--- /dev/null
+/* PSPP - a program for statistical analysis.
+ Copyright (C) 2017, 2019 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
+ the Free Software Foundation, either version 3 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <config.h>
+
+#include "matrix-reader.h"
+
+#include <stdbool.h>
+#include <math.h>
+
+#include "data/casegrouper.h"
+#include "data/casereader.h"
+#include "data/casewriter.h"
+#include "data/data-out.h"
+#include "data/dataset.h"
+#include "data/dictionary.h"
+#include "data/format.h"
+#include "data/variable.h"
+#include "language/command.h"
+#include "language/lexer/lexer.h"
+#include "libpspp/i18n.h"
+#include "libpspp/message.h"
+#include "libpspp/str.h"
+#include "output/pivot-table.h"
+
+#include "gettext.h"
+#define _(msgid) gettext (msgid)
+#define N_(msgid) msgid
+
+struct lexer;
+
+/*
+This module interprets a "data matrix", typically generated by the command
+MATRIX DATA. The dictionary of such a matrix takes the form:
+
+ s_0, s_1, ... s_m, ROWTYPE_, VARNAME_, v_0, v_1, .... v_n
+
+where s_0, s_1 ... s_m are the variables defining the splits, and
+v_0, v_1 ... v_n are the continuous variables.
+
+m >= 0; n >= 0
+
+The ROWTYPE_ variable is of type A8.
+The VARNAME_ variable is a string type whose width is not predetermined.
+The variables s_x are of type F4.0 (although this reader accepts any type),
+and v_x are of any numeric type.
+
+The values of the ROWTYPE_ variable are in the set {MEAN, STDDEV, N, CORR, COV}
+and determine the purpose of that case.
+The values of the VARNAME_ variable must correspond to the names of the varibles
+in {v_0, v_1 ... v_n} and indicate the rows of the correlation or covariance
+matrices.
+
+
+
+A typical example is as follows:
+
+s_0 ROWTYPE_ VARNAME_ v_0 v_1 v_2
+
+0 MEAN 5.0000 4.0000 3.0000
+0 STDDEV 1.0000 2.0000 3.0000
+0 N 9.0000 9.0000 9.0000
+0 CORR V1 1.0000 .6000 .7000
+0 CORR V2 .6000 1.0000 .8000
+0 CORR V3 .7000 .8000 1.0000
+1 MEAN 9.0000 8.0000 7.0000
+1 STDDEV 5.0000 6.0000 7.0000
+1 N 9.0000 9.0000 9.0000
+1 CORR V1 1.0000 .4000 .3000
+1 CORR V2 .4000 1.0000 .2000
+1 CORR V3 .3000 .2000 1.0000
+
+*/
+
+void
+matrix_material_uninit (struct matrix_material *mm)
+{
+ gsl_matrix_free (mm->corr);
+ gsl_matrix_free (mm->cov);
+ gsl_matrix_free (mm->n);
+ gsl_matrix_free (mm->mean_matrix);
+ gsl_matrix_free (mm->var_matrix);
+}
+\f
+static const struct variable *
+find_matrix_string_var (const struct dictionary *dict, const char *name)
+{
+ const struct variable *var = dict_lookup_var (dict, name);
+ if (var == NULL)
+ {
+ msg (SE, _("Matrix dataset lacks a variable called %s."), name);
+ return NULL;
+ }
+ if (!var_is_alpha (var))
+ {
+ msg (SE, _("Matrix dataset variable %s should be of string type."), name);
+ return NULL;
+ }
+ return var;
+}
+
+struct matrix_reader *
+matrix_reader_create (const struct dictionary *dict,
