X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=src%2Flanguage%2Fdata-io%2Fget.c;h=b861aca52ffef7df758d3eabfc506718795081eb;hb=c13375ebe886fc3ed5ce70a260bb2b01cebb4b1f;hp=663d2fc153f3afd008f68b6f387054ec12350e4f;hpb=42489b63e0b4bec2e20c2f55c9791234f7b41764;p=pspp-builds.git diff --git a/src/language/data-io/get.c b/src/language/data-io/get.c index 663d2fc1..b861aca5 100644 --- a/src/language/data-io/get.c +++ b/src/language/data-io/get.c @@ -1,21 +1,18 @@ -/* PSPP - computes sample statistics. - Copyright (C) 1997-9, 2000, 2006 Free Software Foundation, Inc. - Written by Ben Pfaff . +/* PSPP - a program for statistical analysis. + Copyright (C) 1997-9, 2000, 2006, 2007 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 2 of the - License, or (at your option) any later version. + 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. + 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, write to the Free Software - Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA - 02110-1301, USA. */ + along with this program. If not, see . */ #include @@ -23,16 +20,15 @@ #include #include -#include -#include #include -#include -#include +#include +#include +#include +#include #include #include #include #include -#include #include #include #include @@ -41,205 +37,150 @@ #include #include #include -#include #include #include #include #include -#include #include #include +#include + +#include "get-data.h" + +#include "xalloc.h" #include "gettext.h" #define _(msgid) gettext (msgid) -/* Rearranging and reducing a dictionary. */ -static void start_case_map (struct dictionary *); -static struct case_map *finish_case_map (struct dictionary *); -static void map_case (const struct case_map *, - const struct ccase *, struct ccase *); -static void destroy_case_map (struct case_map *); - -static bool parse_dict_trim (struct dictionary *); +static bool parse_dict_trim (struct lexer *, struct dictionary *); /* Reading system and portable files. */ /* Type of command. */ -enum reader_command +enum reader_command { GET_CMD, IMPORT_CMD }; -/* Case reader input program. */ -struct case_reader_pgm - { - struct any_reader *reader; /* File reader. */ - struct case_map *map; /* Map from file dict to active file dict. */ - struct ccase bounce; /* Bounce buffer. */ - }; - -static const struct case_source_class case_reader_source_class; - -static void case_reader_pgm_free (struct case_reader_pgm *); +static void get_translate_case (struct ccase *, struct ccase *, void *map_); +static bool get_destroy_case_map (void *map_); /* Parses a GET or IMPORT command. */ static int -parse_read_command (enum reader_command type) +parse_read_command (struct lexer *lexer, struct dataset *ds, enum reader_command type) { - struct case_reader_pgm *pgm = NULL; + struct casereader *reader = NULL; struct file_handle *fh = NULL; struct dictionary *dict = NULL; + struct case_map *map = NULL; + + if ( type == GET_CMD && lex_match_id (lexer, "DATA") ) + { + return parse_get_data_command (lexer, ds); + } for (;;) { - lex_match ('/'); + lex_match (lexer, '/'); - if (lex_match_id ("FILE") || token == T_STRING) + if (lex_match_id (lexer, "FILE") || lex_token (lexer) == T_STRING) { - lex_match ('='); + lex_match (lexer, '='); - fh = fh_parse (FH_REF_FILE | FH_REF_SCRATCH); + fh = fh_parse (lexer, FH_REF_FILE | FH_REF_SCRATCH); if (fh == NULL) goto error; } - else if (type == IMPORT_CMD && lex_match_id ("TYPE")) + else if (type == IMPORT_CMD && lex_match_id (lexer, "TYPE")) { - lex_match ('='); + lex_match (lexer, '='); - if (lex_match_id ("COMM")) + if (lex_match_id (lexer, "COMM")) type = PFM_COMM; - else if (lex_match_id ("TAPE")) + else if (lex_match_id (lexer, "TAPE")) type = PFM_TAPE; else { - lex_error (_("expecting COMM or TAPE")); + lex_error (lexer, _("expecting COMM or TAPE")); goto error; } } else - break; + break; } - - if (fh == NULL) + + if (fh == NULL) { - lex_sbc_missing ("FILE"); + lex_sbc_missing (lexer, "FILE"); goto error; } - - discard_variables (current_dataset); - - pgm = xmalloc (sizeof *pgm); - pgm->reader = any_reader_open (fh, &dict); - pgm->map = NULL; - case_nullify (&pgm->bounce); - if (pgm->reader == NULL) + + reader = any_reader_open (fh, &dict); + if (reader == NULL) goto error; - case_create (&pgm->bounce, dict_get_next_value_idx (dict)); - - start_case_map (dict); + case_map_prepare_dict (dict); - while (token != '.') + while (lex_token (lexer) != '.') { - lex_match ('/'); - if (!parse_dict_trim (dict)) + lex_match (lexer, '/'); + if (!parse_dict_trim (lexer, dict)) goto error; } + dict_compact_values (dict); - pgm->map = finish_case_map (dict); - - dict_destroy (dataset_dict (current_dataset)); - dataset_set_dict (current_dataset, dict); + map = case_map_from_dict (dict); + if (map != NULL) + reader = casereader_create_translator (reader, + dict_get_next_value_idx (dict), + get_translate_case, + get_destroy_case_map, + map); - proc_set_source (current_dataset, - create_case_source (&case_reader_source_class, pgm)); + proc_set_active_file (ds, reader, dict); return CMD_SUCCESS; error: - case_reader_pgm_free (pgm); + casereader_destroy (reader); if (dict != NULL) dict_destroy (dict); return CMD_CASCADING_FAILURE; } -/* Frees a struct case_reader_pgm. */ -static void -case_reader_pgm_free (struct case_reader_pgm *pgm) -{ - if (pgm != NULL) - { - any_reader_close (pgm->reader); - destroy_case_map (pgm->map); - case_destroy (&pgm->bounce); - free (pgm); - } -} - -/* Clears internal state related to case reader input procedure. */ static void -case_reader_source_destroy (struct case_source *source) +get_translate_case (struct ccase *input, struct ccase *output, + void *map_) { - struct case_reader_pgm *pgm = source->aux; - case_reader_pgm_free (pgm); + struct case_map *map = map_; + case_map_execute (map, input, output); + case_destroy (input); } -/* Reads all the cases from the data file into C and passes them - to WRITE_CASE one by one, passing WC_DATA. - Returns true if successful, false if an I/O error occurred. */ static bool -case_reader_source_read (struct case_source *source, - struct ccase *c, - write_case_func *write_case, write_case_data wc_data) +get_destroy_case_map (void *map_) { - struct case_reader_pgm *pgm = source->aux; - bool ok = true; - - do - { - bool got_case; - if (pgm->map == NULL) - got_case = any_reader_read (pgm->reader, c); - else - { - got_case = any_reader_read (pgm->reader, &pgm->bounce); - if (got_case) - map_case (pgm->map, &pgm->bounce, c); - } - if (!got_case) - break; - - ok = write_case (wc_data); - } - while (ok); - - return ok && !any_reader_error (pgm->reader); + struct case_map *map = map_; + case_map_destroy (map); + return true; } - -static const struct case_source_class case_reader_source_class = - { - "case reader", - NULL, - case_reader_source_read, - case_reader_source_destroy, - }; /* GET. */ int -cmd_get (void) +cmd_get (struct lexer *lexer, struct dataset *ds) { - return parse_read_command (GET_CMD); + return parse_read_command (lexer, ds, GET_CMD); } /* IMPORT. */ int -cmd_import (void) +cmd_import (struct lexer *lexer, struct dataset *ds) { - return parse_read_command (IMPORT_CMD); + return parse_read_command (lexer, ds, IMPORT_CMD); } -/* Writing system and portable files. */ +/* Writing system and portable files. */ /* Type of output file. */ enum writer_type @@ -249,36 +190,12 @@ enum writer_type }; /* Type of a command. */ -enum command_type +enum command_type { XFORM_CMD, /* Transformation. */ PROC_CMD /* Procedure. */ }; -/* File writer plus a case map. */ -struct case_writer - { - struct any_writer *writer; /* File