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

2007-07-11  Ben Pfaff  <blp@gnu.org>

        * get.c (map_case): Create destination case instead of leaving it
        undefined.  Fixes bug #20285.
        Reviewed by John Darrington.

2007-06-06  Ben Pfaff  <blp@gnu.org>

        * get.c: Essentially rewrite MATCH FILES to support FIRST and
        LAST.

2007-06-06  Ben Pfaff  <blp@gnu.org>

        Adapt case sources, sinks, and clients of procedure code to the
        new infrastructure.
        
        * data-list.c: Make DATA LIST into a casereader.

        * get.c: Change GET, IMPORT, SAVE, EXPORT to use casereaders,
        casewriters.

        * inpt-pgm.c: Use caseinit code.  Turn INPUT PROGRAM into a
        casereader.

        * list.q: Adapt to new procedure code.

2007-05-06  Ben Pfaff  <blp@gnu.org>

        Abstract the documents within a dictionary a little better.
        Thanks to John Darrington for suggestion, initial version, and
        review.  Patch #5917.

        * get.c (mtf_merge_dictionary): Simplify creating merged document.

        * sys-file-info.c (display_documents): Use new
        dict_get_document_line_cnt and dict_get_document_line functions.

Thu Feb  1 16:56:02 2007  Ben Pfaff  <blp@gnu.org>

        * file-handle.q (fh_parse): Update to new fh_create_file
        prototype.

Sat Dec 16 22:16:18 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.

        * automake.mk: Removed matrix-data.c.

        * matrix-data.c: Removed.

        * data-list.c (data_list_source_read): Conform with new
        case_source_class interface.
        (data_list_source_destroy): Ditto.

        * get.c (case_reader_source_class): Ditto.
        (case_reader_source_destroy): Ditto.
        (parse_output_proc): Take advantage of new procedure interface.
        (output_proc): Removed.
        (struct mtf_file): Add "struct ccase *" member to allow use of new
        procedure interface.
        (cmd_match_files): Take advantage of new procedure interface.
        (mtf_processing_finish): Removed.
        (mtf_read_nonactive_records): Renamed mtf_read_records.  Now reads
        from every file, without any exception for the active file.
        (mtf_compare_BY_values): Simplify for new interface.
        (mtf_processing): Simplify for new interface.

        * inpt-pgm.c (is_valid_state): New function.
        (input_program_source_read): Conform with new case_source_class
        interface.
        (input_program_source_destroy): Ditto.
        (end_case_trns_proc): Now just needs to return TRNS_END_CASE.

Sat Dec  9 18:43:34 2006  Ben Pfaff  <blp@gnu.org>

        * list.q (cmd_list): Use new var_create, var_destroy functions.

Thu Nov 30 21:51:58 2006  Ben Pfaff  <blp@gnu.org>

        * inpt-pgm.c (cmd_reread): Always return error code upon detecting
        syntax error.  Fixes bug #18419.  Thanks to John Darrington for
        reporting this bug.

Sun Nov 19 09:17:45 2006  Ben Pfaff  <blp@gnu.org>

        * data-list.c (parse_free): Follow documented (but odd) rule that
        N format is treated as F format for free-field input.
        
        * data-reader.c (read_file_record): Drop new-line character from
        input text lines.  This is symmetrical with the recently changed
        dfm_put_record semantics.

Thu Nov  2 20:56:03 2006  Ben Pfaff  <blp@gnu.org>

        Implement SKIP keyword on DATA LIST.  Fixes bug #17099.
        
        * data-list.c: (struct data_list_pgm) Add `skip_records' members.
        (cmd_data_list) Set skip_records based on user input.
        (data_list_source_read) Skip records requested by user.

Tue Oct 31 20:04:06 2006  Ben Pfaff  <blp@gnu.org>

        * placement-parser.c: (PRS_TYPE_T) Now that struct fmt_spec uses
        an enum fmt_type for its type member, we can't depend on the
        ability to put negative values into that member as out-of-band
        values, because enum fmt_type might be an unsigned type.  So use
        values around SCHAR_MAX instead, because we know that SCHAR_MAX
        will fit into any type, signed or unsigned, and there aren't
        nearly that many format types.
        (parse_var_placements) Add for_input parameter to specify whether
        we're parsing input or output formats.  Update all callers.
        (fixed_parse_columns) Ditto.
        (fixed_parse_fortran) Ditto.
        
