We are using a single member in struct file_handle, the "name"

[pspp-builds.git] / src / data / ChangeLog

Thu Feb  1 16:53:37 2007  Ben Pfaff  <blp@gnu.org>

        We are using a single member in struct file_handle, the "name"
        field, for more than one purpose.  When it begins with '"', it's a
        file name; otherwise, it's a token that can be used to identify
        it.  When that assertion fires, it's because we searched for the
        name case-sensitively as a file name (so that there was no match),
        and then we try to insert it case-insensitively as a token, which
        fails because duplicates aren't allowed.

        Solution: break the two purposes into two separate fields.  This
        fixes the problem and likely makes the code easier to read too.

        Fixes bug #18922.  Thanks to John Darrington for bug report and
        review.

        * file-handle-def.c (struct file_handle): New `id' member.
        (fh_from_name): Rename fh_from_id.  Update all callers.
        (create_handle): New `id' parameter.  Update all callers.
        (fh_create_file): Ditto.
        (fh_get_id): New function.

Mon Jan 15 16:18:10 2007  Ben Pfaff  <blp@gnu.org>

        * case.c (case_is_null): Change return type to bool.

Mon Jan 15 10:57:28 2007  Ben Pfaff  <blp@gnu.org>

        Add debugging code.
        
        * case.c (case_clone) [DEBUGGING]: When debugging, don't use
        reference counting to share data.  This makes it easy for
        valgrind, etc. to find accesses to cases that have been destroyed
        but have been kept around by another user's ref-count.  This often
        happens when the data set is small enough to find in memory; if a
        bigger data set that would overflow to disk were used, then data
        corruption would occur.

Mon Jan 15 10:55:18 2007  Ben Pfaff  <blp@gnu.org>

        Simplify code.

        * case.c (case_unshare): Make it check internally whether the
        ref_cnt is greater than 1, so that the callers don't have to.
        Update callers not to check.

Mon Jan 15 10:53:01 2007  Ben Pfaff  <blp@gnu.org>

        Before, I was thinking that I might want to get rid of reference
        counting at some point.  Now, I'm pretty sure that it's here to
        stay.  Thus, because we have to store the value_cnt anyway for
        reference-counted cases, we might as well expose it to users.

        * case.c (case_get_value_cnt): New function.
        (case_resize): Drop OLD_CNT argument.  Update all callers.  Only
        resize case if its size actually changed.

        * casefile.c (casefile_append_xfer): Use case_get_value_cnt
        instead of peeking inside struct case directly.
        (casefile_append): Ditto.

Mon Jan 15 10:50:22 2007  Ben Pfaff  <blp@gnu.org>

        Get rid of the inlines for the case functions, which made the
        header file hard to read.  (Also, in testing with "-O2 -DNDEBUG",
        the inlines didn't speed up "make check" at all, which is not a
        perfect benchmark but seems indicative.)
        
        * case.c: Remove #ifdef DEBUGGING...#endif around many function
        definitions.  Remove some assertions on nonnull pointers that were
        redundant with a pointer dereference soon after in the function.
        Also:
        (struct case_data): Move definition here from case.h.
        (case_data): Ditto.
        (case_num): Ditto.
        (case_str): Ditto.
        (case_data_wr): Ditto.
        
Sun Jan 14 21:41:12 2007  Ben Pfaff  <blp@gnu.org>

        * automake.mk: Add casedeque.h to sources.
        
        * casedeque.h: New file.

        * procedure.c: (struct dataset) Change lag_count, lag_head,
        lag_queue member into single struct casedeque member.  Update all
        users to use the casedeque instead.
        (lag_case) Removed.

Sun Jan 14 21:43:12 2007  Ben Pfaff  <blp@gnu.org>

        * procedure.c: Simplify lagged cases interface.  Updated all
        clients--well, the only client--to use the simplified interface.
        (dataset_n_lag) Removed.
        (dataset_set_n_lag) Removed.
        (dataset_need_lag) New function.

Tue Jan  9 07:20:05 WST 2007 John Darrington <john@darrington.wattle.id.au>

        * dictionary.c procedure.c: More changes to ensure that callbacks occur
        whenever appropriate, but only when the dataset/dictionary is in a 
        consistent state.

Sun Jan  7 08:33:04 WST 2007 John Darrington <john@darrington.wattle.id.au>

        * dictionary.c dictionary.h : Added callbacks for change of filter and 
        split variables.  Refactored some code to ensure that callbacks get
        invoked when appropriate.

        * procedure.c (proc_cancel_temporary_transformations): Make sure that 
        replace_dict callback occurs when permanent_dict replaces the current
        dictionary.

Wed Jan  3 11:02:11 WST 2007 John Darrington <john@darrington.wattle.id.au>

        * dictionary.c dictionary.h : Added callback for when the weight 
        variable of a dictionary changes.

Mon Jan  1 10:36:26 WST 2007 John Darrington <john@darrington.wattle.id.au>

        * dictionary.c dictionary.h : Added replace_source and replace_dict
        callbacks, and functions to deal with them.

Fri Dec 22 13:56:08 2006  Ben Pfaff  <blp@gnu.org>

        Simplify missing value handling.

        * missing-values.h (enum mv_class): New type.
        (enum mv_type): Moved definition into missing-values.c and renamed
        each MV_* to MVT_*, to distinguish them from the exposed mv_class
        enums.  Updated all uses.
        (struct missing_values): Changed type of `type' from `enum
        mv_type' to `int' because the definition is no longer exposed.
        
        * missing-values.c (mv_is_value_missing): Add new enum mv_class
        parameter.  Update all callers.
        (mv_is_num_missing): Ditto.
        (mv_is_str_missing): Ditto.
        (mv_is_value_user_missing): Removed.  Changed callers to use
        mv_is_value_missing.
        (mv_is_num_user_missing): Removed.  Changed callers to use
        mv_is_num_missing.
        (mv_is_str_user_missing): Removed.  Changed callers to use
        mv_is_str_missing.
        (mv_is_value_system_missing): Removed.  Changed callers to use
        mv_is_value_missing.
        (mv_set_type): Removed.  Changed callers to use mv_clear.
        (mv_clear): New function.
        
