1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 2017 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 #include "data/case.h"
20 #include "data/casereader.h"
21 #include "data/casewriter.h"
22 #include "data/dataset.h"
23 #include "data/dictionary.h"
24 #include "data/format.h"
25 #include "data/transformations.h"
26 #include "data/variable.h"
27 #include "language/command.h"
28 #include "language/data-io/data-parser.h"
29 #include "language/data-io/data-reader.h"
30 #include "language/data-io/file-handle.h"
31 #include "language/data-io/inpt-pgm.h"
32 #include "language/data-io/placement-parser.h"
33 #include "language/lexer/lexer.h"
34 #include "language/lexer/variable-parser.h"
35 #include "libpspp/i18n.h"
36 #include "libpspp/message.h"
37 #include "libpspp/misc.h"
40 #include "gl/xalloc.h"
43 #define _(msgid) gettext (msgid)
45 /* DATA LIST transformation data. */
48 struct data_parser *parser; /* Parser. */
49 struct dfm_reader *reader; /* Data file reader. */
50 struct variable *end; /* Variable specified on END subcommand. */
53 static trns_free_func data_list_trns_free;
54 static trns_proc_func data_list_trns_proc;
69 static const int ROWTYPE_WIDTH = 8;
73 enum triangle triangle;
74 enum diagonal diagonal;
75 const struct variable *rowtype;
76 const struct variable *varname;
77 int n_continuous_vars;
78 struct variable **split_vars;
83 valid rowtype_ values:
100 /* Sets the value of OUTCASE which corresponds to VNAME
101 to the value STR. VNAME must be of type string.
104 set_varname_column (struct ccase *outcase, const struct variable *vname,
107 int len = var_get_width (vname);
108 uint8_t *s = value_str_rw (case_data_rw (outcase, vname), len);
110 strncpy ((char *) s, str, len);
114 blank_varname_column (struct ccase *outcase, const struct variable *vname)
116 int len = var_get_width (vname);
117 uint8_t *s = value_str_rw (case_data_rw (outcase, vname), len);
119 memset (s, ' ', len);
122 static struct casereader *
123 preprocess (struct casereader *casereader0, const struct dictionary *dict, void *aux)
125 struct matrix_format *mformat = aux;
126 const struct caseproto *proto = casereader_get_proto (casereader0);
127 struct casewriter *writer;
128 writer = autopaging_writer_create (proto);
129 struct ccase *prev_case = NULL;
130 double **matrices = NULL;
133 const size_t sizeof_matrix =
134 sizeof (double) * mformat->n_continuous_vars * mformat->n_continuous_vars;
137 /* Make an initial pass to populate our temporary matrix */
138 struct casereader *pass0 = casereader_clone (casereader0);
140 unsigned int prev_split_hash = 1;
141 int row = (mformat->triangle == LOWER && mformat->diagonal == NO_DIAGONAL) ? 1 : 0;
142 for (; (c = casereader_read (pass0)) != NULL; case_unref (c))
145 unsigned int split_hash = 0;
146 for (s = 0; s < mformat->n_split_vars; ++s)
148 const struct variable *svar = mformat->split_vars[s];
149 const union value *sv = case_data (c, svar);
150 split_hash = value_hash (sv, var_get_width (svar), split_hash);
153 if (matrices == NULL || prev_split_hash != split_hash)
155 row = (mformat->triangle == LOWER && mformat->diagonal == NO_DIAGONAL) ?
159 matrices = xrealloc (matrices, sizeof (double*) * n_splits);
160 matrices[n_splits - 1] = xmalloc (sizeof_matrix);
163 prev_split_hash = split_hash;
165 int c_offset = (mformat->triangle == UPPER) ? row : 0;
166 if (mformat->triangle == UPPER && mformat->diagonal == NO_DIAGONAL)
168 const union value *v = case_data (c, mformat->rowtype);
169 const char *val = (const char *) value_str (v, ROWTYPE_WIDTH);
170 if (0 == strncasecmp (val, "corr ", ROWTYPE_WIDTH) ||
171 0 == strncasecmp (val, "cov ", ROWTYPE_WIDTH))
173 if (row >= mformat->n_continuous_vars)
176 _("There are %d variable declared but the data has at least %d matrix rows."),
177 mformat->n_continuous_vars, row + 1);
181 for (col = c_offset; col < mformat->n_continuous_vars; ++col)
183 const struct variable *var =
186 var_get_dict_index (mformat->varname));
188 double e = case_data (c, var)->f;
192 /* Fill in the lower triangle */
193 (matrices[n_splits-1])[col + mformat->n_continuous_vars * row] = e;