writer. */ - struct case_map *map; /* Map to output file dictionary - (null pointer for identity mapping). */ - struct ccase bounce; /* Bounce buffer for mapping (if needed). */ - }; - -/* Destroys AW. */ -static bool -case_writer_destroy (struct case_writer *aw) -{ - bool ok = true; - if (aw != NULL) - { - ok = any_writer_close (aw->writer); - destroy_case_map (aw->map); - case_destroy (&aw->bounce); - free (aw); - } - return ok; -} - /* Parses SAVE or XSAVE or EXPORT or XEXPORT command. WRITER_TYPE identifies the type of file to write, and COMMAND_TYPE identifies the type of command. @@ -289,15 +206,17 @@ case_writer_destroy (struct case_writer *aw) included. On failure, returns a null pointer. */ -static struct case_writer * -parse_write_command (enum writer_type writer_type, +static struct casewriter * +parse_write_command (struct lexer *lexer, struct dataset *ds, + enum writer_type writer_type, enum command_type command_type, bool *retain_unselected) { /* Common data. */ struct file_handle *handle; /* Output file. */ struct dictionary *dict; /* Dictionary for output file. */ - struct case_writer *aw; /* Writer. */ + struct casewriter *writer; /* Writer. */ + struct case_map *map; /* Map from input data to data for writer. */ /* Common options. */ bool print_map; /* Print map? TODO. */ @@ -313,213 +232,198 @@ parse_write_command (enum writer_type writer_type, *retain_unselected = true; handle = NULL; - dict = dict_clone (dataset_dict (current_dataset)); - aw = xmalloc (sizeof *aw); - aw->writer = NULL; - aw->map = NULL; - case_nullify (&aw->bounce); + dict = dict_clone (dataset_dict (ds)); + writer = NULL; + map = NULL; print_map = false; print_short_names = false; sysfile_opts = sfm_writer_default_options (); porfile_opts = pfm_writer_default_options (); - start_case_map (dict); + case_map_prepare_dict (dict); dict_delete_scratch_vars (dict); - lex_match ('/'); + lex_match (lexer, '/'); for (;;) { - if (lex_match_id ("OUTFILE")) + if (lex_match_id (lexer, "OUTFILE")) { - if (handle != NULL) + if (handle != NULL) { lex_sbc_only_once ("OUTFILE"); - goto error; + goto error; } - - lex_match ('='); - - handle = fh_parse (FH_REF_FILE | FH_REF_SCRATCH); + + lex_match (lexer, '='); + + handle = fh_parse (lexer, FH_REF_FILE | FH_REF_SCRATCH); if (handle == NULL) goto error; } - else if (lex_match_id ("NAMES")) + else if (lex_match_id (lexer, "NAMES")) print_short_names = true; - else if (lex_match_id ("PERMISSIONS")) + else if (lex_match_id (lexer, "PERMISSIONS")) { bool cw; - - lex_match ('='); - if (lex_match_id ("READONLY")) + + lex_match (lexer, '='); + if (lex_match_id (lexer, "READONLY")) cw = false; - else if (lex_match_id ("WRITEABLE")) + else if (lex_match_id (lexer, "WRITEABLE")) cw = true; else { - lex_error (_("expecting %s or %s"), "READONLY", "WRITEABLE"); + lex_error (lexer, _("expecting %s or %s"), "READONLY", "WRITEABLE"); goto error; } sysfile_opts.create_writeable = porfile_opts.create_writeable = cw; } - else if (command_type == PROC_CMD && lex_match_id ("UNSELECTED")) + else if (command_type == PROC_CMD && lex_match_id (lexer, "UNSELECTED")) { - lex_match ('='); - if (lex_match_id ("RETAIN")) + lex_match (lexer, '='); + if (lex_match_id (lexer, "RETAIN")) *retain_unselected = true; - else if (lex_match_id ("DELETE")) + else if (lex_match_id (lexer, "DELETE")) *retain_unselected = false; else { - lex_error (_("expecting %s or %s"), "RETAIN", "DELETE"); + lex_error (lexer, _("expecting %s or %s"), "RETAIN", "DELETE"); goto error; } } - else if (writer_type == SYSFILE_WRITER && lex_match_id ("COMPRESSED")) + else if (writer_type == SYSFILE_WRITER && lex_match_id (lexer, "COMPRESSED")) sysfile_opts.compress = true; - else if (writer_type == SYSFILE_WRITER && lex_match_id ("UNCOMPRESSED")) + else if (writer_type == SYSFILE_WRITER && lex_match_id (lexer, "UNCOMPRESSED")) sysfile_opts.compress = false; - else if (writer_type == SYSFILE_WRITER && lex_match_id ("VERSION")) + else if (writer_type == SYSFILE_WRITER && lex_match_id (lexer, "VERSION")) { - lex_match ('='); - if (!lex_force_int ()) + lex_match (lexer, '='); + if (!lex_force_int (lexer)) goto error; - sysfile_opts.version = lex_integer (); - lex_get (); + sysfile_opts.version = lex_integer (lexer); + lex_get (lexer); } - else if (writer_type == PORFILE_WRITER && lex_match_id ("TYPE")) + else if (writer_type == PORFILE_WRITER && lex_match_id (lexer, "TYPE")) { - lex_match ('='); - if (lex_match_id ("COMMUNICATIONS")) + lex_match (lexer, '='); + if (lex_match_id (lexer, "COMMUNICATIONS")) porfile_opts.type = PFM_COMM; - else if (lex_match_id ("TAPE")) + else if (lex_match_id (lexer, "TAPE")) porfile_opts.type = PFM_TAPE; else { - lex_error (_("expecting %s or %s"), "COMM", "TAPE"); + lex_error (lexer, _("expecting %s or %s"), "COMM", "TAPE"); goto error; } } - else if (writer_type == PORFILE_WRITER && lex_match_id ("DIGITS")) + else if (writer_type == PORFILE_WRITER && lex_match_id (lexer, "DIGITS")) { - lex_match ('='); - if (!lex_force_int ()) + lex_match (lexer, '='); + if (!lex_force_int (lexer)) goto error; - porfile_opts.digits = lex_integer (); - lex_get (); + porfile_opts.digits = lex_integer (lexer); + lex_get (lexer); } - else if (!parse_dict_trim (dict)) + else if (!parse_dict_trim (lexer, dict)) goto error; - - if (!lex_match ('/')) + + if (!lex_match (lexer, '/')) break; } - if (lex_end_of_command () != CMD_SUCCESS) + if (lex_end_of_command (lexer) != CMD_SUCCESS) goto error; - if (handle == NULL) + if (handle == NULL) { - lex_sbc_missing ("OUTFILE"); + lex_sbc_missing (lexer, "OUTFILE"); goto error; } + dict_delete_scratch_vars (dict); dict_compact_values (dict); - aw->map = finish_case_map (dict); - if (aw->map != NULL) - case_create (&aw->bounce, dict_get_next_value_idx (dict)); - if (fh_get_referent (handle) == FH_REF_FILE) + if (fh_get_referent (handle) == FH_REF_FILE) { - switch (writer_type) + switch (writer_type) { case SYSFILE_WRITER: - aw->writer = any_writer_from_sfm_writer ( - sfm_open_writer (handle, dict, sysfile_opts)); + writer = sfm_open_writer (handle, dict, sysfile_opts); break; case PORFILE_WRITER: - aw->writer = any_writer_from_pfm_writer ( - pfm_open_writer (handle, dict, porfile_opts)); + writer = pfm_open_writer (handle, dict, porfile_opts); break; } } else - aw->writer = any_writer_open (handle, dict); - if (aw->writer == NULL) + writer = any_writer_open (handle, dict); + if (writer == NULL) goto error; + + map = case_map_from_dict (dict); + if (map != NULL) + writer = casewriter_create_translator (writer, + case_map_get_value_cnt (map), + get_translate_case, + get_destroy_case_map, + map); dict_destroy (dict); - - return aw; + + return writer; error: - case_writer_destroy (aw); + casewriter_destroy (writer); dict_destroy (dict); + case_map_destroy (map); return NULL; } - -/* Writes case C to writer AW. */ -static bool -case_writer_write_case (struct case_writer *aw, const struct ccase *c) -{ - if (aw->map != NULL) - { - map_case (aw->map, c, &aw->bounce); - c = &aw->bounce; - } - return any_writer_write (aw->writer, c); -} /* SAVE and EXPORT. */ -static bool output_proc (const struct ccase *, void *); - /* Parses and performs the SAVE or EXPORT procedure. */ static int -parse_output_proc (enum writer_type writer_type) +parse_output_proc (struct lexer *lexer, struct dataset *ds, enum writer_type writer_type) { bool retain_unselected; struct variable *saved_filter_variable; - struct case_writer *aw; + struct casewriter *output; bool ok; - aw = parse_write_command (writer_type, PROC_CMD, &retain_unselected); - if (aw == NULL) + output = parse_write_command (lexer, ds, writer_type, PROC_CMD, + &retain_unselected); + if (output == NULL) return CMD_CASCADING_FAILURE; - saved_filter_variable = dict_get_filter (dataset_dict (current_dataset)); - if (retain_unselected) - dict_set_filter (dataset_dict (current_dataset), NULL); - ok = procedure (current_dataset,output_proc, aw); - dict_set_filter (dataset_dict (current_dataset), saved_filter_variable); + saved_filter_variable = dict_get_filter (dataset_dict (ds)); + if (retain_unselected) + dict_set_filter (dataset_dict (ds), NULL); - case_writer_destroy (aw); - return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE; -} + casereader_transfer (proc_open (ds), output); + ok = casewriter_destroy (output); + ok = proc_commit (ds) && ok; -/* Writes case C to file. */ -static bool -output_proc (const struct ccase *c, void *aw_) -{ - struct case_writer *aw = aw_; - return case_writer_write_case (aw, c); + dict_set_filter (dataset_dict (ds), saved_filter_variable); + + return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE; } int -cmd_save (void) +cmd_save (struct lexer *lexer, struct dataset *ds) { - return parse_output_proc (SYSFILE_WRITER); + return parse_output_proc (lexer, ds, SYSFILE_WRITER); } int -cmd_export (void) +cmd_export (struct lexer *lexer, struct dataset *ds) { - return parse_output_proc (PORFILE_WRITER); + return parse_output_proc (lexer, ds, PORFILE_WRITER); } /* XSAVE and XEXPORT. */ /* Transformation. */ -struct output_trns +struct output_trns { - struct case_writer *aw; /* Writer. */ + struct casewriter *writer; /* Writer. */ }; static trns_proc_func output_trns_proc; @@ -527,26 +431,28 @@ static trns_free_func output_trns_free; /* Parses the XSAVE or XEXPORT transformation command. */ static int -parse_output_trns (enum writer_type writer_type) +parse_output_trns (struct lexer *lexer, struct dataset *ds, enum writer_type writer_type) { struct output_trns *t = xmalloc (sizeof *t); - t->aw = parse_write_command (writer_type, XFORM_CMD, NULL); - if (t->aw == NULL) + t->writer = parse_write_command (lexer, ds, writer_type, XFORM_CMD, NULL); + if (t->writer == NULL) { free (t); return CMD_CASCADING_FAILURE; } - add_transformation (current_dataset, output_trns_proc, output_trns_free, t); + add_transformation (ds, output_trns_proc, output_trns_free, t); return CMD_SUCCESS; } /* Writes case C to the system file specified on XSAVE or XEXPORT. */ static int -output_trns_proc (void *trns_, struct ccase *c, casenum_t case_num UNUSED) +output_trns_proc (void *trns_, struct ccase *c, casenumber case_num UNUSED) { struct output_trns *t = trns_; - case_writer_write_case (t->aw, c); + struct ccase tmp; + case_clone (&tmp, c); + casewriter_write (t->writer, &tmp); return TRNS_CONTINUE; } @@ -556,62 +462,57 @@ static bool output_trns_free (void *trns_) { struct output_trns *t = trns_; - bool ok = true; - - if (t != NULL) - { - ok = case_writer_destroy (t->aw); - free (t); - } + bool ok = casewriter_destroy (t->writer); + free (t); return ok; } /* XSAVE command. */ int -cmd_xsave (void) +cmd_xsave (struct lexer *lexer, struct dataset *ds) { - return parse_output_trns (SYSFILE_WRITER); + return parse_output_trns (lexer, ds, SYSFILE_WRITER); } /* XEXPORT command. */ int -cmd_xexport (void) +cmd_xexport (struct lexer *lexer, struct dataset *ds) { - return parse_output_trns (PORFILE_WRITER); + return parse_output_trns (lexer, ds, PORFILE_WRITER); } -static bool rename_variables (struct dictionary *dict); -static bool drop_variables (struct dictionary *dict); -static bool keep_variables (struct dictionary *dict); +static bool rename_variables (struct lexer *lexer, struct dictionary *dict); +static bool drop_variables (struct lexer *, struct dictionary *dict); +static bool keep_variables (struct lexer *, struct dictionary *dict); /* Commands that read and write system files share a great deal of common syntactic structure for rearranging and dropping variables. This function parses this syntax and modifies DICT appropriately. Returns true on success, false on failure. */ static bool -parse_dict_trim (struct dictionary *dict) +parse_dict_trim (struct lexer *lexer, struct dictionary *dict) { - if (lex_match_id ("MAP")) + if (lex_match_id (lexer, "MAP")) { /* FIXME. */ return true; } - else if (lex_match_id ("DROP")) - return drop_variables (dict); - else if (lex_match_id ("KEEP")) - return keep_variables (dict); - else if (lex_match_id ("RENAME")) - return rename_variables (dict); + else if (lex_match_id (lexer, "DROP")) + return drop_variables (lexer, dict); + else if (lex_match_id (lexer, "KEEP")) + return keep_variables (lexer, dict); + else if (lex_match_id (lexer, "RENAME")) + return rename_variables (lexer, dict); else { - lex_error (_("expecting a valid subcommand")); + lex_error (lexer, _("expecting a valid subcommand")); return false; } } /* Parses and performs the RENAME subcommand of GET and SAVE. */ static bool -rename_variables (struct dictionary *dict) +rename_variables (struct lexer *lexer, struct dictionary *dict) { size_t i; @@ -624,29 +525,30 @@ rename_variables (struct dictionary *dict) int group; - lex_match ('='); - if (token != '(') + lex_match (lexer, '='); + if (lex_token (lexer) != '(') { struct variable *v; - v = parse_dict_variable (dict); + v = parse_variable (lexer, dict); if (v == NULL) return 0; - if (!lex_force_match ('=') - || !lex_force_id ()) + if (!lex_force_match (lexer, '=') + || !lex_force_id (lexer)) return 0; - if (dict_lookup_var (dict, tokid) != NULL) + if (dict_lookup_var (dict, lex_tokid (lexer)) != NULL) { msg (SE, _("Cannot rename %s as %s because there already exists " "a variable named %s. To rename variables with " "overlapping names, use a single RENAME subcommand " "such as \"/RENAME (A=B)(B=C)(C=A)\", or equivalently, " - "\"/RENAME (A B C=B C A)\"."), v->name, tokid, tokid); + "\"/RENAME (A B C=B C A)\"."), + var_get_name (v), lex_tokid (lexer), lex_tokid (lexer)); return 0; } - - dict_rename_var (dict, v, tokid); - lex_get (); + + dict_rename_var (dict, v, lex_tokid (lexer)); + lex_get (lexer); return 1; } @@ -654,33 +556,33 @@ rename_variables (struct dictionary *dict) v = NULL; new_names = 0; group = 1; - while (lex_match ('(')) + while (lex_match (lexer, '(')) { size_t old_nv = nv; - if (!parse_variables (dict, &v, &nv, PV_NO_DUPLICATE | PV_APPEND)) + if (!parse_variables (lexer, dict, &v, &nv, PV_NO_DUPLICATE | PV_APPEND)) goto done; - if (!lex_match ('=')) + if (!lex_match (lexer, '=')) { msg (SE, _("`=' expected after variable list.")); goto done; } - if (!parse_DATA_LIST_vars (&new_names, &nn, PV_APPEND | PV_NO_SCRATCH)) + if (!parse_DATA_LIST_vars (lexer, &new_names, &nn, PV_APPEND | PV_NO_SCRATCH)) goto done; if (nn != nv) { - msg (SE, _("Number of variables on left side of `=' (%d) does not " - "match number of variables on right side (%d), in " + msg (SE, _("Number of variables on left side of `=' (%zu) does not " + "match number of variables on right side (%zu), in " "parenthesized group %d of RENAME subcommand."), - (unsigned) (nv - old_nv), (unsigned) (nn - old_nv), group); + nv - old_nv, nn - old_nv, group); goto done; } - if (!lex_force_match (')')) + if (!lex_force_match (lexer, ')')) goto done; group++; } - if (!dict_rename_vars (dict, v, new_names, nv, &err_name)) + if (!dict_rename_vars (dict, v, new_names, nv, &err_name)) { msg (SE, _("Requested renaming duplicates variable name %s."), err_name); goto done; @@ -699,13 +601,13 @@ rename_variables (struct dictionary *dict) /* Parses and performs the DROP subcommand of GET and SAVE. Returns true if successful, false on failure.