        * variable.c (var_is_value_missing): Add new enum mv_class
        parameter.  Update all callers.
        (var_is_num_missing): Ditto.
        (var_is_str_missing): Ditto.
        (var_is_value_user_missing): Removed.  Changed callers to use
        var_is_value_missing.
        (var_is_num_user_missing): Removed.  Changed callers to use
        var_is_num_missing.
        (var_is_str_user_missing): Removed.  Changed callers to use
        var_is_str_missing.
        (var_is_value_system_missing): Removed.  Changed callers to use
        var_is_value_missing.
        
        * casefilter.c (struct casefilter): Use enum mv_class in place of
        bool.
        (casefilter_variable_missing): Adapt to new member.
        (casefilter_create): Change signature to take enum mv_class,
        update callers.

Fri Dec 22 20:08:38 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * casefile-factory.h fastfile-factory.c fastfile-factory.h: New files.

        * case-sink.c case-sink.h procedure.c procedure.h 
          storage-stream.c: Now uses the factory.
        
Sat Dec 16 22:05:18 2006  Ben Pfaff  <blp@gnu.org>

        Make it possible to pull cases from the active file with a
        function call, instead of requiring indirection through a callback
        function.

        * case-source.h (struct case_source_class): Change ->read function
        to return a single case, instead of calling a callback function
        for each case.  Change ->destroy function to return an error
        status.

        * case-source.c (free_case_source): Pass along the value returned
        by the case_source ->destroy function.

        * procedure.c (struct write_case_data): Removed.
        (struct dataset): Added some members to track procedure state.
        (procedure): Optimize the trivial case at this level.
        (internal_procedure): Re-implement in terms of proc_open,
        proc_read, proc_close.
        (proc_open) New function.
        (proc_read) New function.
        (proc_close) New function.
        (write_case) Moved into proc_read.
        (close_active_file) Moved closing of data source into proc_close.

        * storage-source.c: Rewrote to conform with modified
        case_source_class interface.

        * transformations.c (trns_chain_execute): Added argument to allow
        starting execution from an arbitrary transformation.  Updated
        callers.

        * transformations.h (enum TRNS_NEXT_CASE) Renamed TRNS_END_CASE.

Sat Dec 16 14:09:25 2006  Ben Pfaff  <blp@gnu.org>

        * sys-file-reader.c (read_display_parameters): Don't assume that
        MEASURE_* and ALIGN_* have the same values found in system files.

        * sys-file-writer.c (write_variable_display_parameters): Ditto.

        * variable.h: Change MEASURE_NOMINAL, MEASURE_ORDINAL,
        MEASURE_SCALE to be 0-based instead of 1-based.  This also fixes
        the value of n_MEASURES, which was off by 1 (at least from my
        point of view).

Sat Dec 16 12:17:34 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * dictionary.c dictionary.h vardict.h variable.c: Added optional
                callbacks which are invoked when the dictionary or its 
                variables are changed.  
        
        * missing-values.c missing-values.h value-labels.c: Tidied up
                consistency checks, and made some of them return false 
                instead of assert-failing. 

Wed Dec 13 19:30:11 2006  Ben Pfaff  <blp@gnu.org>

        * calendar.c (calendar_days_in_month): New function.

Mon Dec 11 07:53:39 2006  Ben Pfaff  <blp@gnu.org>

        * value-labels.c (hash_int_val_lab): Only hash as many bytes as
        the value label's width.

Sun Dec 10 14:21:29 2006  Ben Pfaff  <blp@gnu.org>

        * sfm-private.h: Move contents into sys-file-writer.c, which is
        the only remaining user.  Removed Borland C++-specific directives.
        
        * sys-file-reader.c: Clean up and rewrite entire file.  The
        rewritten version is simpler and better abstracted, and should be
        easier to maintain and extend.  It avoids using structures to read
        file data, which is prone to padding variations among compilers.
        It should also handle non-IEEE 754 system files, although I
        haven't been able to find any.  It has been tested against many
        .sav files obtained from the Web and found to produce the same
        results as the earlier version of the code, or in some cases
        improved results.  It is more tolerant of format variations found
        in the wild.

        * sys-file-reader.h (struct sfm_read_info): Removed `big_endian'
        member, putting an enum integer_format in its place.  New member
        `float_format'.  Changed `compressed' member to type bool.

Sun Dec 10 13:48:53 2006  Ben Pfaff  <blp@gnu.org>

        * dictionary.c (dict_delete_consecutive_vars): New function.

Sat Dec  9 20:08:25 2006  Ben Pfaff  <blp@gnu.org>

        * file-name.c (fn_search_path): Remove prefix arg that was unused
        by any caller.  Updated all callers.

Sat Dec  9 20:04:22 2006  Ben Pfaff  <blp@gnu.org>

        * format.c (fmt_dollar_template): Use user's decimal point
        character.  Add assertion.

Sat Dec  9 20:02:25 2006  Ben Pfaff  <blp@gnu.org>

        * format.c (fmt_dollar_template): New function, based on
        dollar_format_template from var-type-dialog.c.

Sat Dec  9 18:05:59 2006  Ben Pfaff  <blp@gnu.org>

        * data-out.c (output_scientific): Fix bad assumption that "buf" is
        null-terminated.
        
Sat Dec  9 17:23:23 2006  Ben Pfaff  <blp@gnu.org>

        Finish converting struct variable to an opaque type.  In this
        phase, we add remaining setter and getter functions, convert the
        remaining PSPP code to use them, and do a bunch of cleanup.  The
        resulting changes are pervasive but mostly trivial, and only the
        notable changes are logged.
        
        * automake.mk (src_data_libdata_a_SOURCES): Add the new source
        files.
        
        * case.c (case_data): Renamed case_data_idx.
        (case_num): Renamed case_num_idx.
        (case_str): Renamed case_str_idx.
        (case_data_rw): Renamed case_data_rw_idx.

        * case.h (case_data): New function with old name and an interface
        that takes a variable instead of an index, which is easier to
        use.  Updated all callers to use the new interface, or to use the
        new *_idx function (see above).
        (case_num): Ditto.
        (case_str): Ditto.
        (case_data_rw): Ditto.
        
