xmalloc (sizeof *mv);
717 mv->var_type = var_type;
718 mv->sub_type = sub_type;
719 var_attach_aux (v, mv, var_dtor_free);
722 /* Matrix tokenizer. */
724 /* Matrix token types. */
725 enum matrix_token_type
731 /* A MATRIX DATA parsing token. */
734 enum matrix_token_type type;
735 double number; /* MNUM: token value. */
736 char *string; /* MSTR: token string; not null-terminated. */
737 int length; /* MSTR: tokstr length. */
740 static int mget_token (struct matrix_token *, struct dfm_reader *);
743 #define mget_token(TOKEN, READER) mget_token_dump(TOKEN, READER)
746 mdump_token (const struct matrix_token *token)
751 printf (" #%g", token->number);
754 printf (" '%.*s'", token->length, token->string);
763 mget_token_dump (struct matrix_token *token, struct dfm_reader *reader)
765 int result = (mget_token) (token, reader);
771 /* Return the current position in READER. */
773 context (struct dfm_reader *reader)
775 static struct string buf = DS_EMPTY_INITIALIZER;
778 if (dfm_eof (reader))
779 ds_assign_cstr (&buf, "at end of file");
784 p = dfm_get_record (reader);
785 ss_ltrim (&p, ss_cstr (CC_SPACES));
787 ds_assign_cstr (&buf, "at end of line");
789 ds_put_format (&buf, "before `%.*s'",
790 (int) ss_cspan (p, ss_cstr (CC_SPACES)), ss_data (p));
793 return ds_cstr (&buf);
796 /* Is there at least one token left in the data file? */
798 another_token (struct dfm_reader *reader)
805 if (dfm_eof (reader))
808 p = dfm_get_record (reader);
809 space_cnt = ss_span (p, ss_cstr (CC_SPACES));
810 if (space_cnt < ss_length (p))
812 dfm_forward_columns (reader, space_cnt);
816 dfm_forward_record (reader);
820 /* Parse a MATRIX DATA token from READER into TOKEN. */
822 (mget_token) (struct matrix_token *token, struct dfm_reader *reader)
824 struct substring line, p;
828 if (!another_token (reader))
831 line = p = dfm_get_record (reader);
833 /* Three types of fields: quoted with ', quoted with ", unquoted. */
835 if (c == '\'' || c == '"')
838 if (!ss_get_until (&p, c, &s))
839 msg (SW, _("Scope of string exceeds line."));
843 bool is_num = isdigit (c) || c == '.';
844 const char *start = ss_data (p);
849 if (strchr (CC_SPACES ",-+", c) != NULL)
854 if (strchr ("deDE", c) && strchr ("+-", ss_at (p, 1)))
862 s = ss_buffer (start, ss_data (p) - start);
870 di.v = (union value *) &token->number;
871 di.f1 = dfm_get_column (reader, di.s);
872 di.format = make_output_format (FMT_F, token->length, 0);
879 token->string = ss_data (s);
880 token->length = ss_length (s);
882 dfm_reread_record (reader, dfm_get_column (reader, ss_end (s)));
887 /* Forcibly skip the end of a line for content type CONTENT in
890 force_eol (struct dfm_reader *reader, const char *content)
894 if (dfm_eof (reader))
897 p = dfm_get_record (reader);
898 if (ss_span (p, ss_cstr (CC_SPACES)) != ss_length (p))
900 msg (SE, _("End of line expected %s while reading %s."),
901 context (reader), content);
905 dfm_forward_record (reader);
909 /* Back end, omitting ROWTYPE_. */
913 struct matrix_data_pgm *mx; /* MATRIX DATA program. */
914 double ***data; /* MATRIX DATA data. */
915 double *factor_values; /* Factor values. */
916 int max_cell_idx; /* Max-numbered cell that we have
917 read so far, plus one. */
918 double *split_values; /* SPLIT FILE variable values. */
921 static int nr_read_splits (struct nr_aux_data *, int compare);
922 static int nr_read_factors (struct nr_aux_data *, int cell);
923 static bool nr_output_data (struct nr_aux_data *, struct ccase *,
924 write_case_func *, write_case_data);
925 static bool matrix_data_read_without_rowtype (struct case_source *source,
930 /* Read from the data file and write it to the active file.