*/ static bool -drop_variables (struct dictionary *dict) +drop_variables (struct lexer *lexer, struct dictionary *dict) { struct variable **v; size_t nv; - lex_match ('='); - if (!parse_variables (dict, &v, &nv, PV_NONE)) + lex_match (lexer, '='); + if (!parse_variables (lexer, dict, &v, &nv, PV_NONE)) return false; dict_delete_vars (dict, v, nv); free (v); @@ -721,19 +623,19 @@ drop_variables (struct dictionary *dict) /* Parses and performs the KEEP subcommand of GET and SAVE. Returns true if successful, false on failure.*/ static bool -keep_variables (struct dictionary *dict) +keep_variables (struct lexer *lexer, struct dictionary *dict) { struct variable **v; size_t nv; size_t i; - lex_match ('='); - if (!parse_variables (dict, &v, &nv, PV_NONE)) + lex_match (lexer, '='); + if (!parse_variables (lexer, dict, &v, &nv, PV_NONE)) return false; /* Move the specified variables to the beginning. */ dict_reorder_vars (dict, v, nv); - + /* Delete the remaining variables. */ v = xnrealloc (v, dict_get_var_cnt (dict) - nv, sizeof *v); for (i = nv; i < dict_get_var_cnt (dict); i++) @@ -747,194 +649,180 @@ keep_variables (struct dictionary *dict) /* MATCH FILES. */ /* File types. */ -enum +enum mtf_type { MTF_FILE, /* Specified on FILE= subcommand. */ MTF_TABLE /* Specified on TABLE= subcommand. */ }; -/* One of the files on MATCH FILES. */ +/* One of the FILEs or TABLEs on MATCH FILES. */ struct mtf_file { - struct mtf_file *next, *prev; /* Next, previous in the list of files. */ - struct mtf_file *next_min; /* Next in the chain of minimums. */ - - int type; /* One of MTF_*. */ - struct variable **by; /* List of BY variables for this file. */ - struct file_handle *handle; /* File handle. */ - struct any_reader *reader; /* File reader. */ - struct dictionary *dict; /* Dictionary from system file. */ + struct ll ll; /* In list of all files and tables. */ + + enum mtf_type type; + int sequence; + + const struct variable **by; /* List of BY variables for this file. */ + struct mtf_variable *vars; /* Variables to copy to output. */ + size_t var_cnt; /* Number of other variables. */ + + struct file_handle *handle; /* Input file handle. */ + struct dictionary *dict; /* Input file dictionary. */ + struct casereader *reader; /* Input reader. */ + struct ccase input; /* Input record (null at end of file). */ /* IN subcommand. */ char *in_name; /* Variable name. */ struct variable *in_var; /* Variable (in master dictionary). */ + }; - struct ccase input; /* Input record. */ +struct mtf_variable + { + struct variable *in_var; + struct variable *out_var; }; /* MATCH FILES procedure. */ -struct mtf_proc +struct mtf_proc { - struct mtf_file *head; /* First file mentioned on FILE or TABLE. */ - struct mtf_file *tail; /* Last file mentioned on FILE or TABLE. */ + struct ll_list files; /* List of "struct mtf_file"s. */ + int nonempty_files; /* FILEs that are not at end-of-file. */ bool ok; /* False if I/O error occurs. */ - size_t by_cnt; /* Number of variables on BY subcommand. */ - - /* Names of FIRST, LAST variables. */ - char first[LONG_NAME_LEN + 1], last[LONG_NAME_LEN + 1]; - struct dictionary *dict; /* Dictionary of output file. */ - struct casefile *output; /* MATCH FILES output. */ - struct ccase mtf_case; /* Case used for output. */ + struct casewriter *output; /* MATCH FILES output. */ - unsigned seq_num; /* Have we initialized this variable? */ - unsigned *seq_nums; /* Sequence numbers for each var in dict. */ + size_t by_cnt; /* Number of variables on BY subcommand. */ + + /* FIRST, LAST. + Only if "first" or "last" is nonnull are the remaining + members used. */ + struct variable *first; /* Variable specified on FIRST (if any). */ + struct variable *last; /* Variable specified on LAST (if any). */ + struct ccase buffered_case; /* Case ready for output except that we don't + know the value for the LAST variable yet. */ + struct ccase prev_BY_case; /* Case with values of last set of BY vars. */ + const struct variable **prev_BY; /* Last set of BY variables. */ }; -static bool mtf_free (struct mtf_proc *); -static bool mtf_close_file (struct mtf_file *); -static int mtf_merge_dictionary (struct dictionary *const, struct mtf_file *); -static bool mtf_delete_file_in_place (struct mtf_proc *, struct mtf_file **); +static void mtf_free (struct mtf_proc *); -static bool mtf_read_nonactive_records (void *); -static bool mtf_processing_finish (void *); -static bool mtf_processing (const struct ccase *, void *); +static bool mtf_close_all_files (struct mtf_proc *); +static bool mtf_merge_dictionary (struct dictionary *const, struct mtf_file *); +static bool mtf_read_record (struct mtf_proc *mtf, struct mtf_file *); -static char *var_type_description (struct variable *); +static void mtf_process_case (struct mtf_proc *); -static void set_master (struct variable *, struct variable *master); -static struct variable *get_master (struct variable *); +static bool create_flag_var (const char *subcommand_name, const char *var_name, + struct dictionary *, struct variable **); +static char *var_type_description (struct variable *); /* Parse and execute the MATCH FILES command. */ int -cmd_match_files (void) +cmd_match_files (struct lexer *lexer, struct dataset *ds) { struct mtf_proc mtf; - struct mtf_file *first_table = NULL; - struct mtf_file *iter; - - bool used_active_file = false; - bool saw_table = false; + struct ll *first_table; + struct mtf_file *file, *next; + bool saw_in = false; + struct casereader *active_file = NULL; - bool ok; - - mtf.head = mtf.tail = NULL; - mtf.by_cnt = 0; - mtf.first[0] = '\0'; - mtf.last[0] = '\0'; + char first_name[LONG_NAME_LEN + 1] = ""; + char last_name[LONG_NAME_LEN + 1] = ""; + + struct taint *taint = NULL; + + size_t i; + + ll_init (&mtf.files); + mtf.nonempty_files = 0; + first_table = ll_null (&mtf.files); mtf.dict = dict_create (); mtf.output = NULL; - case_nullify (&mtf.mtf_case); - mtf.seq_num = 0; - mtf.seq_nums = NULL; - dict_set_case_limit (mtf.dict, dict_get_case_limit (dataset_dict (current_dataset))); - - lex_match ('/'); - while (token == T_ID - && (lex_id_match ("FILE", tokid) || lex_id_match ("TABLE", tokid))) - { - struct mtf_file *file = xmalloc (sizeof *file); + mtf.by_cnt = 0; + mtf.first = mtf.last = NULL; + case_nullify (&mtf.buffered_case); + case_nullify (&mtf.prev_BY_case); + mtf.prev_BY = NULL; - if (lex_match_id ("FILE")) - file->type = MTF_FILE; - else if (lex_match_id ("TABLE")) - { - file->type = MTF_TABLE; - saw_table = true; - } - else - NOT_REACHED (); - lex_match ('='); + dict_set_case_limit (mtf.dict, dict_get_case_limit (dataset_dict (ds))); + lex_match (lexer, '/'); + while (lex_token (lexer) == T_ID + && (lex_id_match (ss_cstr ("FILE"), ss_cstr (lex_tokid (lexer))) + || lex_id_match (ss_cstr ("TABLE"), ss_cstr (lex_tokid (lexer))))) + { + struct mtf_file *file = xmalloc (sizeof *file); file->by = NULL; file->handle = NULL; file->reader = NULL; file->dict = NULL; file->in_name = NULL; file->in_var = NULL; + file->var_cnt = 0; + file->vars = NULL; case_nullify (&file->input); - /* FILEs go first, then TABLEs. */ - if (file->type == MTF_TABLE || first_table == NULL) + if (lex_match_id (lexer, "FILE")) { - file->next = NULL; - file->prev = mtf.tail; - if (mtf.tail) - mtf.tail->next = file; - mtf.tail = file; - if (mtf.head == NULL) - mtf.head = file; - if (file->type == MTF_TABLE && first_table == NULL) - first_table = file; + file->type = MTF_FILE; + ll_insert (first_table, &file->ll); + mtf.nonempty_files++; } - else + else if (lex_match_id (lexer, "TABLE")) { - assert (file->type == MTF_FILE); - file->next = first_table; - file->prev = first_table->prev; - if (first_table->prev) - first_table->prev->next = file; - else - mtf.head = file; - first_table->prev = file; + file->type = MTF_TABLE; + ll_push_tail (&mtf.files, &file->ll); + if (first_table == ll_null (&mtf.files)) + first_table = &file->ll; } + else + NOT_REACHED (); + lex_match (lexer, '='); - if (lex_match ('*')) + if (lex_match (lexer, '*')) { - file->handle = NULL; - file->reader = NULL; - - if (used_active_file) - { - msg (SE, _("The active file may not be specified more " - "than once.")); - goto error; - } - used_active_file = true; - - if (!proc_has_source (current_dataset)) + if (!proc_has_active_file (ds)) { msg (SE, _("Cannot specify the active file since no active " "file has been defined.")); goto error; } - if (proc_make_temporary_transformations_permanent (current_dataset)) + if (proc_make_temporary_transformations_permanent (ds)) msg (SE, _("MATCH FILES may not be used after TEMPORARY when " "the active file is an input source. " "Temporary transformations will be made permanent.")); - file->dict = dataset_dict (current_dataset); + file->dict = dict_clone (dataset_dict (ds)); } else { - file->handle = fh_parse (FH_REF_FILE | FH_REF_SCRATCH); + file->handle = fh_parse (lexer, FH_REF_FILE | FH_REF_SCRATCH); if (file->handle == NULL) goto error; file->reader = any_reader_open (file->handle, &file->dict); if (file->reader == NULL) goto error; - - case_create (&file->input, dict_get_next_value_idx (file->dict)); } - while (lex_match ('/')) - if (lex_match_id ("RENAME")) + while (lex_match (lexer, '/')) + if (lex_match_id (lexer, "RENAME")) { - if (!rename_variables (file->dict)) - goto error; + if (!rename_variables (lexer, file->dict)) + goto error; } - else if (lex_match_id ("IN")) + else if (lex_match_id (lexer, "IN")) { - lex_match ('='); - if (token != T_ID) + lex_match (lexer, '='); + if (lex_token (lexer) != T_ID) { - lex_error (NULL); + lex_error (lexer, NULL); goto error; } @@ -944,110 +832,118 @@ cmd_match_files (void) "TABLE.")); goto error; } - file->in_name = xstrdup (tokid); - lex_get (); + file->in_name = xstrdup (lex_tokid (lexer)); + lex_get (lexer); saw_in = true; } mtf_merge_dictionary (mtf.dict, file); } - - while (token != '.') + + while (lex_token (lexer) != '.') { - if (lex_match (T_BY)) + if (lex_match (lexer, T_BY)) { + struct mtf_file *file; struct variable **by; - + bool ok; + if (mtf.by_cnt) { - msg (SE, _("BY may appear at most once.")); + lex_sbc_only_once ("BY"); goto error; } - - lex_match ('='); - if (!parse_variables (mtf.dict, &by, &mtf.by_cnt, + + lex_match (lexer, '='); + if (!parse_variables (lexer, mtf.dict, &by, &mtf.by_cnt, PV_NO_DUPLICATE | PV_NO_SCRATCH)) goto error; - for (iter = mtf.head; iter != NULL; iter = iter->next) + ok = true; + ll_for_each (file, struct mtf_file, ll, &mtf.files) { size_t i; - - iter->by = xnmalloc (mtf.by_cnt, sizeof *iter->by); + file->by = xnmalloc (mtf.by_cnt, sizeof *file->by); for (i = 0; i < mtf.by_cnt; i++) { - iter->by[i] = dict_lookup_var (iter->dict, by[i]->name); - if (iter->by[i] == NULL) + const char *var_name = var_get_name (by[i]); + file->by[i] = dict_lookup_var (file->dict, var_name); + if (file->by[i] == NULL) { - msg (SE, _("File %s lacks BY variable %s."), - iter->handle ? fh_get_name (iter->handle) : "*", - by[i]->name); - free (by); - goto error; + if (file->handle != NULL) + msg (SE, _("File %s lacks BY variable %s."), + fh_get_name (file->handle), var_name); + else + msg (SE, _("Active file lacks BY variable %s."), + var_name); + ok = false; } } } free (by); + + if (!ok) + goto error; } - else if (lex_match_id ("FIRST")) + else if (lex_match_id (lexer, "FIRST")) { - if (mtf.first[0] != '\0') + if (first_name[0] != '\0') { - msg (SE, _("FIRST may appear at most once.")); + lex_sbc_only_once ("FIRST"); goto error; } - - lex_match ('='); - if (!lex_force_id ()) + + lex_match (lexer, '='); + if (!lex_force_id (lexer)) goto error; - strcpy (mtf.first, tokid); - lex_get (); + strcpy (first_name, lex_tokid (lexer)); + lex_get (lexer); } - else if (lex_match_id ("LAST")) + else if (lex_match_id (lexer, "LAST")) { - if (mtf.last[0] != '\0') + if (last_name[0] != '\0') { - msg (SE, _("LAST may appear at most once.")); + lex_sbc_only_once ("LAST"); goto error; } - - lex_match ('='); - if (!lex_force_id ()) + + lex_match (lexer, '='); + if (!lex_force_id (lexer)) goto error; - strcpy (mtf.last, tokid); - lex_get (); + strcpy (last_name, lex_tokid (lexer)); + lex_get (lexer); } - else if (lex_match_id ("MAP")) + else if (lex_match_id (lexer, "MAP")) { /* FIXME. */ } - else if (lex_match_id ("DROP")) + else if (lex_match_id (lexer, "DROP")) { - if (!drop_variables (mtf.dict)) + if (!drop_variables (lexer, mtf.dict)) goto error; } - else if (lex_match_id ("KEEP")) + else if (lex_match_id (lexer, "KEEP")) { - if (!keep_variables (mtf.dict)) + if (!keep_variables (lexer, mtf.dict)) goto error; } else { - lex_error (NULL); + lex_error (lexer, NULL); goto error; } - if (!lex_match ('/') && token != '.') + if (!lex_match (lexer, '/') && lex_token (lexer) != '.') { - lex_end_of_command (); + lex_end_of_command (lexer); goto error; } } if (mtf.by_cnt == 0) { - if (saw_table) + if (first_table != ll_null (&mtf.files)) { msg (SE, _("BY is required when TABLE is specified.")); goto error; @@ -1061,442 +957,313 @@ cmd_match_files (void) /* Set up mapping from each file's variables to master variables. */ - for (iter = mtf.head; iter != NULL; iter = iter->next) + ll_for_each (file, struct mtf_file, ll, &mtf.files) { - struct dictionary *d = iter->dict; - int i; + size_t in_var_cnt = dict_get_var_cnt (file->dict); - for (i = 0; i < dict_get_var_cnt (d); i++) + file->vars = xnmalloc (in_var_cnt, sizeof *file->vars); + file->var_cnt = 0; + for (i = 0; i < in_var_cnt; i++) { - struct variable *v = dict_get_var (d, i); - struct variable *mv = dict_lookup_var (mtf.dict, v->name); - if (mv != NULL) - set_master (v, mv); + struct variable *in_var = dict_get_var (file->dict, i); + struct variable *out_var = dict_lookup_var (mtf.dict, + var_get_name (in_var)); + + if (out_var != NULL) + { + struct mtf_variable *mv = &file->vars[file->var_cnt++]; + mv->in_var = in_var; + mv->out_var = out_var; + } } } - /* Add IN variables to master dictionary. */ - for (iter = mtf.head; iter != NULL; iter = iter->next) - if (iter->in_name != NULL) - { - iter->in_var = dict_create_var (mtf.dict, iter->in_name, 0); - if (iter->in_var == NULL) - { - msg (SE, _("IN variable name %s duplicates an " - "existing variable name."), - iter->in_var->name); - goto error; - } - iter->in_var->print = iter->in_var->write - = make_output_format (FMT_F, 1, 0); - } - - /* MATCH FILES performs an n-way merge on all its input files. - Abstract algorithm: - - 1. Read one input record from every input FILE. - - 2. If no FILEs are left, stop. Otherwise, proceed to step 3. - - 3. Find the FILE input record(s) that have minimum BY - values. Store all the values from these input records into - the output record. - - 4. For every TABLE, read another record as long as the BY values - on the TABLE's input record are less than the FILEs' BY values. - If an exact match is found, store all the values from the TABLE - input record into the output record. - - 5. Write the output record. - - 6. Read another record from each input file FILE and TABLE that - we stored values from above. If we come to the end of one of the - input files, remove it from the list of input files. - - 7. Repeat from step 2. - - Unfortunately, this algorithm can't be implemented in a - straightforward way because there's