        * category.c (cat_stored_values_destroy): Changed interface to
        take a struct cat_vals * instead of a struct variable *.

        * dictionary.c (dict_clone): Use new vector_clone function.     
        (dict_clear) Use new var_destroy function.
        (add_var) New function.
        (dict_create_var) Rewrite in terms of dict_create_var_assert.
        (dict_create_var_assert) Rewrite in terms of add_var.
        (dict_clone_var) Rewrite in terms of dict_clone_var_assert.
        (dict_clone_var_assert) Rewrite in terms of var_clone, add_var.
        (dict_lookup_var) Use new var_create, var_destroy functions.
        (dict_contains_var) Rewrite in terms of new vardict functionality.
        (set_var_dict_index) New function.
        (set_var_case_index) New function.
        (reindex_vars) New function.
        (dict_delete_var) Rewrite in terms of new vardict functionality.
        (dict_reorder_var) Ditto.
        (dict_reorder_vars) Ditto.
        (rename_var) New function.
        (dict_rename_var) Use rename_var.
        (dict_rename_vars) Use pool to simplify code.  Use rename_var.
        (dict_get_compacted_idx_to_fv) Rename
        dict_get_compacted_dict_index_to_case_index, update callers.
        (dict_create_vector) Use new vector_create function.
        (dict_clear_vectors) Use new vector_destroy function.
        (set_var_short_name_suffix) Move here from variable.c, renamed
        from var_set_short_name_suffix, make static, update caller.

        * sys-file-private.c: New file.  
        (sfm_width_to_bytes) Moved here from variable.c, renamed from
        width_to_bytes, update callers.

        * sys-file-private.h: New file.  Later it will supplant
        sfm-private.h; for now it supplements it.
        (macro MIN_VERY_LONG_STRING) New macro.
        (macro EFFECTIVE_LONG_STRING_LENGTH) New macro, from value.h.

        * sys-file-reader.c: Use MIN_VERY_LONG_STRING - 1 where
        MAX_LONG_STRING was used before.

        * sys-file-writer.c: Ditto.

        * value-labels.c: Change the paradigm here to be that a null
        pointer is OK for a struct val_labs * in most cases; it just
        represents an empty set of value labels.
        (val_labs_copy) A copy of a null set is a null set.
        (val_labs_count) A null set has 0 labels.
        (val_labs_replace) Change return type to void.  Rewrite for
        simplicity.
        (val_labs_find) A null set does not contain the value.
        (value_to_string) Moved to variable.c, renamed var_get_value_name,
        transposed argument order, updated all callers.

        * value.c: New file.
        (value_dup) Moved here from variable.c.
        (compare_values) Ditto.
        (hash_value) Ditto.

        * value.h: (macro MAX_SHORT_STRING) Rewrote for simplicity.
        (macro MAX_LONG_STRING) Removed, because it was only interesting
        for system files, not for general code.
        (macro MAX_VERY_LONG_STRING) Ditto.
        (macro EFFECTIVE_LONG_STRING_LENGTH) Moved to sys-file-private.h.
        (macro MAX_ELEMS_PER_VALUE) Removed, as it was unused.

        * vardict.h: New file, for an interface between variables and
        their dictionaries.

        * variable.c: A lot of functions were moved around, for better
        organization.
        (struct variable) Move definition here, from variable.h.
        (var_type_adj) Removed--makes i18n hard.
        (var_type_noun) Ditto.
        (var_create) New function.
        (var_clone) New function.
        (var_destroy) New function.
        (var_set_name) Assert that variable is not in a dictionary.
        (compare_var_names) Rename compare_vars_by_name and fix a couple
        of callers who thought the args were strings.
        (hash_var_name) Rename hash_var_by_name.
        (compare_var_ptr_names) Rename compare_var_ptrs_by_name.
        (hash_var_ptr_name) Rename hash_var_ptr_by_name.
        (var_is_very_long_string) Removed, because it was only interesting
        to system file code.
        (var_set_missing_values) Allow the argument to be the wrong width,
        as long as we can resize it.  Simplify callers who were doing the
        resizing themselves.
        (var_get_value_labels) New function.
        (var_has_value_labels) New function.
        (var_set_value_labels) New function.
        (alloc_value_labels) New function.
        (var_add_value_label) New function.
        (var_replace_value_label) New function.
        (var_clear_value_labels) New function.
        (var_lookup_value_label) New function.
        (var_get_value_name) Moved here from variable.c, renamed from
        var_get_value_name, transposed argument order, updated all
        callers.
        (var_to_string) Moved here, from variable-label.c.
        (var_set_leave) New function.
        (var_get_leave) New function.
        (var_must_leave) New function.
        (var_set_short_name_suffix) Moved to dictionary.c, renamed
        set_var_short_name_suffix.
        (var_get_dict_index) New function.
        (var_get_case_index) New function.
        (var_get_obs_vals) New function.
        (var_set_obs_vals) New function.
        (var_has_obs_vals) New function.
        (var_get_vardict) New function.
        (var_set_vardict) New function.
        (var_has_vardict) New function.
        (var_clear_vardict) New function.
        (value_dup) Moved to value.c.
        (compare_values) Ditto.
        (hash_value) Ditto.

        * variable.h: (enum NUMERIC) Rename VAR_NUMERIC, update all users.
        (enum ALPHA) Rename VAR_STRING, update all users.

        * vector.c: New file.
        (struct vector) Moved here, from variable.h.
        (check_widths) New function.
        (vector_create) New function.
        (vector_clone) New function.
        (vector_destroy) New function.
        (vector_get_name) New function.
        (vector_get_var) New function.
        (vector_get_var_cnt) New function.
        (compare_vector_ptrs_by_name) New function.

        * vector.h: New file.

Sun Dec 10 11:32:56 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * casefilter.c (casefilter_variable_missing): Avoided comparision of
        string variables to SYSMIS.  Thanks to Ben Pfaff for reporting this
        problem.