195 if (mformat->triangle != FULL)
196 /* Fill in the upper triangle */
197 (matrices[n_splits-1]) [row + mformat->n_continuous_vars * col] = e;
202 casereader_destroy (pass0);
204 /* Now make a second pass to fill in the other triangle from our
206 const int idx = var_get_dict_index (mformat->varname);
211 for (; (c = casereader_read (casereader0)) != NULL; prev_case = c)
214 unsigned int split_hash = 0;
215 for (s = 0; s < mformat->n_split_vars; ++s)
217 const struct variable *svar = mformat->split_vars[s];
218 const union value *sv = case_data (c, svar);
219 split_hash = value_hash (sv, var_get_width (svar), split_hash);
221 if (prev_split_hash != split_hash)
227 prev_split_hash = split_hash;
229 case_unref (prev_case);
230 struct ccase *outcase = case_create (proto);
231 case_copy (outcase, 0, c, 0, caseproto_get_n_widths (proto));
232 const union value *v = case_data (c, mformat->rowtype);
233 const char *val = (const char *) value_str (v, ROWTYPE_WIDTH);
234 if (0 == strncasecmp (val, "corr ", ROWTYPE_WIDTH) ||
235 0 == strncasecmp (val, "cov ", ROWTYPE_WIDTH))
238 const struct variable *var = dict_get_var (dict, idx + 1 + row);
239 set_varname_column (outcase, mformat->varname, var_get_name (var));
240 value_copy (case_data_rw (outcase, mformat->rowtype), v, ROWTYPE_WIDTH);
242 for (col = 0; col < mformat->n_continuous_vars; ++col)
244 union value *dest_val =
245 case_data_rw_idx (outcase,
246 1 + col + var_get_dict_index (mformat->varname));
247 dest_val->f = (matrices[n_splits - 1])[col + mformat->n_continuous_vars * row];
248 if (col == row && mformat->diagonal == NO_DIAGONAL)
255 blank_varname_column (outcase, mformat->varname);
258 /* Special case for SD and N_VECTOR: Rewrite as STDDEV and N respectively */
259 if (0 == strncasecmp (val, "sd ", ROWTYPE_WIDTH))
261 value_copy_buf_rpad (case_data_rw (outcase, mformat->rowtype), ROWTYPE_WIDTH,
262 (uint8_t *) "STDDEV", 6, ' ');
264 else if (0 == strncasecmp (val, "n_vector", ROWTYPE_WIDTH))
266 value_copy_buf_rpad (case_data_rw (outcase, mformat->rowtype), ROWTYPE_WIDTH,
267 (uint8_t *) "N", 1, ' ');
270 casewriter_write (writer, outcase);
273 /* If NODIAGONAL is specified, then a final case must be written */
274 if (mformat->diagonal == NO_DIAGONAL)
277 struct ccase *outcase = case_create (proto);
280 case_copy (outcase, 0, prev_case, 0, caseproto_get_n_widths (proto));
282 const struct variable *var = dict_get_var (dict, idx + 1 + row);
283 set_varname_column (outcase, mformat->varname, var_get_name (var));
285 for (col = 0; col < mformat->n_continuous_vars; ++col)
287 union value *dest_val =
288 case_data_rw_idx (outcase, 1 + col +
289 var_get_dict_index (mformat->varname));
290 dest_val->f = (matrices[n_splits - 1]) [col + mformat->n_continuous_vars * row];
291 if (col == row && mformat->diagonal == NO_DIAGONAL)
295 casewriter_write (writer, outcase);
300 case_unref (prev_case);
303 for (i = 0 ; i < n_splits; ++i)
306 struct casereader *reader1 = casewriter_make_reader (writer);
307 casereader_destroy (casereader0);
313 case_unref (prev_case);
315 for (i = 0 ; i < n_splits; ++i)
318 casereader_destroy (casereader0);
323 cmd_matrix (struct lexer *lexer, struct dataset *ds)
325 struct dictionary *dict;
326 struct data_parser *parser;
327 struct dfm_reader *reader;
328 struct file_handle *fh = NULL;
329 char *encoding = NULL;
330 struct matrix_format mformat;
335 mformat.triangle = LOWER;
336 mformat.diagonal = DIAGONAL;
337 mformat.n_split_vars = 0;
338 mformat.split_vars = NULL;
340 dict = (in_input_program ()
342 : dict_create (get_default_encoding ()));
343 parser = data_parser_create (dict);
346 data_parser_set_type (parser, DP_DELIMITED);
347 data_parser_set_warn_missing_fields (parser, false);
348 data_parser_set_span (parser, false);
350 mformat.rowtype = dict_create_var (dict, "ROWTYPE_", ROWTYPE_WIDTH);
352 mformat.n_continuous_vars = 0;
353 mformat.n_split_vars = 0;
355 if (! lex_force_match_id (lexer, "VARIABLES"))
358 lex_match (lexer, T_EQUALS);
360 if (! parse_mixed_vars (lexer, dict, &names, &n_names, PV_NO_DUPLICATE))
363 for (i = 0; i < n_names; ++i)
369 int longest_name = 0;