no function to read a - record from the active file. Instead, it has to be written - as a state machine. - - FIXME: For merging large numbers of files (more than 10?) a - better algorithm would use a heap for finding minimum - values. */ - - if (!used_active_file) - discard_variables (current_dataset); + /* Add IN, FIRST, and LAST variables to master dictionary. */ + ll_for_each (file, struct mtf_file, ll, &mtf.files) + if (!create_flag_var ("IN", file->in_name, mtf.dict, &file->in_var)) + goto error; + if (!create_flag_var ("FIRST", first_name, mtf.dict, &mtf.first) + || !create_flag_var ("LAST", last_name, mtf.dict, &mtf.last)) + goto error; + dict_delete_scratch_vars (mtf.dict); dict_compact_values (mtf.dict); - mtf.output = fastfile_create (dict_get_next_value_idx (mtf.dict)); - mtf.seq_nums = xcalloc (dict_get_var_cnt (mtf.dict), sizeof *mtf.seq_nums); - case_create (&mtf.mtf_case, dict_get_next_value_idx (mtf.dict)); + mtf.output = autopaging_writer_create (dict_get_next_value_idx (mtf.dict)); + taint = taint_clone (casewriter_get_taint (mtf.output)); - if (!mtf_read_nonactive_records (&mtf)) - goto error; - - if (used_active_file) + ll_for_each (file, struct mtf_file, ll, &mtf.files) { - proc_set_sink (current_dataset, - create_case_sink (&null_sink_class, - dataset_dict (current_dataset), NULL)); - ok = procedure (current_dataset,mtf_processing, &mtf) && mtf_processing_finish (&mtf); + if (file->reader == NULL) + { + if (active_file == NULL) + { + proc_discard_output (ds); + file->reader = active_file = proc_open (ds); + } + else + file->reader = casereader_clone (active_file); + } + taint_propagate (casereader_get_taint (file->reader), taint); } - else - ok = mtf_processing_finish (&mtf); - discard_variables (current_dataset); + ll_for_each_safe (file, next, struct mtf_file, ll, &mtf.files) + mtf_read_record (&mtf, file); + while (mtf.nonempty_files > 0) + mtf_process_case (&mtf); + if ((mtf.first != NULL || mtf.last != NULL) && mtf.prev_BY != NULL) + { + if (mtf.last != NULL) + case_data_rw (&mtf.buffered_case, mtf.last)->f = 1.0; + casewriter_write (mtf.output, &mtf.buffered_case); + case_nullify (&mtf.buffered_case); + } + mtf_close_all_files (&mtf); + if (active_file != NULL) + proc_commit (ds); - dict_destroy (dataset_dict (current_dataset)); - dataset_set_dict (current_dataset, mtf.dict); + proc_set_active_file (ds, casewriter_make_reader (mtf.output), mtf.dict); mtf.dict = NULL; - proc_set_source (current_dataset, storage_source_create (mtf.output)); mtf.output = NULL; - - if (!mtf_free (&mtf)) - ok = false; - return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE; - + + mtf_free (&mtf); + + return taint_destroy (taint) ? CMD_SUCCESS : CMD_CASCADING_FAILURE; + error: + if (active_file != NULL) + proc_commit (ds); mtf_free (&mtf); + taint_destroy (taint); return CMD_CASCADING_FAILURE; } -/* Repeats 2...7 an arbitrary number of times. */ +/* If VAR_NAME is a nonnull pointer to a non-empty string, + attempts to create a variable named VAR_NAME, with format + F1.0, in DICT, and stores a pointer to the variable in *VAR. + Returns true if successful, false if the variable name is a + duplicate (in which case a message saying that the variable + specified on the given SUBCOMMAND is a duplicate is emitted). + Also returns true, without doing anything, if VAR_NAME is null + or empty. */ static bool -mtf_processing_finish (void *mtf_) +create_flag_var (const char *subcommand, const char *var_name, + struct dictionary *dict, struct variable **var) { - struct mtf_proc *mtf = mtf_; - struct mtf_file *iter; - - /* Find the active file and delete it. */ - for (iter = mtf->head; iter; iter = iter->next) - if (iter->handle == NULL) - { - if (!mtf_delete_file_in_place (mtf, &iter)) - NOT_REACHED (); - break; - } - - while (mtf->head && mtf->head->type == MTF_FILE) - if (!mtf_processing (NULL, mtf)) - return false; - + if (var_name != NULL && var_name[0] != '\0') + { + struct fmt_spec format = fmt_for_output (FMT_F, 1, 0); + *var = dict_create_var (dict, var_name, 0); + if (*var == NULL) + { + msg (SE, _("Variable name %s specified on %s subcommand " + "duplicates an existing variable name."), + subcommand, var_name); + return false; + } + var_set_both_formats (*var, &format); + } + else + *var = NULL; return true; } -/* Return a string in a static buffer describing V's variable type and - width. */ +/* Return a string in an allocated buffer describing V's variable + type and width. */ static char * var_type_description (struct variable *v) { - static char buf[2][32]; - static int x = 0; - char *s; - - x ^= 1; - s = buf[x]; - - if (v->type == NUMERIC) - strcpy (s, "numeric"); + if (var_is_numeric (v)) + return xstrdup ("numeric"); else - { - assert (v->type == ALPHA); - sprintf (s, "string with width %d", v->width); - } - return s; + return xasprintf ("string with width %d", var_get_width (v)); } -/* Closes FILE and frees its associated data. - Returns true if successful, false if an I/O error - occurred on FILE. */ +/* Closes all the files in MTF and frees their associated data. + Returns true if successful, false if an I/O error occurred on + any of the files. */ static bool -mtf_close_file (struct mtf_file *file) +mtf_close_all_files (struct mtf_proc *mtf) { - bool ok = file->reader == NULL || !any_reader_error (file->reader); - free (file->by); - any_reader_close (file->reader); - if (file->handle != NULL) - dict_destroy (file->dict); - case_destroy (&file->input); - free (file->in_name); - free (file); - return ok; -} - -/* Free all the data for the MATCH FILES procedure. - Returns true if successful, false if an I/O error - occurred. */ -static bool -mtf_free (struct mtf_proc *mtf) -{ - struct mtf_file *iter, *next; + struct mtf_file *file; bool ok = true; - for (iter = mtf->head; iter; iter = next) + ll_for_each_preremove (file, struct mtf_file, ll, &mtf->files) { - next = iter->next; - assert (iter->dict != mtf->dict); - if (!mtf_close_file (iter)) - ok = false; + casereader_destroy (file->reader); + free (file->by); + dict_destroy (file->dict); + free (file->in_name); + case_destroy (&file->input); + free (file->vars); + free (file); } - - if (mtf->dict) - dict_destroy (mtf->dict); - case_destroy (&mtf->mtf_case); - free (mtf->seq_nums); return ok; } -/* Remove *FILE from the mtf_file chain. Make *FILE point to the next - file in the chain, or to NULL if was the last in the chain. - Returns true if successful, false if an I/O error occurred. */ -static bool -mtf_delete_file_in_place (struct mtf_proc *mtf, struct mtf_file **file) +/* Frees all the data for the MATCH FILES procedure. */ +static void +mtf_free (struct mtf_proc *mtf) { - struct mtf_file *f = *file; - int i; - - if (f->prev) - f->prev->next = f->next; - if (f->next) - f->next->prev = f->prev; - if (f == mtf->head) - mtf->head = f->next; - if (f == mtf->tail) - mtf->tail = f->prev; - *file = f->next; - - if (f->in_var != NULL) - case_data_rw (&mtf->mtf_case, f->in_var->fv)->f = 0.; - for (i = 0; i < dict_get_var_cnt (f->dict); i++) - { - struct variable *v = dict_get_var (f->dict, i); - struct variable *mv = get_master (v); - if (mv != NULL) - { - union value *out = case_data_rw (&mtf->mtf_case, mv->fv); - - if (v->type == NUMERIC) - out->f = SYSMIS; - else - memset (out->s, ' ', v->width); - } - } - - return mtf_close_file (f); + mtf_close_all_files (mtf); + dict_destroy (mtf->dict); + casewriter_destroy (mtf->output); + case_destroy (&mtf->buffered_case); + case_destroy (&mtf->prev_BY_case); } -/* Read a record from every input file except the active file. - Returns true if