Sat Dec  9 07:18:03 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * value-labels.c (destroy_atoms): New function.
        * value-labels.c (atom_create): Call destroy_atoms in atexit handler.

Thu Dec  7 17:38:26 2006  Ben Pfaff  <blp@gnu.org>

        Thanks to Jason Stover for pointing out this problem.
        
        * data-out.c (output_number): Use gsl_finite from GSL, which is
        portable, instead of isfinite, which is not.
        (power256) Ditto.

Thu Dec  7 15:22:38 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * variable.c variable.h (value_dup): New function.

Mon Dec  4 22:20:17 2006  Ben Pfaff  <blp@gnu.org>

        Start converting struct variable to an opaque type.  In this
        phase, we add a bunch of setter and getter functions and convert
        most of the PSPP code to use them.  The resulting changes are
        pervasive but mostly trivial, and only the notable changes are
        logged.
        
        * format.c (fmt_equal): New function.
        
        * variable.c (var_type_is_valid): New function.
        (measure_is_valid) Moved here, from format.c.
        (alignment_is_valid) Moved here, from format.c.
        (var_get_name) New function.
        (var_set_name) New function.
        (width_to_type) New function.
        (var_get_type) New function.
        (var_get_width) New function.
        (var_set_width) New function.
        (var_is_numeric) New function.
        (var_is_alpha) New function.
        (var_is_short_string) New function.
        (var_is_long_string) New function.
        (var_is_very_long_string) New function.
        (var_get_missing_values) New function.
        (var_set_missing_values) New function.
        (var_clear_missing_values) New function.
        (var_has_missing_values) New function.
        (var_is_value_missing) New function.
        (var_is_num_missing) New function.
        (var_is_str_missing) New function.
        (var_is_value_user_missing) New function.
        (var_is_num_user_missing) New function.
        (var_is_str_user_missing) New function.
        (var_is_value_system_missing) New function.
        (var_get_print_format) New function.
        (var_set_print_format) New function.
        (var_get_write_format) New function.
        (var_set_write_format) New function.
        (var_set_both_formats) New function.
        (var_get_label) New function.
        (var_set_label) New function.
        (var_clear_label) New function.
        (var_has_label) New function.
        (var_get_measure) New function.
        (var_set_measure) New function.
        (var_get_display_width) New function.
        (var_set_display_width) New function.
        (var_get_alignment) New function.
        (var_set_alignment) New function.
        (var_get_value_cnt) New function.
        (var_get_leave) New function.
        (var_get_short_name) New function.

        * variable.h: (struct variable) Removed "type" and "nv" members;
        they are now computed from "width" where needed.

Mon Dec  4 21:38:40 2006  Ben Pfaff  <blp@gnu.org>

        * missing-values.c (mv_resize): Don't write beyond end of the
        allocated buffer when resizing a long string.

Sat Dec  2 16:28:32 2006  Ben Pfaff  <blp@gnu.org>

        Clean up identifier code: don't require identifier enumerations to
        be in a particular order; make better use of string library;
        expose less of the internals.
                
        * identifier.c: (lex_skip_identifier) Rename lex_id_get_length,
        change interface.  Updated all callers.
        (lex_id_match) Change interface to use struct substring, update
        all callers.
        (lex_id_match_len) Removed.  Update callers to use lex_id_match.
        (global array keywords[]) Make static, change form.  Update all
        users to use lex_id_name instead.
        (lex_is_keyword) New function.
        (lex_id_to_token) Change interface to use struct substring, update
        all callers.
        (lex_id_name) New function.

        * identifier.h: (T_FIRST_KEYWORD) Removed.  Changed users to call
        lex_is_keyword instead.
        (T_LAST_KEYWORD) Removed.
        (T_N_KEYWORDS) Removed.
        
Sat Nov 18 20:46:35 2006  Ben Pfaff  <blp@gnu.org>

        * format.c: (fmt_date_template) Distinguish characters for which a
        space is output and any date delimiter is allowed on input, from
        those for which a space is output and only a space is allowed on
        input.  The former is represented by X, the latter by a space.
        Also, drop distinction between h and H, changing the former to the
        latter.

        * data-in.c: Completely rewrite internals to conform to SPSS input
        formats as closely as possible.
        (data_in) Changed external interface by replacing the structure
        that was used as a single argument by a set of arguments.  Updated
        all callers.
        (data_in_finite_line) Removed.  Converted all callers to use plain
        data_in.
        (data_in_get_integer_format) New function.
        (data_in_set_integer_format) New function.
        (data_in_get_float_format) New function.
        (data_in_set_float_format) New function.

        * data-in.h: (enums DI_IGNORE_ERROR, DI_IMPLIED_DECIMALS) Removed.
        (struct data_in) Removed.

        * data-out.c: (output_date) Drop each component from the input as
        it is output, to allow us to drop the distinction between h (a
        count of hours) and H (the hour of day) template characters.
        Also, handle new X template character.
        (output_scientific) Follow more rational rule on when to drop
        fraction introduced between SPSS 13 and 15.  Updated test case to
        match new behavior.

Sat Nov 11 11:41:26 2006  Ben Pfaff  <blp@gnu.org>

        Fix buffer overflow reported by John Darrington.

        * data-out.c (output_bcd_integer): In case of SYSMIS, etc.,
        realize that DIGITS is a count of nibbles, not of bytes.

Sat Nov  4 15:59:56 2006  Ben Pfaff  <blp@gnu.org>

        * calendar.c (calendar_offset_to_gregorian) Also return the
        year-of-day.  Change callers to new interface.

        * data-out.c: Completely rewrite internals to conform to SPSS
        output formats as completely as possible.
        (data_out) Change interface to put input parameters before output
        parameters, for consistency with the style I now prefer.  Update
        all callers.
        (data_out_get_integer_format) New public function.
        (data_out_set_integer_format) New public function.
        (data_out_get_float_format) New public function.
        (data_out_set_float_format) New public function.

        * data-out.h: New file.  Move prototype for data_out here, from
        format.h.

        * format.c: (fmt_step_width) Use equality comparison instead of
        bitwise and, for clarity.
        (fmt_is_string) Ditto.
        (fmt_input_to_output) Fix categories that are translated to F
        format.