+ struct casereader *in_reader)
+{
+ const struct variable *varname = find_matrix_string_var (dict, "VARNAME_");
+ const struct variable *rowtype = find_matrix_string_var (dict, "ROWTYPE_");
+ if (!varname || !rowtype)
+ return NULL;
+
+ size_t varname_idx = var_get_dict_index (varname);
+ size_t rowtype_idx = var_get_dict_index (rowtype);
+ if (varname_idx < rowtype_idx)
+ {
+ msg (SE, _("Variable %s must precede %s in matrix file dictionary."),
+ "ROWTYPE_", "VARNAME_");
+ return NULL;
+ }
+
+ for (size_t i = 0; i < dict_get_n_vars (dict); i++)
+ {
+ const struct variable *v = dict_get_var (dict, i);
+ if (!var_is_numeric (v) && v != rowtype && v != varname)
+ {
+ msg (SE, _("Matrix dataset variable %s should be numeric."),
+ var_get_name (v));
+ return NULL;
+ }
+ }
+
+ size_t n_vars;
+ const struct variable **vars;
+ dict_get_vars (dict, &vars, &n_vars, DC_SCRATCH);
+
+ /* Different kinds of variables. */
+ size_t first_svar = 0;
+ size_t n_svars = rowtype_idx;
+ size_t first_fvar = rowtype_idx + 1;
+ size_t n_fvars = varname_idx - rowtype_idx - 1;
+ size_t first_cvar = varname_idx + 1;
+ size_t n_cvars = n_vars - varname_idx - 1;
+ if (!n_cvars)
+ {
+ msg (SE, _("Matrix dataset does not have any continuous variables."));
+ free (vars);
+ return NULL;
+ }
+
+ struct matrix_reader *mr = xmalloc (sizeof *mr);
+ *mr = (struct matrix_reader) {
+ .dict = dict,
+ .grouper = casegrouper_create_vars (in_reader, &vars[first_svar], n_svars),
+ .svars = xmemdup (vars + first_svar, n_svars * sizeof *mr->svars),
+ .n_svars = n_svars,
+ .rowtype = rowtype,
+ .fvars = xmemdup (vars + first_fvar, n_fvars * sizeof *mr->fvars),
+ .n_fvars = n_fvars,
+ .varname = varname,
+ .cvars = xmemdup (vars + first_cvar, n_cvars * sizeof *mr->cvars),
+ .n_cvars = n_cvars,
+ };
+ free (vars);
+
+ return mr;
+}
+
+bool
+matrix_reader_destroy (struct matrix_reader *mr)
+{
+ if (mr == NULL)
+ return false;
+ bool ret = casegrouper_destroy (mr->grouper);
+ free (mr->svars);
+ free (mr->cvars);
+ free (mr->fvars);
+ free (mr);
+ return ret;
+}
+
+
+/*
+ Allocates MATRIX if necessary,
+ and populates row MROW, from the data in C corresponding to
+ variables in VARS. N_VARS is the length of VARS.
+*/
+static void
+matrix_fill_row (gsl_matrix **matrix,
+ const struct ccase *c, int mrow,
+ const struct variable **vars, size_t n_vars)
+{
+ int col;
+ if (*matrix == NULL)
+ {
+ *matrix = gsl_matrix_alloc (n_vars, n_vars);
+ gsl_matrix_set_all (*matrix, SYSMIS);
+ }
+
+ for (col = 0; col < n_vars; ++col)
+ {
+ const struct variable *cv = vars [col];
+ double x = case_num (c, cv);
+ assert (col < (*matrix)->size2);
+ assert (mrow < (*matrix)->size1);
+ gsl_matrix_set (*matrix, mrow, col, x);
+ }
+}
+
+static int
+find_varname (const struct variable **vars, int n_vars,
+ const char *varname)
+{
+ for (int i = 0; i < n_vars; i++)