successful, false if an I/O error occurred. */ +/* Reads the next record into FILE, if possible, and update MTF's + nonempty_files count if not. */ static bool -mtf_read_nonactive_records (void *mtf_) +mtf_read_record (struct mtf_proc *mtf, struct mtf_file *file) { - struct mtf_proc *mtf = mtf_; - struct mtf_file *iter, *next; - bool ok = true; - - for (iter = mtf->head; ok && iter != NULL; iter = next) + case_destroy (&file->input); + if (!casereader_read (file->reader, &file->input)) { - next = iter->next; - if (iter->handle && !any_reader_read (iter->reader, &iter->input)) - if (!mtf_delete_file_in_place (mtf, &iter)) - ok = false; + mtf->nonempty_files--; + return false; } - return ok; + else + return true; } -/* Compare the BY variables for files A and B; return -1 if A < B, 0 - if A == B, 1 if A > B. */ +/* Compare the BY variables for files A and B; return -1 if A < + B, 0 if A == B, 1 if A > B. (If there are no BY variables, + then all records are equal.) */ static inline int mtf_compare_BY_values (struct mtf_proc *mtf, - struct mtf_file *a, struct mtf_file *b, - const struct ccase *c) + struct mtf_file *a, struct mtf_file *b) { - const struct ccase *ca = case_is_null (&a->input) ? c : &a->input; - const struct ccase *cb = case_is_null (&b->input) ? c : &b->input; - assert ((a == NULL) + (b == NULL) + (c == NULL) <= 1); - return case_compare_2dict (ca, cb, a->by, b->by, mtf->by_cnt); + return case_compare_2dict (&a->input, &b->input, a->by, b->by, mtf->by_cnt); } -/* Perform one iteration of steps 3...7 above. - Returns true if successful, false if an I/O error occurred. */ -static bool -mtf_processing (const struct ccase *c, void *mtf_) +/* Processes input files and write one case to the output file. */ +static void +mtf_process_case (struct mtf_proc *mtf) { - struct mtf_proc *mtf = mtf_; + struct ccase c; + struct mtf_file *min; + struct mtf_file *file; + int min_sequence; + size_t i; + + /* Find the set of one or more FILEs whose BY values are + minimal, as well as the set of zero or more TABLEs whose BY + values equal those of the minimum FILEs. + + After each iteration of the loop, this invariant holds: the + FILEs with minimum BY values thus far have "sequence" + members equal to min_sequence, and "min" points to one of + the mtf_files whose case has those minimum BY values, and + similarly for TABLEs. */ + min_sequence = 0; + min = NULL; + ll_for_each (file, struct mtf_file, ll, &mtf->files) + if (case_is_null (&file->input)) + file->sequence = -1; + else if (file->type == MTF_FILE) + { + int cmp = min != NULL ? mtf_compare_BY_values (mtf, min, file) : 1; + if (cmp <= 0) + file->sequence = cmp < 0 ? -1 : min_sequence; + else + { + file->sequence = ++min_sequence; + min = file; + } + } + else + { + int cmp; + assert (min != NULL); + do + { + cmp = mtf_compare_BY_values (mtf, min, file); + } + while (cmp > 0 && mtf_read_record (mtf, file)); + file->sequence = cmp == 0 ? min_sequence : -1; + } - /* Do we need another record from the active file? */ - bool read_active_file; + /* Form the output case from the input cases. */ + case_create (&c, dict_get_next_value_idx (mtf->dict)); + for (i = 0; i < dict_get_var_cnt (mtf->dict); i++) + { + struct variable *v = dict_get_var (mtf->dict, i); + value_set_missing (case_data_rw (&c, v), var_get_width (v)); + } + ll_for_each_reverse (file, struct mtf_file, ll, &mtf->files) + { + bool include_file = file->sequence == min_sequence; + if (include_file) + for (i = 0; i < file->var_cnt; i++) + { + const struct mtf_variable *mv = &file->vars[i]; + const union value *in = case_data (&file->input, mv->in_var); + union value *out = case_data_rw (&c, mv->out_var); + value_copy (out, in, var_get_width (mv->in_var)); + } + if (file->in_var != NULL) + case_data_rw (&c, file->in_var)->f = include_file; + } - assert (mtf->head != NULL); - if (mtf->head->type == MTF_TABLE) - return true; - - do + /* Write the output case. */ + if (mtf->first == NULL && mtf->last == NULL) { - struct mtf_file *min_head, *min_tail; /* Files with minimum BY values. */ - struct mtf_file *max_head, *max_tail; /* Files with non-minimum BYs. */ - struct mtf_file *iter, *next; - - read_active_file = false; - - /* 3. Find the FILE input record(s) that have minimum BY - values. Store all the values from these input records into - the output record. */ - min_head = min_tail = mtf->head; - max_head = max_tail = NULL; - for (iter = mtf->head->next; iter && iter->type == MTF_FILE; - iter = iter->next) + /* With no FIRST or LAST variables, it's trivial. */ + casewriter_write (mtf->output, &c); + } + else + { + /* It's harder with LAST, because we can't know whether + this case is the last in a group until we've prepared + the *next* case also. Thus, we buffer the previous + output case until the next one is ready. + + We also have to save a copy of one of the previous input + cases, so that we can compare the BY variables. We + can't compare the BY variables between the current + output case and the saved one because the BY variables + might not be in the output (the user is allowed to drop + them). */ + bool new_BY; + if (mtf->prev_BY != NULL) { - int cmp = mtf_compare_BY_values (mtf, min_head, iter, c); - if (cmp < 0) - { - if (max_head) - max_tail = max_tail->next_min = iter; - else - max_head = max_tail = iter; - } - else if (cmp == 0) - min_tail = min_tail->next_min = iter; - else /* cmp > 0 */ - { - if (max_head) - { - max_tail->next_min = min_head; - max_tail = min_tail; - } - else - { - max_head = min_head; - max_tail = min_tail; - } - min_head = min_tail = iter; - } + new_BY = case_compare_2dict (&min->input, &mtf->prev_BY_case, + min->by, mtf->prev_BY, + mtf->by_cnt); + if (mtf->last != NULL) + case_data_rw (&mtf->buffered_case, mtf->last)->f = new_BY; + casewriter_write (mtf->output, &mtf->buffered_case); } - - /* 4. For every TABLE, read another record as long as the BY - values on the TABLE's input record are less than the FILEs' - BY values. If an exact match is found, store all the values - from the TABLE input record into the output record. */ - for (; iter != NULL; iter = next) - { - assert (iter->type == MTF_TABLE); - - next = iter->next; - for (;;) - { - int cmp = mtf_compare_BY_values (mtf, min_head, iter, c); - if (cmp < 0) - { - if (max_head) - max_tail = max_tail->next_min = iter; - else - max_head = max_tail = iter; - } - else if (cmp == 0) - min_tail = min_tail->next_min = iter; - else /* cmp > 0 */ - { - if (iter->handle == NULL) - return true; - if (any_reader_read (iter->reader, &iter->input)) - continue; - if (!mtf_delete_file_in_place (mtf, &iter)) - return false; - } - break; - } - } - - /* Next sequence number. */ - mtf->seq_num++; - - /* Store data to all the records we are using. */ - if (min_tail) - min_tail->next_min = NULL; - for (iter = min_head; iter; iter = iter->next_min) - { - int i; - - for (i = 0; i < dict_get_var_cnt (iter->dict); i++) - { - struct variable *v = dict_get_var (iter->dict, i); - struct variable *mv = get_master (v); - - if (mv != NULL && mtf->seq_nums[mv->index] != mtf->seq_num) - { - const struct ccase *record - = case_is_null (&iter->input) ? c : &iter->input; - union value *out = case_data_rw (&mtf->mtf_case, mv->fv); - - mtf->seq_nums[mv->index] = mtf->seq_num; - if (v->type == NUMERIC) - out->f = case_num (record, v->fv); - else - memcpy (out->s, case_str (record, v->fv), v->width); - } - } - if (iter->in_var != NULL) - case_data_rw (&mtf->mtf_case, iter->in_var->fv)->f = 1.; - - if (iter->type == MTF_FILE && iter->handle == NULL) - read_active_file = true; - } - - /* Store missing values