Sun Nov  5 08:29:34 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * casefilter.c casefilter.h (new files), casefile.c casefile.h 
        casefile-private.h: Added casefilter to assist commands with missing 
        values.

Sat Nov  4 11:47:09 2006  Ben Pfaff  <blp@gnu.org>

        Implement SET ERRORS, SHOW ERRORS.  Fixes bug #17609.
        
        * settings.c: (route_errors_to_terminal) New variable.
        (route_errors_to_listing) New variable. 
        (get_error_routing_to_terminal) New function.
        (set_error_routing_to_terminal) New function.
        (get_error_routing_to_listing) New function.
        (set_error_routing_to_listing) New function.

        * settings.h: (SET_ROUTE_* enums) Removed, because unused.

Tue Oct 31 19:58:27 2006  Ben Pfaff  <blp@gnu.org>

        * format.c: Completely rewrite, to achieve better abstraction.
        Rewrite all references to formats in other files.
        
        * format.def: Rewrite and reorganize.

        * settings.c: Move everything related to custom currency formats
        into format.[ch], changing them in form, so as to group related
        code and definitions better.  Changed all references to use the
        new functions.
        (static var decimal) Removed.
        (static var grouping) Removed.
        (static var cc) Removed.
        (get_decimal) Removed.
        (set_decimal) Removed.
        (get_grouping) Removed.
        (set_grouping) Removed.
        (get_cc) Removed.
        (set_cc) Removed.

        * settings.h: (macro CC_CNT) Removed.
        (macro CC_WIDTH) Removed.
        (struct custom_currency) Removed.

Tue Oct 31 19:56:19 2006  Ben Pfaff  <blp@gnu.org>

        * data-in.c (data_in): Use switch statement instead of table, to
        avoid dependence on the order of the FMT_* enums.

Tue Oct 31 19:35:36 2006  Ben Pfaff  <blp@gnu.org>

        * data-out.c: (num_to_string) Removed, because it was dead code.

Tue Oct 31 18:09:24 2006  Ben Pfaff  <blp@gnu.org>

        * data-in.c (parse_trailer): Fix error message.

Sat Oct 28 11:56:50 2006  Ben Pfaff  <blp@gnu.org>

        * format.c (fmt_is_binary): New function.

Thu Oct 19 22:59:56 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * procedure.c procedure.h: Encapsulated the static data into a single
        struct.  

Sat Oct 14 16:56:44 2006  Ben Pfaff  <blp@gnu.org>

        * casefile.c (casereader_read_xfer): Always initialize the case,
        even on an error condition.

Wed Sep 27 09:37:49 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * procedure.c (case_limit_trns_proc): Fixed buglet which rendered the 
        entire function useless.

Mon Sep 25 17:11:46 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * casefile-private.h casefile.c casefile.h fastfile.c: Created new
        casereader method casereader_clone.   
        
        * procedure.c pransformations.h: Introduced new type casenum_t

Thu Sep 21 07:00:30 2006  Ben Pfaff  <blp@gnu.org>

        * variable.c: (width_to_bytes) Rephrase code for clarify.

Sun Jul 16 19:52:03 2006  Ben Pfaff  <blp@gnu.org>

        * format.c: (fmt_type_from_string) New function.
        (fmt_to_string) Include decimals in output if the format has
        decimals, even if the format type does not.  This way, we can
        accurately reproduce incorrect formats in user output.
        (check_common_specifier) Make the check for a bad format type an
        assertion, so we get bug reports if they show up.  Fix message.
        Check for decimal places with a format type that doesn't allow
        them.
        (check_input_specifier) Remove check for FMT_X, which has been
        deleted.
        (check_output_specifier) Ditto. 

        * format.def: Remove FMT_T, FMT_X, FMT_DESCEND, FMT_NEWREC.

        * format.h: (macro FMT_TYPE_LEN_MAX) New macro.
        (struct fmt_desc) Use FMT_TYPE_LEN_MAX in definition.
        (enum fmt_parse_flags) Removed.

Mon Jul 17 18:26:21 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * casefile.c casefile.h: Converted to  an abstract base class.
        * casefile-private.h fastfile.c fastfile.h: New files.
        * automake.mk procedure.c scratch-writer.c storage-stream.c

Wed Jul 12 21:02:26 2006  Ben Pfaff  <blp@gnu.org>

        * procedure.c (internal_procedure): Create sink_case with only as
        many values as the compacted dictionary.

Wed Jul 12 21:01:00 2006  Ben Pfaff  <blp@gnu.org>

        Remove "debugging" code that caused plenty of false positives and
        no true positives.
        
        * case.h (struct ccase): [DEBUGGING] Remove `this' member.

        * case.c: Remove all references to `this' member.

Thu Jul  6 19:09:53 2006  Ben Pfaff  <blp@gnu.org>

        Fix link error noted by Jason Stover.
        
        * storage-stream.c: Include <assert.h>.

Tue Jul  4 08:47:35 2006  Ben Pfaff  <blp@gnu.org>

        Fix bug #15766 (/KEEP subcommand on SAVE doesn't fully support
        ALL) and additional underlying system file issues.

        Thanks to John Darrington for review.

        First problem: var_hash points to variables not owned by the
        sys-file-reader, which the caller may free or modify.  Use an
        array of sfm_vars instead, as done earlier (e.g. CVS version
        1.12).
        
        * sys-file-reader.c (struct sfm_reader): Remove var_hash, svars
        members and remove all code that references it.  Add vars, var_cnt
        members.  Remove fix_specials member, which was unused.
        (struct sfm_var) Remove name member, which was unused.
        (sfm_close_reader) Free vars member instead of var_hash.
        (compare_var_shortnames) Removed.
        (hash_var_shortname) Removed.
        (sfm_open_reader) Fill out vars array.
        (compare_var_index) Removed.
        (sfm_read_case) Use vars instead of var_hash.
        
        Second problem: we're confused about when we actually have very
        long strings, causing us to choose incorrectly between slow path
        and fast path in sfm_read_case.