Tue Oct 31 18:21:48 2006  Ben Pfaff  <blp@gnu.org>

        * print-space.c (print_space_trns_proc): Let dfm_put_record add
        the new-line character, to match dfm_put_record change below.

Sat Oct 28 11:57:19 2006  Ben Pfaff  <blp@gnu.org>

        * data-writer.c (struct dfm_writer): Removed `bounce' member, and
        all references to it.
        (dfm_put_record) Change semantics so that it adds formatting
        itself, such as new-line characters, instead of putting that
        responsibility on the caller.  Also, pad binary records with
        spaces instead of zeros, for compatibility.

        * print.c (struct prt_out_spec) New member `sysmis_as_spaces'.
        (struct print_trns) Remove `omit_new_lines' and all references,
        since dfm_put_record() is taking care of that.  Add
        `include_prefix'.
        (internal_cmd_print) Allow an empty set of data to print.  Set
        include_prefix.
        (parse_specs) Allow an empty set of data to print.
        (parse_variable_argument) Only add space with PRINT command.  Set
        sysmis_as_spaces.
        (print_trns_proc) Indent records if include_prefix is set, for
        compatibility.  Output SYSMIS as spaces if sysmis_as_spaces is
        set.  Put "1" in first column if PRINT EJECT is used with an
        external output file.
        (flush_records) Ditto.

Sat Oct 28 16:19:57 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * data-reader.c: Eliminated references to extern variable getl_buf

Sat Aug  5 08:25:07 2006  Ben Pfaff  <blp@gnu.org>

        Fix bug #17329 in REREAD parsing, reported by John Darrington.

        * inpt-pgm.c (cmd_reread): Fix file handle parsing.

Mon Jul 31 10:32:31 2006  Ben Pfaff  <blp@gnu.org>

        * print.c (parse_specs): Allow a comma between specifications.

Sun Jul 16 19:57:10 2006  Ben Pfaff  <blp@gnu.org>

        * automake.mk: (src_language_data_io_libdata_io_a_SOURCE) Add
        print-space.c, placement-parser.c, placement-parser.h.

        * data-list.c: Basically rewrote the whole thing.  Broke out a lot
        of code into placement-parser.c.  Code is much cleaner now.

        * placement-parser.c: New file.

        * placement-parser.h: New file.

        * print.c: Basically rewrote the whole thing.  Broke out PRINT
        SPACE into print-space.c.  Code is much cleaner now.

        * print-space.c: New file.

Sat Jul  1 17:39:40 2006  Ben Pfaff  <blp@gnu.org>

        Fix bug #11612, "q2c documentation does not agree with code".
        
        * list.q: Audit use of q2c "+" prefixes that indicate that a
        command may appear multiple times.

Sat Jul  1 20:44:22 2006  Ben Pfaff  <blp@gnu.org>

        Fix bug #15786: System File Creation crashes if directoy is
        nonexistent.
        
        * get.c (parse_write_command): Check that the any_writer open
        succeeds.

Tue Jun 27 22:44:28 2006  Ben Pfaff  <blp@gnu.org>

        Fix regression in command name completion reported by John
        Darrington.  Now completion is again state-dependent and occurs
        only on the first line of a command.
        
        * inpt-pgm.c: (cmd_input_program) Reading of first token in
        command moved into cmd_parse.

Fri Jun  9 13:56:00 2006  Ben Pfaff  <blp@gnu.org>

        Reform string library.
        
        * matrix-data.c (context): Use dynamic string.
        (another_token) Deal with changed dfm_get_record() interface.
        (mget_token) Ditto.
        (force_eol) Ditto.

        * data-list.c (struct data_list_pgm) Delete delims, delim_cnt
        members, replacing them by struct string delims.  Update all
        references to use struct string functions.
        (cut_field) Change interface to avoid needing "end_blank", by
        getting the data-reader to remember that state for us.  Change
        internals to use substring.  Update both callers.