to all the records we're not - using. */ - if (max_tail) - max_tail->next_min = NULL; - for (iter = max_head; iter; iter = iter->next_min) - { - int i; - - for (i = 0; i < dict_get_var_cnt (iter->dict); i++) - { - struct variable *v = dict_get_var (iter->dict, i); - struct variable *mv = get_master (v); - - if (mv != NULL && mtf->seq_nums[mv->index] != mtf->seq_num) - { - union value *out = case_data_rw (&mtf->mtf_case, mv->fv); - mtf->seq_nums[mv->index] = mtf->seq_num; - - if (v->type == NUMERIC) - out->f = SYSMIS; - else - memset (out->s, ' ', v->width); - } - } - if (iter->in_var != NULL) - case_data_rw (&mtf->mtf_case, iter->in_var->fv)->f = 0.; - } + else + new_BY = true; - /* 5. Write the output record. */ - casefile_append (mtf->output, &mtf->mtf_case); + case_move (&mtf->buffered_case, &c); + if (mtf->first != NULL) + case_data_rw (&mtf->buffered_case, mtf->first)->f = new_BY; - /* 6. Read another record from each input file FILE and TABLE - that we stored values from above. If we come to the end of - one of the input files, remove it from the list of input - files. */ - for (iter = min_head; iter && iter->type == MTF_FILE; iter = next) - { - next = iter->next_min; - if (iter->reader != NULL - && !any_reader_read (iter->reader, &iter->input)) - if (!mtf_delete_file_in_place (mtf, &iter)) - return false; - } + if (new_BY) + { + mtf->prev_BY = min->by; + case_destroy (&mtf->prev_BY_case); + case_clone (&mtf->prev_BY_case, &min->input); + } } - while (!read_active_file - && mtf->head != NULL && mtf->head->type == MTF_FILE); - return true; + /* Read another record from each input file FILE with minimum + values. */ + ll_for_each (file, struct mtf_file, ll, &mtf->files) + if (file->type == MTF_FILE) + { + if (file->sequence == min_sequence) + mtf_read_record (mtf, file); + } + else + break; } /* Merge the dictionary for file F into master dictionary M. */ -static int +static bool mtf_merge_dictionary (struct dictionary *const m, struct mtf_file *f) { struct dictionary *d = f->dict; @@ -1508,202 +1275,56 @@ mtf_merge_dictionary (struct dictionary *const m, struct mtf_file *f) d_docs = dict_get_documents (d); m_docs = dict_get_documents (m); - if (d_docs != NULL) + if (d_docs != NULL) { if (m_docs == NULL) dict_set_documents (m, d_docs); else { - char *new_docs; - size_t new_len; - - new_len = strlen (m_docs) + strlen (d_docs); - new_docs = xmalloc (new_len + 1); - strcpy (new_docs, m_docs); - strcat (new_docs, d_docs); + char *new_docs = xasprintf ("%s%s", m_docs, d_docs); dict_set_documents (m, new_docs); free (new_docs); } } - + for (i = 0; i < dict_get_var_cnt (d); i++) { struct variable *dv = dict_get_var (d, i); - struct variable *mv = dict_lookup_var (m, dv->name); + struct variable *mv = dict_lookup_var (m, var_get_name (dv)); - if (dict_class_from_id (dv->name) == DC_SCRATCH) + if (dict_class_from_id (var_get_name (dv)) == DC_SCRATCH) continue; if (mv != NULL) { - if (mv->width != dv->width) + if (var_get_width (mv) != var_get_width (dv)) { + char *dv_description = var_type_description (dv); + char *mv_description = var_type_description (mv); msg (SE, _("Variable %s in file %s (%s) has different " "type or width from the same variable in " "earlier file (%s)."), - dv->name, fh_get_name (f->handle), - var_type_description (dv), var_type_description (mv)); - return 0; + var_get_name (dv), fh_get_name (f->handle), + dv_description, mv_description); + free (dv_description); + free (mv_description); + return false; } - - if (dv->width == mv->width) + + if (var_get_width (dv) == var_get_width (mv)) { - if (val_labs_count (dv->val_labs) - && !val_labs_count (mv->val_labs)) - { - val_labs_destroy (mv->val_labs); - mv->val_labs = val_labs_copy (dv->val_labs); - } - if (!mv_is_empty (&dv->miss) && mv_is_empty (&mv->miss)) - mv_copy (&mv->miss, &dv->miss); + if (var_has_value_labels (dv) && !var_has_value_labels (mv)) + var_set_value_labels (mv, var_get_value_labels (dv)); + if (var_has_missing_values (dv) && !var_has_missing_values (mv)) + var_set_missing_values (mv, var_get_missing_values (dv)); } - if (dv->label && !mv->label) - mv->label = xstrdup (dv->label); + if (var_get_label (dv) && !var_get_label (mv)) + var_set_label (mv, var_get_label (dv)); } else - mv = dict_clone_var_assert (m, dv, dv->name); + mv = dict_clone_var_assert (m, dv, var_get_name (dv)); } - return 1; -} - -/* Marks V's master variable as MASTER. */ -static void -set_master (struct variable *v, struct variable *master) -{ - var_attach_aux (v, master, NULL); -} - -/* Returns the master variable corresponding to V, - as set with set_master(). */ -static struct variable * -get_master (struct variable *v) -{ - return v->aux; -} - - - -/* Case map. - - A case map copies data from a case that corresponds for one - dictionary to a case that corresponds to a second dictionary - derived from the first by, optionally, deleting, reordering, - or renaming variables. (No new variables may be created.) - */ - -/* A case map. */ -struct case_map - { - size_t value_cnt; /* Number of values in map. */ - int *map; /* For each destination index, the - corresponding source index. */ - }; - -/* Prepares dictionary D for producing a case map. Afterward, - the caller may delete, reorder, or rename variables within D - at will before using finish_case_map() to produce the case - map. - - Uses D's aux members, which must otherwise not be in use. */ -static void -start_case_map (struct dictionary *d) -{ - size_t var_cnt = dict_get_var_cnt (d); - size_t i; - - for (i = 0; i < var_cnt; i++) - { - struct variable *v = dict_get_var (d, i); - int *src_fv = xmalloc (sizeof *src_fv); - *src_fv = v->fv; - var_attach_aux (v, src_fv, var_dtor_free); - } -} - -/* Produces a case map from dictionary D, which must have been - previously prepared with start_case_map(). - - Does not retain any reference to D, and clears the aux members - set up by start_case_map(). - - Returns the new case map, or a null pointer if no mapping is - required (that is, no data has changed position). */ -static struct case_map * -finish_case_map (struct dictionary *d) -{ - struct case_map *map; - size_t var_cnt = dict_get_var_cnt (d); - size_t i; - int identity_map; - - map = xmalloc (sizeof *map); - map->value_cnt = dict_get_next_value_idx (d); - map->map = xnmalloc (map->value_cnt, sizeof *map->map); - for (i = 0; i < map->value_cnt; i++) - map->map[i] = -1; - - identity_map = 1; - for (i = 0; i < var_cnt; i++) - { - struct variable *v = dict_get_var (d, i); - int *src_fv = (int *) var_detach_aux (v); - size_t idx; - - if (v->fv != *src_fv) - identity_map = 0; - - for (idx = 0; idx < v->nv; idx++) - { - int src_idx = *src_fv + idx; - int dst_idx = v->fv + idx; - - assert (map->map[dst_idx] == -1); - map->map[dst_idx] = src_idx; - } - free (src_fv); - } - - if (identity_map) - { - destroy_case_map (map); - return NULL; - } - - while (map->value_cnt > 0 && map->map[map->value_cnt - 1] == -1) - map->value_cnt--; - - return map; -} - -/* Maps from SRC to DST, applying case map MAP. */ -static void -map_case (const struct case_map *map, - const struct ccase *src, struct ccase *dst) -{ - size_t dst_idx; - - assert (map != NULL); - assert (src != NULL); - assert (dst != NULL); - assert (src != dst); - - for (dst_idx = 0; dst_idx < map->value_cnt; dst_idx++) - { - int src_idx = map->map[dst_idx]; - if (src_idx != -1) - *case_data_rw (dst, dst_idx) = *case_data (src, src_idx); - } -} - -/* Destroys case map MAP. */ -static void -destroy_case_map (struct case_map *map) -{ - if (map != NULL) - { - free (map->map); - free (map); - } + return true; }