931 Returns true if successful, false if an I/O error occurred. */
933 read_matrices_without_rowtype (struct matrix_data_pgm *mx)
935 struct nr_aux_data nr;
943 nr.factor_values = xnmalloc (mx->n_factors * mx->cells,
944 sizeof *nr.factor_values);
946 nr.split_values = xnmalloc (dict_get_split_cnt (default_dict),
947 sizeof *nr.split_values);
949 proc_set_source (create_case_source (
950 &matrix_data_without_rowtype_source_class, &nr));
952 ok = procedure (NULL, NULL);
954 free (nr.split_values);
955 free (nr.factor_values);
960 /* Mirror data across the diagonal of matrix CP which contains
961 CONTENT type data. */
963 fill_matrix (struct matrix_data_pgm *mx, int content, double *cp)
965 int type = content_type[content];
967 if (type == 1 && mx->section != FULL)
969 if (mx->diag == NODIAGONAL)
971 const double fill = content == CORR ? 1.0 : SYSMIS;
974 for (i = 0; i < mx->n_continuous; i++)
975 cp[i * (1 + mx->n_continuous)] = fill;
981 if (mx->section == LOWER)
983 int n_lines = mx->n_continuous;
984 if (mx->section != FULL && mx->diag == NODIAGONAL)
987 for (r = 1; r < n_lines; r++)
988 for (c = 0; c < r; c++)
989 cp[r + c * mx->n_continuous] = cp[c + r * mx->n_continuous];
993 assert (mx->section == UPPER);
994 for (r = 1; r < mx->n_continuous; r++)
995 for (c = 0; c < r; c++)
996 cp[c + r * mx->n_continuous] = cp[r + c * mx->n_continuous];
1004 for (c = 1; c < mx->n_continuous; c++)
1009 /* Read data lines for content type CONTENT from the data file.
1010 If PER_FACTOR is nonzero, then factor information is read from
1011 the data file. Data is for cell number CELL. */
1013 nr_read_data_lines (struct nr_aux_data *nr,
1014 int per_factor, int cell, int content, int compare)
1016 struct matrix_data_pgm *mx = nr->mx;
1017 const int type = content_type[content]; /* Content type. */
1018 int n_lines; /* Number of lines to parse from data file for this type. */
1019 double *cp; /* Current position in vector or matrix. */
1026 n_lines = mx->n_continuous;
1027 if (mx->section != FULL && mx->diag == NODIAGONAL)
1031 cp = nr->data[content][cell];
1032 if (type == 1 && mx->section == LOWER && mx->diag == NODIAGONAL)
1033 cp += mx->n_continuous;
1035 for (i = 0; i < n_lines; i++)
1039 if (!nr_read_splits (nr, 1))
1041 if (per_factor && !nr_read_factors (nr, cell))
1048 n_cols = mx->n_continuous;
1051 switch (mx->section)
1058 n_cols = mx->n_continuous - i;
1059 if (mx->diag == NODIAGONAL)
1066 n_cols = mx->n_continuous;
1084 for (j = 0; j < n_cols; j++)
1086 struct matrix_token token;
1087 if (!mget_token (&token, mx->reader))
1089 if (token.type != MNUM)
1091 msg (SE, _("expecting value for %s %s"),
1092 dict_get_var (default_dict, j)->name,
1093 context (mx->reader));
1097 *cp++ = token.number;
1100 && !force_eol (mx->reader, content_names[content]))
1104 if (mx->section == LOWER)
1105 cp += mx->n_continuous - n_cols;
1108 fill_matrix (mx, content, nr->data[content][cell]);
1113 /* When ROWTYPE_ does not appear in the data, reads the matrices and
1114 writes them to the output file.