        * data-reader.c (read_file_record): Use ds_read_stream().
        (dfm_get_record) Change interface to return substring.  Updated
        all callers.
        (dfm_expand_tabs) Use ds_find_char().  Now maintain position
        relative to end-of-line.  Use ds_swap().
        (dfm_reread_record) Don't limit position by line length.
        (dfm_column_start) Make parameter const.
        (dfm_columns_past_end) New function.
        (dfm_get_column) New function.

Thu May 25 18:26:26 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * print.c (print_trns_free): Made the code agree with the comment, by
        not freeing PRT.  Has the side effect that the command no longer
        crashes on invalid syntax.

Tue May  9 20:55:46 2006  Ben Pfaff  <blp@gnu.org>

        * get.c (cmd_match_files): Fix memory leak replacing default_dict.

Sat May  6 22:25:09 2006  Ben Pfaff  <blp@gnu.org>

        Fix segfault.
        
        * list.q (write_fallback_headers): (write_fallback_headers)
        Properly record width of leader and pass it to write_varname().

Sat May  6 19:03:13 2006  Ben Pfaff  <blp@gnu.org>

        * get.c: (mtf_merge_dictionary) Fix value label memory leak.

Sat May  6 13:51:16 2006  Ben Pfaff  <blp@gnu.org>

        Use a casefile, instead of a case sink, for MATCH FILES output.
        It's more straightforward, although it has the same effect.
        
        * get.c: (struct mtf_proc) Replace `sink' case sink member by
        `output' casefile member.
        (cmd_match_files) Work with casefile instead of sink.
        (mtf_processing) Add case to casefile instead of sink.

Sat May  6 10:43:07 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.

        * list.q (write_all_headers): Call output_split_file_values().

Wed May  3 23:00:17 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.
         
        * data-list.c: (data_list_trns_proc) Return TRNS_END_FILE at end
        of file.  (Why didn't we do this before?)
        (cmd_match_files) Direct procedure output to null sink.
        Use discard_variables() instead of indirect version.

        * inpt-pgm.c: Use transformation chain.
        (struct input_program_pgm) Add trns_chain member.
        (cmd_input_program) Initialize trns_chain member and capture
        transformations with proc_capture_transformations().
        (input_program_source_read) Use trns_chain_execute().
        (destroy_input_program) Destroy input chain.

Tue May  2 10:39:56 WST 2006 John Darrington <john@darrington.wattle.id.au>

        * list.q Changed from using fixed length char buffers to struct 
        string so that any length variables can be used.

Mon May  1 18:21:19 2006  Ben Pfaff  <blp@gnu.org>

        Further clean up the CMD_* command result codes.
        
        * (enum cmd_result_extensions) New.  Add CMD_END_INPUT_PROGRAM and
        CMD_END_CASE result codes.
        (struct input_program_pgm) Added case_nr, write_case, wc_data
        members for use by END CASE transformation.
        (emit_END_CASE) New function.
        (cmd_input_program) Interpret CMD_END_CASE by outputting an END
        CASE transformation.  If none is output by the input program
        itself, add one automatically at the end.  Change lack of
        variables from warning to error.
        (cmd_end_input_program) Return CMD_END_INPUT_PROGRAM instead of
        CMD_END_SUBLOOP.
        (input_program_source_read) No longer any need to special-case END
        CASE.  Handle TRNS_DROP_CASE properly.  Initialize new members in
        inp for use by END CASE transformation.
        (destroy_input_program) New function.
        (input_program_source_destroy) Just call destroy_input_program().
        (cmd_end_case) Just return CMD_END_CASE.
        (end_case_trns_proc) No longer a stub handled by
        input_program_source_read().  Actually output the case and
        increment the case number.

Mon May  1 16:06:30 2006  Ben Pfaff  <blp@gnu.org>

        Remove vestiges of REPEATING DATA support.
        
        * data-list.c: (struct rpd_num_or_var) Removed.
        (struct repeating_data_trns) Removed.
        (cmd_repeating_data) Removed.
        (find_variable_input_spec) Removed.
        (parse_num_or_var) Removed.
        (parse_repeating_data) Removed.
        (realize_value) Removed.
        (struct rpd_parse_info) Removed.
        (rpd_parse_record) Removed.
        (repeating_data_trns_proc) Removed.
        (repeating_data_trns_free) Removed.
        (repeating_data_set_write_case) Removed.
        (rpd_msg) Removed.