        * sys-file-reader.c: (sfm_open_reader) Only mark has_vls if we
        have very long strings, not when we have long variable names,
        which is an unrelated feature.

Tue Jun 27 12:06:49 2006  Ben Pfaff  <blp@gnu.org>

        * variable.h: Move var_set and variable parsing declarations to
        new header, src/language/lexer/variable-parser.h.  Modified lots
        of files to include the new header.

Sun Jun 25 22:39:32 2006  Ben Pfaff  <blp@gnu.org>

        * value-labels.c (value_to_string): When there's no value label,
        format the variable according to its print format, instead of
        always effectively using A or F format.

Mon Jun 19 18:05:42 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * casefile.c (casefile_get_random_reader): Nasty hack to get around 
        the mode assertion.

        * format.c: Removed tortological assertion.

Fri Jun  9 12:20:09 2006  Ben Pfaff  <blp@gnu.org>

        Reform string library.
        
        * file-name.c (fn_interp_vars): Change interface to take a
        substring as input.  Updated all users.
        
Fri Jun  9 12:11:24 2006  Ben Pfaff  <blp@gnu.org>

        * format.c (measure_is_valid): Really return false when m >=
        n_MEASURES.

Tue Jun  6 18:46:26 2006  Ben Pfaff  <blp@gnu.org>

        Implement random access to casefiles, for use in GUI.
        
        * casefile.c: (struct casereader) Add `random', `file_ofs',
        `buffer_ofs' members.
        (casefile_get_random_reader) New function.
        (read_open_file) Break part into new function
        seek_and_fill_buffer().
        (fill_buffer) Update buffer_ofs, file_ofs.
        (casereader_seek) New function.

Tue May 30 19:52:33 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * settings.c: Added call to i18n{done, init}.

Tue May  9 21:09:17 2006  Ben Pfaff  <blp@gnu.org>

        * procedure.h: Add WARN_UNUSED_RESULT to procedure function
        prototypes.

Tue May  9 21:08:05 2006  Ben Pfaff  <blp@gnu.org>

        * casefile.c: Convert many uses of `int' to `bool'.

Sat May  6 22:49:43 2006  Ben Pfaff  <blp@gnu.org>

        * transformations.c (trns_chain_destroy): Destroy chain's trns
        member, to fix memory leak.

Sat May  6 22:48:30 2006  Ben Pfaff  <blp@gnu.org>

        * storage-stream.c (storage_source_decapsulate): Destroy case
        source to fix memory leak.

Sat May  6 22:46:47 2006  Ben Pfaff  <blp@gnu.org>

        * scratch-reader.c (scratch_reader_read_case): Copy into existing
        case passed as argument instead of initializing the argument as a
        case.  Fixes memory leak that showed up in
        tests/command/aggregate.sh with scratch files.

Sat May  6 22:45:55 2006  Ben Pfaff  <blp@gnu.org>

        * procedure.c (proc_done): Destroy default_dict, to fix memory
        leak.

Sat May  6 22:44:44 2006  Ben Pfaff  <blp@gnu.org>

        Simplify procedure_with_splits().
        
        * procedure.c (struct split_aux_data): Removed case_count member.
        (procedure_with_splits) Don't initialize case_count.
        (split_procedure_case_func) Check whether prev_case is null
        instead of case_count.
        (split_procedure_end_func) Ditto.

Sat May  6 22:42:23 2006  Ben Pfaff  <blp@gnu.org>

        * case.c (case_move): Do nothing if dst and src are the same
        object.
        (case_try_create) Merge two similar cases.
        (case_copy) Unshare only if data must be actually copied.

Sun May  7 10:04:06 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * data-in.c data-out.c dictionary.c sys-file-reader.c
        sys-file-writer.c variable.c variable.h:  Reworked very long string
        support for better encapsulation.

Sat May  6 19:02:00 2006  Ben Pfaff  <blp@gnu.org>

        * value-labels.c (val_labs_can_set_width): New function.
        (val_labs_set_width) Clear labels if increasing width to long
        string.
        (val_labs_destroy) Remove unneeded test for null.
        
Sat May  6 16:14:08 2006  Ben Pfaff  <blp@gnu.org>

        * value-labels.h: Remove unneeded dependency on variable.h.

Sat May  6 15:58:36 2006  Ben Pfaff  <blp@gnu.org>

        Get rid of `char *c' member in union value, for cleanliness.
        
        * value.h: (union value) Remove `c' member.

Sat May  6 15:36:59 2006  Ben Pfaff  <blp@gnu.org>

        Make dictionary compacting functions a little more general.
        
        * sys-file-writer.c (sfm_open_writer): Use
        dict_compacting_would_change().
        (does_dict_need_translation) Removed.

Sat May  6 15:35:42 2006  Ben Pfaff  <blp@gnu.org>

        Make dictionary compacting functions a little more general.
        
        * dictionary.c (dict_needs_compaction): Rename
        dict_compacting_would_shrink().  Update all callers.
        (dict_compacting_would_change) New function.
        
Sat May  6 14:25:49 2006  Ben Pfaff  <blp@gnu.org>

        * sys-file-writer.c: (does_dict_need_translation) Fix bug:
        inverted return value (!).

Sat May  6 13:37:52 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  

        * procedure.c: Search and replace "vfm" by "proc".  Notably:
        (static var vfm_source) Rename proc_source.  Update all
        references.
        (static var vfm_sink) Rename proc_sink.  Update all references.
        
Sat May  6 12:38:55 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  In this phase, remove
        PROCESS IF, which was deprecated anyway and can be easily
        simulated with TEMPORARY followed by SELECT IF.

        * procedure.c: (open_active_file) Don't call
        add_process_if_trns().
        (discard_variables) Get rid of redundant call to
        proc_cancel_all_transformations().
        (add_process_if_trns) Removed.
        (process_if_trns_proc) Removed.
        (process_if_trns_free) Removed.

Sat May  6 10:58:05 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  In this phase, add
        `const' to the case passed to procedure()'s callback.

        Updated all users of procedure() as well.
        