+ if (!strcasecmp (var_get_name (vars[i]), varname))
+ return i;
+ return -1;
+}
+
+struct substring
+matrix_reader_get_string (const struct ccase *c, const struct variable *var)
+{
+ struct substring s = case_ss (c, var);
+ ss_rtrim (&s, ss_cstr (CC_SPACES));
+ return s;
+}
+
+void
+matrix_reader_set_string (struct ccase *c, const struct variable *var,
+ struct substring src)
+{
+ struct substring dst = case_ss (c, var);
+ for (size_t i = 0; i < dst.length; i++)
+ dst.string[i] = i < src.length ? src.string[i] : ' ';
+}
+
+bool
+matrix_reader_next (struct matrix_material *mm, struct matrix_reader *mr,
+ struct casereader **groupp)
+{
+ struct casereader *group;
+ if (!casegrouper_get_next_group (mr->grouper, &group))
+ {
+ *mm = (struct matrix_material) MATRIX_MATERIAL_INIT;
+ if (groupp)
+ *groupp = NULL;
+ return false;
+ }
+
+ if (groupp)
+ *groupp = casereader_clone (group);
+
+ const struct variable **vars = mr->cvars;
+ size_t n_vars = mr->n_cvars;
+
+ *mm = (struct matrix_material) { .n = NULL };
+
+ struct matrix
+ {
+ const char *name;
+ gsl_matrix **m;
+ size_t good_rows;
+ size_t bad_rows;
+ };
+ struct matrix matrices[] = {
+ { .name = "CORR", .m = &mm->corr },
+ { .name = "COV", .m = &mm->cov },
+ };
+ enum { N_MATRICES = 2 };
+
+ struct ccase *c;
+ for (; (c = casereader_read (group)); case_unref (c))
+ {
+ struct substring rowtype = matrix_reader_get_string (c, mr->rowtype);
+
+ gsl_matrix **v
+ = (ss_equals_case (rowtype, ss_cstr ("N")) ? &mm->n
+ : ss_equals_case (rowtype, ss_cstr ("MEAN")) ? &mm->mean_matrix
+ : ss_equals_case (rowtype, ss_cstr ("STDDEV")) ? &mm->var_matrix
+ : NULL);
+ if (v)
+ {
+ if (!*v)
+ *v = gsl_matrix_calloc (n_vars, n_vars);
+
+ for (int x = 0; x < n_vars; ++x)
+ {
+ double n = case_num (c, vars[x]);
+ if (v == &mm->var_matrix)
+ n *= n;
+ for (int y = 0; y < n_vars; ++y)
+ gsl_matrix_set (*v, y, x, n);
+ }
+ continue;
+ }
+
+ struct matrix *m = NULL;
+ for (size_t i = 0; i < N_MATRICES; i++)
+ if (ss_equals_case (rowtype, ss_cstr (matrices[i].name)))
+ {
+ m = &matrices[i];
+ break;
+ }
+ if (m)
+ {
+ struct substring varname_raw = case_ss (c, mr->varname);
+ struct substring varname = ss_cstr (
+ recode_string (UTF8, dict_get_encoding (mr->dict),
+ varname_raw.string, varname_raw.length));
+ ss_rtrim (&varname, ss_cstr (CC_SPACES));
+ varname.string[varname.length] = '\0';
+
+ int y = find_varname (vars, n_vars, varname.string);
+ if (y >= 0)
+ {
+ m->good_rows++;
+ matrix_fill_row (m->m, c, y, vars, n_vars);
+ }
+ else
+ m->bad_rows++;
+ ss_dealloc (&varname);
+ }
+ }
+ casereader_destroy (group);
+
+ for (size_t i = 0; i < N_MATRICES; i++)
+ if (matrices[i].good_rows && matrices[i].good_rows != n_vars)
+ msg (SW, _("%s matrix has %zu columns but %zu rows named variables "
+ "to be analyzed (and %zu rows named unknown variables)."),