1115 Returns true if successful, false if an I/O error occurred. */
1117 matrix_data_read_without_rowtype (struct case_source *source,
1119 write_case_func *write_case,
1120 write_case_data wc_data)
1122 struct nr_aux_data *nr = source->aux;
1123 struct matrix_data_pgm *mx = nr->mx;
1128 nr->data = pool_nalloc (mx->container, PROX + 1, sizeof *nr->data);
1133 for (i = 0; i <= PROX; i++)
1137 for (cp = mx->contents; *cp != EOC; cp++)
1138 if (*cp != LPAREN && *cp != RPAREN)
1140 int per_factor = mx->is_per_factor[*cp];
1143 n_entries = mx->n_continuous;
1144 if (content_type[*cp] == 1)
1145 n_entries *= mx->n_continuous;
1148 int n_vectors = per_factor ? mx->cells : 1;
1151 nr->data[*cp] = pool_nalloc (mx->container,
1152 n_vectors, sizeof **nr->data);
1154 for (i = 0; i < n_vectors; i++)
1155 nr->data[*cp][i] = pool_nalloc (mx->container,
1156 n_entries, sizeof ***nr->data);
1165 if (!nr_read_splits (nr, 0))
1168 for (bp = mx->contents; *bp != EOC; bp = np)
1172 /* Trap the CONTENTS that we should parse in this pass
1173 between bp and ep. Set np to the starting bp for next
1178 while (*ep != RPAREN)
1186 while (*ep != EOC && *ep != LPAREN)
1195 for (i = 0; i < (per_factor ? mx->cells : 1); i++)
1199 for (cp = bp; cp < ep; cp++)
1200 if (!nr_read_data_lines (nr, per_factor, i, *cp, cp != bp))
1206 if (!nr_output_data (nr, c, write_case, wc_data))
1209 if (dict_get_split_cnt (default_dict) == 0
1210 || !another_token (mx->reader))
1215 /* Read the split file variables. If COMPARE is 1, compares the
1216 values read to the last values read and returns 1 if they're equal,
1219 nr_read_splits (struct nr_aux_data *nr, int compare)
1221 struct matrix_data_pgm *mx = nr->mx;
1222 static int just_read = 0; /* FIXME: WTF? */
1226 if (compare && just_read)
1232 if (dict_get_split_vars (default_dict) == NULL)
1235 if (mx->single_split)
1239 struct mxd_var *mv = dict_get_split_vars (default_dict)[0]->aux;
1240 nr->split_values[0] = ++mv->sub_type;
1248 split_cnt = dict_get_split_cnt (default_dict);
1249 for (i = 0; i < split_cnt; i++)
1251 struct matrix_token token;
1252 if (!mget_token (&token, mx->reader))
1254 if (token.type != MNUM)
1256 msg (SE, _("Syntax error expecting SPLIT FILE value %s."),
1257 context (mx->reader));
1262 nr->split_values[i] = token.number;
1263 else if (nr->split_values[i] != token.number)
1265 msg (SE, _("Expecting value %g for %s."),
1266 nr->split_values[i],
1267 dict_get_split_vars (default_dict)[i]->name);
1275 /* Read the factors for cell CELL. If COMPARE is 1, compares the
1276 values read to the last values read and returns 1 if they're equal,
1279 nr_read_factors (struct nr_aux_data *nr, int cell)
1281 struct matrix_data_pgm *mx = nr->mx;
1284 if (mx->n_factors == 0)
1287 assert (nr->max_cell_idx >= cell);
1288 if (cell != nr->max_cell_idx)
1299 for (i = 0; i < mx->n_factors; i++)
1301 struct matrix_token token;
1302 if (!mget_token (&token, mx->reader))
1304 if (token.type != MNUM)
1306 msg (SE, _("Syntax error expecting factor value %s."),
1307 context (mx->reader));
1312 nr->factor_values[i + mx->n_factors * cell] = token.number;
1313 else if (nr->factor_values[i + mx->n_factors * cell] != token.number)
1315 msg (SE, _("Syntax error expecting value %g for %s %s."),
1316 nr->factor_values[i + mx->n_factors * cell],
1317 mx->factors[i]->name, context (mx->reader));