        * inpt-pgm.c: (input_program_source_read) Don't deal with
        REPEATING DATA.

        * data-list.h: Removed.

        * automake.mk (src_language_data_io_libdata_io_a_SOURCES): Removed
        data-list.h.

Mon May  1 15:58:28 2006  Ben Pfaff  <blp@gnu.org>

        Remove vestiges of FILE TYPE support. 
        
        * data-list.c: (cmd_data_list) Don't check for FILE TYPE.
        (cmd_repeating_data) Ditto.
        
        * automake.mk (src_language_data_io_libdata_io_a_SOURCES): Remove
        file-type.c, file-type.h.

        * file-type.c: Removed.
        
        * file-type.h: Removed.

Wed Apr 26 13:16:28 2006  Ben Pfaff  <blp@gnu.org>

        Improve the way we handle the various parsing "states".  Until now
        we've hard-coded the state transitions in the command definition
        file, but that's error-prone and, worse, it's redundant--we can
        figure out what state we're in anyhow.  We can cleanly handle
        INPUT PROGRAM and FILE TYPE with a nested command-processing loop.
        
        * data-list.c: (cmd_data_list) Use in_file_type() or
        in_input_program() in place of case_source_is_class() or
        case_source_is_complex().

        * file-type.c: NB: Not really fixed except minimally to compile,
        because it doesn't work anyway.
        (in_file_type) New function.
        (cmd_record_type) No need to check that we're in FILE TYPE.
        (cmd_end_file_type) Ditto.
        (var file_type_source_class) Make static.

        * get.c: (cmd_match_files) Check vfm_source instead of pgm_state.

        * inpt-pgm.c: (in_input_program) New function.
        (cmd_input_program) Rewrite to include nested command processing
        loop.
        (cmd_end_input_program) Just return CMD_END_SUBLOOP.
        (var input_program_source_class) Make static.
        (cmd_end_case) No need to check that we're in INPUT PROGRAM.
        (cmd_end_file) Ditto.
        
        * automake.mk (src_language_data_io_libdata_io_a_SOURCES): Add
        file-type.h, inpt-pgm.h.

        * file-type.h: New file.

        * inpt-pgm.h: New file.

Tue Apr 25 13:11:55 2006  Ben Pfaff  <blp@gnu.org>

        * print.c: Don't special-case MS-DOS line terminators.
        (macro LINE_END_WIDTH) Removed.
        (alloc_line) Line ends are 1 byte.
        (print_trns_proc) Just output \n for line end.

Sun Apr 23 22:05:58 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.
        
        * data-list.c: (macro RPD_ERR) Removed.
        (rpd_msg) New function.  Updated all references to tmsg() to call
        this function instead.

Sat Apr 15 19:38:13 2006  Ben Pfaff  <blp@gnu.org>

        Remove last users of struct file_ext so we can get rid of it
        entirely.
        
        * data-reader.c: (struct dfm_reader) Change file member from
        struct file_ext to FILE *.  Updated all references.
        (dfm_close_reader) Close file with fn_close() instead of
        fn_close_ext().  Also, make a copy of the file name from the file
        handle before closing it, because we can't extract it after we
        close the file.
        (dfm_open_reader) Open file with fn_open() instead of
        fn_open_ext().

        * data-writer.c: (struct dfm_writer) Change file member 
        struct file_ext to FILE *.  Updated all references.
        (dfm_close_writer) Close file with fn_close() instead of
        fn_close_ext().  Also, make a copy of the file name from the file
        handle before closing it, because we can't extract it after we
        close the file.
        (dfm_open_writer) Open file with fn_open() instead of
        fn_open_ext().

Sat Apr 15 18:00:32 2006  Ben Pfaff  <blp@gnu.org>

        * data-list.c: Add prototype to suppress warning for
        cmd_repeating_data().
        
Thu Mar  2 08:40:33 WST 2006 John Darrington <john@darrington.wattle.id.au>
        
        * Moved files from src directory