+ matrices[i].name, n_vars, matrices[i].good_rows,
+ matrices[i].bad_rows);
+
+ return true;
+}
+
+int
+cmd_debug_matrix_read (struct lexer *lexer, struct dataset *ds)
+{
+ if (lex_match_id (lexer, "NODATA"))
+ {
+ struct casereader *cr = casewriter_make_reader (
+ mem_writer_create (dict_get_proto (dataset_dict (ds))));
+ struct matrix_reader *mr = matrix_reader_create (dataset_dict (ds), cr);
+ if (!mr)
+ {
+ casereader_destroy (cr);
+ return CMD_FAILURE;
+ }
+ matrix_reader_destroy (mr);
+ return CMD_SUCCESS;
+ }
+
+ struct matrix_reader *mr = matrix_reader_create (dataset_dict (ds),
+ proc_open (ds));
+ if (!mr)
+ return CMD_FAILURE;
+
+ struct pivot_table *pt = pivot_table_create ("Debug Matrix Reader");
+
+ enum mm_stat
+ {
+ MM_CORR,
+ MM_COV,
+ MM_N,
+ MM_MEAN,
+ MM_STDDEV,
+ };
+ const char *mm_stat_names[] = {
+ [MM_CORR] = "Correlation",
+ [MM_COV] = "Covariance",
+ [MM_N] = "N",
+ [MM_MEAN] = "Mean",
+ [MM_STDDEV] = "Standard Deviation",
+ };
+ enum { N_STATS = sizeof mm_stat_names / sizeof *mm_stat_names };
+ for (size_t i = 0; i < 2; i++)
+ {
+ struct pivot_dimension *d = pivot_dimension_create (
+ pt,
+ i ? PIVOT_AXIS_COLUMN : PIVOT_AXIS_ROW,
+ i ? "Column" : "Row");
+ if (!i)
+ pivot_category_create_leaf_rc (d->root, pivot_value_new_text ("Value"),
+ PIVOT_RC_CORRELATION);
+ for (size_t j = 0; j < mr->n_cvars; j++)
+ pivot_category_create_leaf_rc (
+ d->root, pivot_value_new_variable (mr->cvars[j]),
+ PIVOT_RC_CORRELATION);
+ }
+
+ struct pivot_dimension *stat = pivot_dimension_create (pt, PIVOT_AXIS_ROW,
+ "Statistic");
+ for (size_t i = 0; i < N_STATS; i++)
+ pivot_category_create_leaf (stat->root,
+ pivot_value_new_text (mm_stat_names[i]));
+
+ struct pivot_dimension *split = pivot_dimension_create (
+ pt, PIVOT_AXIS_ROW, "Split");
+
+ int split_num = 0;
+
+ struct matrix_material mm = MATRIX_MATERIAL_INIT;
+ while (matrix_reader_next (&mm, mr, NULL))
+ {
+ pivot_category_create_leaf (split->root,
+ pivot_value_new_integer (split_num + 1));
+
+ const gsl_matrix *m[N_STATS] = {
+ [MM_CORR] = mm.corr,
+ [MM_COV] = mm.cov,
+ [MM_N] = mm.n,
+ [MM_MEAN] = mm.mean_matrix,
+ [MM_STDDEV] = mm.var_matrix,
+ };
+
+ for (size_t i = 0; i < N_STATS; i++)
+ if (m[i])
+ {
+ if (i == MM_COV || i == MM_CORR)
+ {
+ for (size_t y = 0; y < mr->n_cvars; y++)
+ for (size_t x = 0; x < mr->n_cvars; x++)
+ pivot_table_put4 (
+ pt, y + 1, x, i, split_num,
+ pivot_value_new_number (gsl_matrix_get (m[i], y, x)));
+ }
+ else
+ for (size_t x = 0; x < mr->n_cvars; x++)
+ {
+ double n = gsl_matrix_get (m[i], 0, x);
+ if (i == MM_STDDEV)
+ n = sqrt (n);
+ pivot_table_put4 (pt, 0, x, i, split_num,
+ pivot_value_new_number (n));
+ }
+ }
+
+ split_num++;
+ matrix_material_uninit (&mm);
+ }
+ pivot_table_submit (pt);
+
+ proc_commit (ds);
+
+ matrix_reader_destroy (mr);
+ return CMD_SUCCESS;
+}
projects / pspp / blobdiff