1326 /* Write the contents of a cell having content type CONTENT and data
1327 CP to the active file.
1328 Returns true if successful, false if an I/O error occurred. */
1330 dump_cell_content (struct matrix_data_pgm *mx, int content, double *cp,
1332 write_case_func *write_case, write_case_data wc_data)
1334 int type = content_type[content];
1337 buf_copy_str_rpad (case_data_rw (c, mx->rowtype_->fv)->s, 8,
1338 content_names[content]);
1341 memset (case_data_rw (c, mx->varname_->fv)->s, ' ', 8);
1345 int n_lines = (type == 1) ? mx->n_continuous : 1;
1348 for (i = 0; i < n_lines; i++)
1352 for (j = 0; j < mx->n_continuous; j++)
1354 int fv = dict_get_var (default_dict, mx->first_continuous + j)->fv;
1355 case_data_rw (c, fv)->f = *cp;
1359 buf_copy_str_rpad (case_data_rw (c, mx->varname_->fv)->s, 8,
1360 dict_get_var (default_dict,
1361 mx->first_continuous + i)->name);
1362 if (!write_case (wc_data))
1369 /* Finally dump out everything from nr_data[] to the output file. */
1371 nr_output_data (struct nr_aux_data *nr, struct ccase *c,
1372 write_case_func *write_case, write_case_data wc_data)
1374 struct matrix_data_pgm *mx = nr->mx;
1377 struct variable *const *split;
1381 split_cnt = dict_get_split_cnt (default_dict);
1382 split = dict_get_split_vars (default_dict);
1383 for (i = 0; i < split_cnt; i++)
1384 case_data_rw (c, split[i]->fv)->f = nr->split_values[i];
1391 for (cell = 0; cell < mx->cells; cell++)
1396 for (factor = 0; factor < mx->n_factors; factor++)
1397 case_data_rw (c, mx->factors[factor]->fv)->f
1398 = nr->factor_values[factor + cell * mx->n_factors];
1404 for (content = 0; content <= PROX; content++)
1405 if (mx->is_per_factor[content])
1407 assert (nr->data[content] != NULL
1408 && nr->data[content][cell] != NULL);
1410 if (!dump_cell_content (mx, content, nr->data[content][cell],
1411 c, write_case, wc_data))
1424 for (factor = 0; factor < mx->n_factors; factor++)
1425 case_data_rw (c, mx->factors[factor]->fv)->f = SYSMIS;
1428 for (content = 0; content <= PROX; content++)
1429 if (!mx->is_per_factor[content] && nr->data[content] != NULL)
1431 if (!dump_cell_content (mx, content, nr->data[content][0],
1432 c, write_case, wc_data))
1440 /* Back end, with ROWTYPE_. */
1442 /* All the data for one set of factor values. */
1446 int n_rows[PROX + 1];
1447 double *data[PROX + 1];
1448 struct factor_data *next;
1451 /* With ROWTYPE_ auxiliary data. */
1454 struct matrix_data_pgm *mx; /* MATRIX DATA program. */
1455 int content; /* Type of current row. */
1456 double *split_values; /* SPLIT FILE variable values. */
1457 struct factor_data *data; /* All the data. */
1458 struct factor_data *current; /* Current factor. */
1461 static int wr_read_splits (struct wr_aux_data *, struct ccase *,
1462 write_case_func *, write_case_data);
1463 static bool wr_output_data (struct wr_aux_data *, struct ccase *,
1464 write_case_func *, write_case_data);
1465 static int wr_read_rowtype (struct wr_aux_data *,
1466 const struct matrix_token *, struct dfm_reader *);
1467 static int wr_read_factors (struct wr_aux_data *);
1468 static int wr_read_indeps (struct wr_aux_data *);
1469 static bool matrix_data_read_with_rowtype (struct case_source *,
1474 /* When ROWTYPE_ appears in the data, reads the matrices and writes
1475 them to the output file.
1476 Returns true if successful, false if an I/O error occurred. */
1478 read_matrices_with_rowtype (struct matrix_data_pgm *mx)
1480 struct wr_aux_data wr;
1485 wr.split_values = NULL;
1490 proc_set_source (create_case_source (&matrix_data_with_rowtype_source_class,
1492 ok = procedure (NULL, NULL);
1494 free (wr.split_values);
1498 /* Read from the data file and write it to the active file.