370 for (i = 0; i < n_names; ++i)
372 maximize_int (&longest_name, strlen (names[i]));
375 mformat.varname = dict_create_var (dict, "VARNAME_",
376 8 * DIV_RND_UP (longest_name, 8));
378 for (i = 0; i < n_names; ++i)
380 if (0 == strcasecmp (names[i], "ROWTYPE_"))
382 const struct fmt_spec fmt = fmt_for_input (FMT_A, 8, 0);
383 data_parser_add_delimited_field (parser,
385 var_get_case_index (mformat.rowtype),
390 const struct fmt_spec fmt = fmt_for_input (FMT_F, 10, 4);
391 struct variable *v = dict_create_var (dict, names[i], 0);
392 var_set_both_formats (v, &fmt);
393 data_parser_add_delimited_field (parser,
395 var_get_case_index (mformat.varname) +
396 ++mformat.n_continuous_vars,
400 for (i = 0; i < n_names; ++i)
404 while (lex_token (lexer) != T_ENDCMD)
406 if (! lex_force_match (lexer, T_SLASH))
409 if (lex_match_id (lexer, "FORMAT"))
411 lex_match (lexer, T_EQUALS);
413 while (lex_token (lexer) != T_SLASH && (lex_token (lexer) != T_ENDCMD))
415 if (lex_match_id (lexer, "LIST"))
417 data_parser_set_span (parser, false);
419 else if (lex_match_id (lexer, "FREE"))
421 data_parser_set_span (parser, true);
423 else if (lex_match_id (lexer, "UPPER"))
425 mformat.triangle = UPPER;
427 else if (lex_match_id (lexer, "LOWER"))
429 mformat.triangle = LOWER;
431 else if (lex_match_id (lexer, "FULL"))
433 mformat.triangle = FULL;
435 else if (lex_match_id (lexer, "DIAGONAL"))
437 mformat.diagonal = DIAGONAL;
439 else if (lex_match_id (lexer, "NODIAGONAL"))
441 mformat.diagonal = NO_DIAGONAL;
445 lex_error (lexer, NULL);
450 else if (lex_match_id (lexer, "FILE"))
452 lex_match (lexer, T_EQUALS);
454 fh = fh_parse (lexer, FH_REF_FILE | FH_REF_INLINE, NULL);
458 else if (lex_match_id (lexer, "SPLIT"))
460 lex_match (lexer, T_EQUALS);
461 if (! parse_variables (lexer, dict, &mformat.split_vars, &mformat.n_split_vars, 0))
463 free (mformat.split_vars);
467 for (i = 0; i < mformat.n_split_vars; ++i)
469 const struct fmt_spec fmt = fmt_for_input (FMT_F, 4, 0);
470 var_set_both_formats (mformat.split_vars[i], &fmt);
472 dict_reorder_vars (dict, mformat.split_vars, mformat.n_split_vars);
473 mformat.n_continuous_vars -= mformat.n_split_vars;
477 lex_error (lexer, NULL);
482 if (mformat.diagonal == NO_DIAGONAL && mformat.triangle == FULL)
484 msg (SE, _("FORMAT = FULL and FORMAT = NODIAGONAL are mutually exclusive."));
489 fh = fh_inline_file ();
490 fh_set_default_handle (fh);
492 if (!data_parser_any_fields (parser))
494 msg (SE, _("At least one variable must be specified."));
498 if (lex_end_of_command (lexer) != CMD_SUCCESS)
501 reader = dfm_open_reader (fh, lexer, encoding);
505 if (in_input_program ())
507 struct data_list_trns *trns = xmalloc (sizeof *trns);
508 trns->parser = parser;
509 trns->reader = reader;
511 add_transformation (ds, data_list_trns_proc, data_list_trns_free, trns);
515 data_parser_make_active_file (parser, ds, reader, dict, preprocess, &mformat);
520 free (mformat.split_vars);
522 return CMD_DATA_LIST;
525 data_parser_destroy (parser);
526 if (!in_input_program ())
530 free (mformat.split_vars);
531 return CMD_CASCADING_FAILURE;
535 /* Input procedure. */
537 /* Destroys DATA LIST transformation TRNS.
538 Returns true if successful, false if an I/O error occurred. */
540 data_list_trns_free (void *trns_)
542 struct data_list_trns *trns = trns_;
543 data_parser_destroy (trns->parser);
544 dfm_close_reader (trns->reader);
549 /* Handle DATA LIST transformation TRNS, parsing data into *C. */
551 data_list_trns_proc (void *trns_, struct ccase **c, casenumber case_num UNUSED)
553 struct data_list_trns *trns = trns_;
556 *c = case_unshare (*c);
557 if (data_parser_parse (trns->parser, trns->reader, *c))
558 retval = TRNS_CONTINUE;
559 else if (dfm_reader_error (trns->reader) || dfm_eof (trns->reader) > 1)
561 /* An I/O error, or encountering end of file for a second
562 time, should be escalated into a more serious error. */
566 retval = TRNS_END_FILE;
568 /* If there was an END subcommand handle it. */
569 if (trns->end != NULL)
571 double *end = &case_data_rw (*c, trns->end)->f;
572 if (retval == TRNS_END_FILE)
575 retval = TRNS_CONTINUE;