        * procedure.c: (struct write_case_data) Add "const" to ccase
        parameter for case_func member.
        (procedure) Add "const" to ccase parameter for proc_func
        parameter.
        (multipass_case_func) Make ccase parameter const.
        (internal_procedure) Add "const" to ccase parameter for case_func
        parameter.
        (split_procedure_case_func) Make ccase parameter const.
        (multipass_split_case_func) Make ccase parameter const.
        
Sat May  6 10:30:33 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  In this phase, get rid of
        the output code for SPLIT FILE groups in procedure.c, which really
        shouldn't be doing any output.  Move it into the individual
        procedures instead.  This also adds some flexibility.

        Updated many users of procedure_with_splits() and
        multipass_procedure_with_splits() to call
        output_split_file_values() and to deal with increased use of
        const.

        * procedure.c: (struct split_aux_data) Add "const struct ccase *"
        parameter to begin_func member.
        (procedure_with_splits) Add "const struct ccase *" parameter to
        begin_func parameter.  Make ccase parameter const in proc_func
        parameter.
        (split_procedure_case_func) Don't dump split file group.  Pass
        case to begin_func.
        (dump_splits) Moved to language/dictionary/split-file.c as
        output_split_file_values().
        (struct multipass_split_aux_data) Add "const struct ccase *"
        parameter to split_func member.
        (multipass_procedure_with_splits) Add "const struct ccase *"
        parameter to split_func parameter.
        (multipass_split_case_func) Save new SPLIT FILE case before
        outputting case.
        (multipass_split_output) Pass saved SPLIT FILE case to split_func.
        
Fri May  5 22:48:50 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  Change
        internal_procedure() so that it calls open_active_file() and
        close_active_file(), which isolates most of the actual procedure
        functionality.

        * procedure.c: (struct write_case_data) Rename `proc_func' member
        to `case_func' and update all references.
        (procedure) Rewrite as one-line wrapper around
        internal_procedure().
        (struct multipass_aux_data) New.
        (multipass_callback) Renamed multipass_case_func().  Use struct
        multipass_aux_data as auxiliary data.
        (multipass_end_func) New function.
        (multipass_procedure) Rewrite as wrapper for internal_procedure()
        that uses multipass_case_func, multipass_end_func.
        (internal_procedure) Add `end_func' argument.  Move optimization
        of trivial case in here.  Move call to open_active_file() and
        close_active_file() in here.  Now assert that vfm_source is
        non-null.
        (procedure_with_splits_callback) Rename
        split_procedure_case_func().
        (split_procedure_end_func) New function.
        (multipass_split_callback) Rename multipass_split_case_func.
        (multipass_split_end_func) New function.
        (discard_variables) No need to test for nonnull vfm_source.

Fri May  5 21:34:02 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  Get rid of unused member.

        * procedure.c: (struct write_case_data) Remove `cases_analyzed'
        member.
        (write_case) Don't increment cases_analyzed.

Thu May  4 21:50:11 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  In this phase, move
        procedure.c and procedure.h from src to src/data.  Update
        makefiles and #includes accordingly.

        * procedure.c: Moved here from src/.

        * procedure.h: Moved here from src/.

Wed May  3 22:42:12 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  In this phase, get rid of
        many global variables, consolidating procedure execution in
        procedure.c.  Encapsulate transformations in new "struct
        trns_chain".  Also, change implementation of N OF CASES, FILTER,
        and PROCESS IF from special cases to transformations.
         
        * automake.mk: (src_data_libdata_a_SOURCES) Add transformations.c,
        transformations.h.

        * dictionary.c: (global variable default_dict) Move to
        src/procedure.c.

        * variable.h: (TRNS_*) Move to transformations.h.
        (struct transformation) Move to transformations.c.

Thu May  4 13:47:06 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * sys-file-reader.c: Fixed invalid read problems.

Tue May  2 15:57:10 2006  Ben Pfaff  <blp@gnu.org>

        * storage-stream.c: Add missing function comments.

Tue May  2 15:50:21 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  In this phase, add some
        new, needed functionality to storage-stream.

        * storage-stream.c: (storage_source_decapsulate) New function.

Tue May  2 15:43:36 2006  Ben Pfaff  <blp@gnu.org>

        * variable.c (width_to_bytes): Declarations must precede
        statements for C90 compliance.

Tue May  2 10:42:05 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * data-out.c, data-in.c, variable.c, variable.h: New functions 
        copy_mangle and copy_demangle for reading/writing cases; emulates the 
        way SPSS deals with strings > 255 bytes.

        * sys-file-reader.c sys-file-writer.c: Added support for Record 7, 
        subtype 14 needed for strings longer than 255 bytes.

        * dictionary.c, format.def, value.c : Updated to use MAX_STRING 
        instead of literal values. Also fixed some constness issues.

        * format.h: Constness

        * sfm-private.h: Renamed the case_size identifier, since I discovered 
        that SPSS's respect for this variable is very nominal.

Mon May  1 15:45:42 2006  Ben Pfaff  <blp@gnu.org>

        Change case limit type from int to size_t.

        * dictionary.c: (struct dictionary) Change type of case_limit
        member.
        (dict_get_case_limit) Change return type.
        (dict_set_case_limit) Change parameter type.

Wed Apr 26 20:01:19 2006  Ben Pfaff  <blp@gnu.org>

        * variable.h: (struct variable) Rename `reinit' member as `leave'
        and invert sense.  Fix up all references.

Wed Apr 26 19:39:28 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming procedure execution.  In this phase, break
        procedure.c into multiple files.
        
        * automake.mk: (src_data_libdata_a_SOURCES) Add all the new files.

        * case-sink.c: New file.

        * case-sink.h: New file.
        
        * case-source.c: New file.

        * case-source.h: New file.
        
        * storage-stream.c: New file.

        * storage-stream.h: New file.

Wed Apr 26 14:55:19 2006  Ben Pfaff  <blp@gnu.org>

        * variable.h: (struct variable) Remove `init' member and all
        references to it from other files.  It was initialized in several
        places, but nothing really ever used it for anything worthwhile.
        Thanks to Jason Stover for pointing out how confusing this
        member is.