1499 Returns true if successful, false if an I/O error occurred. */
1501 matrix_data_read_with_rowtype (struct case_source *source,
1503 write_case_func *write_case,
1504 write_case_data wc_data)
1506 struct wr_aux_data *wr = source->aux;
1507 struct matrix_data_pgm *mx = wr->mx;
1511 if (!wr_read_splits (wr, c, write_case, wc_data))
1514 if (!wr_read_factors (wr))
1517 if (!wr_read_indeps (wr))
1520 while (another_token (mx->reader));
1522 return wr_output_data (wr, c, write_case, wc_data);
1525 /* Read the split file variables. If they differ from the previous
1526 set of split variables then output the data. Returns success. */
1528 wr_read_splits (struct wr_aux_data *wr,
1530 write_case_func *write_case, write_case_data wc_data)
1532 struct matrix_data_pgm *mx = wr->mx;
1536 split_cnt = dict_get_split_cnt (default_dict);
1540 if (wr->split_values)
1545 wr->split_values = xnmalloc (split_cnt, sizeof *wr->split_values);
1552 for (i = 0; i < split_cnt; i++)
1554 struct matrix_token token;
1555 if (!mget_token (&token, mx->reader))
1557 if (token.type != MNUM)
1559 msg (SE, _("Syntax error %s expecting SPLIT FILE value."),
1560 context (mx->reader));
1564 if (compare && wr->split_values[i] != token.number && !different)
1566 if (!wr_output_data (wr, c, write_case, wc_data))
1571 wr->split_values[i] = token.number;
1578 /* Compares doubles A and B, treating SYSMIS as greatest. */
1580 compare_doubles (const void *a_, const void *b_, void *aux UNUSED)
1582 const double *a = a_;
1583 const double *b = b_;
1587 else if (*a == SYSMIS)
1589 else if (*b == SYSMIS)
1597 /* Return strcmp()-type comparison of the MX->n_factors factors at _A and
1598 _B. Sort missing values toward the end. */
1600 compare_factors (const void *a_, const void *b_, void *mx_)
1602 struct matrix_data_pgm *mx = mx_;
1603 struct factor_data *const *pa = a_;
1604 struct factor_data *const *pb = b_;
1605 const double *a = (*pa)->factors;
1606 const double *b = (*pb)->factors;
1608 return lexicographical_compare_3way (a, mx->n_factors,
1611 compare_doubles, NULL);
1614 /* Write out the data for the current split file to the active
1616 Returns true if successful, false if an I/O error occurred. */
1618 wr_output_data (struct wr_aux_data *wr,
1620 write_case_func *write_case, write_case_data wc_data)
1622 struct matrix_data_pgm *mx = wr->mx;
1626 struct variable *const *split;
1630 split_cnt = dict_get_split_cnt (default_dict);
1631 split = dict_get_split_vars (default_dict);
1632 for (i = 0; i < split_cnt; i++)
1633 case_data_rw (c, split[i]->fv)->f = wr->split_values[i];
1636 /* Sort the wr->data list. */
1638 struct factor_data **factors;
1639 struct factor_data *iter;
1642 factors = xnmalloc (mx->cells, sizeof *factors);
1644 for (i = 0, iter = wr->data; iter; iter = iter->next, i++)
1647 sort (factors, mx->cells, sizeof *factors, compare_factors, mx);
1649 wr->data = factors[0];
1650 for (i = 0; i < mx->cells - 1; i++)
1651 factors[i]->next = factors[i + 1];
1652 factors[mx->cells - 1]->next = NULL;
1657 /* Write out records for every set of factor values. */
1659 struct factor_data *iter;
1661 for (iter = wr->data; iter; iter = iter->next)
1666 for (factor = 0; factor < mx->n_factors; factor++)
1667 case_data_rw (c, mx->factors[factor]->fv)->f
1668 = iter->factors[factor];
1674 for (content = 0; content <= PROX; content++)
1676 if (!iter->n_rows[content])
1680 int type = content_type[content];
1681 int n_lines = (type == 1
1683 - (mx->section != FULL && mx->diag == NODIAGONAL))
1686 if (n_lines != iter->n_rows[content])
1688 msg (SE, _("Expected %d lines of data for %s content; "
1689 "actually saw %d lines. No data will be "
1690 "output for this content."),
1691 n_lines, content_names[content],
1692 iter->n_rows[content]);
1697 fill_matrix (mx, content, iter->data[content]);
1699 ok = dump_cell_content (mx, content, iter->data[content],
1700 c, write_case, wc_data);
1708 pool_destroy (mx->container);
1709 mx->container = pool_create ();
1711 wr->data = wr->current = NULL;
1716 /* Sets ROWTYPE_ based on the given TOKEN read from READER.