Sun Apr 23 22:04:45 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming error message support.  In this phase, get rid
        of message "titles" and put the message text in `struct error'.
        Now `struct error' encapsulates a message more properly.
        
        * casefile.c: (io_error) Use err_msg() instead of err_vmsg().
        Format message ourselves.

        * data-in.c: (vdls_error) Ditto.

        * por-file-reader.c: (error) Ditto.

        * sys-file-reader.c: (corrupt_msg) Ditto.

Sun Apr 16 18:49:51 2006  Ben Pfaff  <blp@gnu.org>

        GNU standards require "file name" instead of "filename" in
        documentation.  It's nice for our code to follow the convention
        too.
        
        * casefile.c: (struct casefile) Rename `filename' member to
        `file_name'.  Updated all references.

        * file-name.c: [!unix] (struct file_identity) Rename
        normalized_filename member to normalized_file_name.  Updated all
        references.

Sun Apr 16 18:35:33 2006  Ben Pfaff  <blp@gnu.org>

        We don't really support anything but Unix-like environments well,
        so we might as well de-obfuscate by writing directory and path
        separators explicitly.

        * file-name.h: (macro DIR_SEPARATOR) Removed.  Changed all usages
        to just '/'.
        (macro PATH_SEPARATOR) Removed.  Changed all usages to just ':'.
        (macro DIR_SEPARATOR_STRING) Removed.  Changed all usages to just
        "/".
        (macro PATH_SEPARATOR_STRING) Removed.  Changed all usages to just
        ":"

Sun Apr 16 18:28:35 2006  Ben Pfaff  <blp@gnu.org>

        GNU standards require "file name" instead of "filename" in
        documentation.  It's nice for our code to follow the convention
        too.
        
        * filename.c: Rename to file-name.c.

        * filename.h: Rename to file-name.h.  Update all inclusions.
        Update header guards.

        * automake.mk: Update file names.

Sun Apr 16 16:42:47 2006  Ben Pfaff  <blp@gnu.org>

        * filename.c: (fn_dirname) Renamed fn_dir_name(), all references
        updated.
        (fn_basename) Removed (dead code).
        (fn_absolute_p) Renamed fn_is_absolute(), all references updated.
        (fn_special_p) Renamed fn_is_special(), all references updated.
        (fn_exists_p) Renamed fn_exists(), all references updated.

Sun Apr 16 16:33:58 2006  Ben Pfaff  <blp@gnu.org>

        * filename.c: (fn_tilde_expand) Rewrite for cleaner code.  
        Also, now it only tilde-expands file names, not paths.
        (fn_search_path) Tilde-expand one directory at a time.

Sun Apr 16 16:28:06 2006  Ben Pfaff  <blp@gnu.org>

        * filename.c: (fn_search_path) rewrite for cleaner code.  Also,
        get rid of non-Unixlike version of the code, which has probably
        never been tested.
        (fn_prepend_dir) Removed (dead code).

        * filename.h: (macro DIR_SEPARATOR_STRING) New.
        (macro PATH_SEPARATOR_STRING) New.
Sun Apr 16 16:05:28 2006  Ben Pfaff  <blp@gnu.org>

        Continue reforming error message support.  In this phase, we get
        rid of VM() and the other msg() support for "verbosity", replacing
        it by a new function verbose_msg().

        * filename.c: (fn_search_path) Use verbose_msg() instead of
        msg(VM(), ...).  

Sat Apr 15 19:53:19 2006  Ben Pfaff  <blp@gnu.org>

        * sfm-private.h: Get rid of #defines after #error, which makes no
        sense.

Sat Apr 15 19:48:57 2006  Ben Pfaff  <blp@gnu.org>

        Get rid of our own int32 type in favor of the standard int32_t
        type.
        
        * sfm-private.h: (int32 macro) Don't define this anymore.  Do
        include <stdint.h>.

        * sys-file-reader.c: Use int32_t instead of int32 throughout.
        
        * sys-file-writer.c: Use int32_t instead of int32 throughout.

Sat Apr 15 19:36:47 2006  Ben Pfaff  <blp@gnu.org>

        Remove ill-considered file routines that are no longer used.
        
        * filename.c: (fn_open_ext) Removed.
        (fn_close_ext) Removed.

        * filename.h: (struct file_ext) Removed.

Mon Apr  3 13:22:39 2006  Ben Pfaff  <blp@gnu.org>

        * variable.c (var_is_valid_name): Move declarations before code
        for C90 compliance.

Tue Apr  4 15:28:40 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * filename.ch (fn_interp_vars): Fixed small buglet.

Tue Mar 28 13:47:16 WST 2006 John Darrington <john@darrington.wattle.id.au>
        
        * filename.[ch] (fn_interp_vars): Changed the signature and semantics
        so as to modify the string inline.   Thus makeing it easier to
        destroy the results when no longer needed.
        
2006-03-25  Jason Stover  <jhs@math.gcsu.edu>

        * category.c (cat_stored_values_destroy): Fixed memory leak.

Fri Mar 24 18:15:41 2006  Ben Pfaff  <blp@gnu.org>

        Add some missing frees.  Thanks to John Darrington for reporting
        these.

        * any-writer.c (any_writer_close): Free writer.

        * any-reader.c (any_reader_close): Free reader.

Mon Mar 20 16:33:53 2006  Ben Pfaff  <blp@gnu.org>

        * por-file-reader.c: (error) Mark as NO_RETURN.

Sat Mar 11 15:06:07 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * settings.c: Changed default value of scompress to true.

Sat Mar  4 13:22:51 2006  Ben Pfaff  <blp@gnu.org>

        * sfm-private.h: Include variable.h, to get SHORT_NAME_LEN.

        * value.h: Remove check on MAX_SHORT_STRING, which I don't think
        really applies.

        * variable.h: Move definition of SHORT_NAME_LEN, LONG_NAME_LEN
        here from pref.h.orig.

Sat Mar  4 12:50:48 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * sys-file-reader.c: Fixed bug reading compressed files.

Thu Mar  2 08:40:33 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * Numerous renames.  See src/ChangeLog for details.
        
        * Moved files from src directory