1719 wr_read_rowtype (struct wr_aux_data *wr,
1720 const struct matrix_token *token,
1721 struct dfm_reader *reader)
1723 if (wr->content != -1)
1725 msg (SE, _("Multiply specified ROWTYPE_ %s."), context (reader));
1728 if (token->type != MSTR)
1730 msg (SE, _("Syntax error %s expecting ROWTYPE_ string."),
1739 memcpy (s, token->string, min (15, token->length));
1740 s[min (15, token->length)] = 0;
1742 for (cp = s; *cp; cp++)
1743 *cp = toupper ((unsigned char) *cp);
1745 wr->content = string_to_content_type (s, NULL);
1748 if (wr->content == -1)
1750 msg (SE, _("Syntax error %s."), context (reader));
1757 /* Read the factors for the current row. Select a set of factors and
1758 point wr_current to it. */
1760 wr_read_factors (struct wr_aux_data *wr)
1762 struct matrix_data_pgm *mx = wr->mx;
1763 double *factor_values = local_alloc (sizeof *factor_values * mx->n_factors);
1769 for (i = 0; i < mx->n_factors; i++)
1771 struct matrix_token token;
1772 if (!mget_token (&token, mx->reader))
1774 if (token.type == MSTR)
1776 if (!wr_read_rowtype (wr, &token, mx->reader))
1778 if (!mget_token (&token, mx->reader))
1781 if (token.type != MNUM)
1783 msg (SE, _("Syntax error expecting factor value %s."),
1784 context (mx->reader));
1788 factor_values[i] = token.number;
1791 if (wr->content == -1)
1793 struct matrix_token token;
1794 if (!mget_token (&token, mx->reader))
1796 if (!wr_read_rowtype (wr, &token, mx->reader))
1800 /* Try the most recent factor first as a simple caching
1806 for (i = 0; i < mx->n_factors; i++)
1807 if (factor_values[i] != wr->current->factors[i])
1812 /* Linear search through the list. */
1815 struct factor_data *iter;
1817 for (iter = wr->data; iter; iter = iter->next)
1821 for (i = 0; i < mx->n_factors; i++)
1822 if (factor_values[i] != iter->factors[i])
1832 /* Not found. Make a new item. */
1834 struct factor_data *new = pool_alloc (mx->container, sizeof *new);
1836 new->factors = pool_nalloc (mx->container,
1837 mx->n_factors, sizeof *new->factors);
1842 for (i = 0; i < mx->n_factors; i++)
1843 new->factors[i] = factor_values[i];
1849 for (i = 0; i <= PROX; i++)
1852 new->data[i] = NULL;
1856 new->next = wr->data;
1857 wr->data = wr->current = new;
1862 local_free (factor_values);
1866 local_free (factor_values);
1870 /* Read the independent variables into wr->current. */
1872 wr_read_indeps (struct wr_aux_data *wr)
1874 struct matrix_data_pgm *mx = wr->mx;
1875 struct factor_data *c = wr->current;
1876 const int type = content_type[wr->content];
1877 const int n_rows = c->n_rows[wr->content];
1881 /* Allocate room for data if necessary. */
1882 if (c->data[wr->content] == NULL)
1884 int n_items = mx->n_continuous;
1886 n_items *= mx->n_continuous;
1888 c->data[wr->content] = pool_nalloc (mx->container,
1889 n_items, sizeof **c->data);
1892 cp = &c->data[wr->content][n_rows * mx->n_continuous];
1894 /* Figure out how much to read from this line. */
1901 msg (SE, _("Duplicate specification for %s."),
1902 content_names[wr->content]);
1906 n_cols = mx->n_continuous;
1911 if (n_rows >= mx->n_continuous - (mx->section != FULL && mx->diag == NODIAGONAL))
1913 msg (SE, _("Too many rows of matrix data for %s."),
1914 content_names[wr->content]);
1918 switch (mx->section)
1921 n_cols = n_rows + 1;
1922 if (mx->diag == NODIAGONAL)
1923 cp += mx->n_continuous;
1927 n_cols = mx->n_continuous - n_rows;
1928 if (mx->diag == NODIAGONAL)
1935 n_cols = mx->n_continuous;
1946 c->n_rows[wr->content]++;
1948 /* Read N_COLS items at CP. */
1952 for (j = 0; j < n_cols; j++)
1954 struct matrix_token token;
1955 if (!mget_token (&token, mx->reader))
1957 if (token.type != MNUM)
1959 msg (SE, _("Syntax error expecting value for %s %s."),
1960 dict_get_var (default_dict, mx->first_continuous + j)->name,
1961 context (mx->reader));
1965 *cp++ = token.number;
1968 && !force_eol (mx->reader, content_names[wr->content]))
1975 /* Matrix source. */
1977 static const struct case_source_class matrix_data_with_rowtype_source_class =
1981 matrix_data_read_with_rowtype,
1985 static const struct case_source_class
1986 matrix_data_without_rowtype_source_class =
1990 matrix_data_read_without_rowtype,