Ben Pfaff [Sun, 6 Feb 2011 05:10:10 +0000 (21:10 -0800)]
sys-file-reader: Refactor to clean up character encoding support.
The system file format is unusual in that it does not record the encoding
used by character strings at the beginning or at any fixed place in the
file. Instead, it can be recorded practically anywhere in the file. It
never precedes all of the actual character strings in the file, which makes
it impossible to interpret those strings completely and correctly until it
is encountered.
Until now, the system file reader has dealt with this situation by
stuffing uninterpreted character strings into data structures until the
encoding is known, then at that point fetching out the character strings,
reencoding them, and stuffing them back into the data structures. This
does work, but it has the disadvantage that all of the PSPP data
structures have to tolerate character strings with unknown encoding. In
some cases this seems like an ugly situation. For example, arbitrary
variable names have to be supported, even though the syntax for variable
names is circumscribed by the language, because the syntax rules for
variable names cannot be completely and correctly applied to a string that
is in an unknown encoding.
This commit fixes that problem by adopting a new way to read system files.
Each record in the system file dictionary is essentially slurped into
memory as a chunk, then the character encoding is extracted from it, then
the rest of the dictionary is interpreted based on that encoding. The
actual implementation is a little more intricate because the format of
system file records is somewhat non-uniform.
Ben Pfaff [Sat, 12 Mar 2011 06:10:54 +0000 (22:10 -0800)]
FREQUENCIES: Fix percentiles calculation.
The condition for using a variate directly instead of interpolating was
just wrong. It would interpolate in cases where it clearly should not,
which produced incorrect results in many cases.
Thanks to Fabio Bordignon <bordignon@demos.it> for reporting the problem
and supplying a simple test case.
Ben Pfaff [Fri, 11 Mar 2011 06:53:17 +0000 (22:53 -0800)]
T-TEST: Fix use-after-free with TEMPORARY and independent samples.
When TEMPORARY is in effect, proc_commit() destroys the temporary
dictionary. This means that any procedure that does not somehow disable
temporary transformations and refers to a variable following proc_commit()
has a use-after-free error.
T-TEST has two different bugs of this type. First, the loop that destroys
group statistics refers to destroyed variables. This commit fixes this
problem by instead using variable aux data destructors to destroy group
statistics.
Second, when there is an independent variable, destroying its values
requires knowing the variable's width. This commit fixes this problem by
destroying the values before calling proc_commit().
The AUTORECODE, DESCRIPTIVES, RANK, and REGRESSION procedures appear to
have similar issues (not fixed by this commit).
Reported by Jeremy Lavergne <jeremy@lavergne.gotdns.org>.
Ben Pfaff [Fri, 25 Feb 2011 05:31:44 +0000 (21:31 -0800)]
README.Git: Update to newer commit.
This is the Gnulib commit I'm testing against locally. It should fix the
problem that showed up in the nightly build, in which the dtoastr module
was missing.
Ben Pfaff [Sat, 19 Feb 2011 20:55:54 +0000 (12:55 -0800)]
data-out: Optimize and fix some bad assumptions.
Until now, data_out_pool() and its wrapper function data_out() have always
done at least two memory allocations: one to fill in the initial version
of the result and another to recode it to UTF-8. However, recoding to
UTF-8 is usually unnecessary, because most output formats always produce
output in UTF-8 anyway. Only binary formats and the string A format ever
produce data in other encodings, so this commit drops recoding entirely
except for those cases. Binary formats are a particularly special case:
usually it doesn't make any sense to use these formats for text output,
but this commit does its best to translate the binary output bytes into
valid UTF-8, at least up to the first null byte.
This commit also finishes fixing up display widths.
The closely related data_out_legacy() function, which only has one user
in three also needed some work. It was badly named, so I renamed it to
data_out_recode(). It made the bad assumption that the data passed in
was encoded in ASCII (written C_ENCODING). It also made the bad
assumption that the number of bytes output would be exactly the format's
width. This rewrite fixes these problems.
Ben Pfaff [Sat, 19 Feb 2011 05:55:18 +0000 (21:55 -0800)]
format: Count prefix and suffix width in terms of display columns.
Until now, the prefixes and suffixes for custom currency formats
(CCA, etc.) have been considered to occupy one display column per
byte. This is fine for prefixes and suffixes like "$" or "%", but
falls down badly with U+00A5 (¥) or U+20AC (€), which occupy two
or three bytes, respectively, in UTF-8, while occupying only a
single display column.
This commit fixes the problem. It doesn't add a test yet because
there are still some higher-level issues, but that will come in
a later commit when those remaining issues are resolved.
Ben Pfaff [Wed, 16 Feb 2011 06:00:33 +0000 (22:00 -0800)]
format: Create a new "struct fmt_affix" for prefix and suffix strings.
This combines two changes: changing the string type for numeric
prefix and suffix strings from "struct substring" to plain "char *",
and putting the string inside a new structure. Both of these will
make more sense after the following commit, which adds another member
to the new structure and stops using the length of the string in so
many places (which is the reason that "struct substring" was a good
choice).
Ben Pfaff [Sat, 19 Feb 2011 06:30:00 +0000 (22:30 -0800)]
data-out: Make each converter responsible for storing null terminator.
Until now, every converter has produced output that is exactly as many
bytes long as the format's width. In upcoming patches this will change,
because in UTF-8 a character that occupies a single display column can
require multiple bytes. In preparation, this commit requires the
individual converters to write out their own null byte, giving a bit
more flexibility regarding length.
Ben Pfaff [Tue, 15 Feb 2011 07:17:11 +0000 (23:17 -0800)]
format: Increase abstraction of fmt_number_style.
Upcoming commits will make some changes to fmt_number_style, so it
seems best to avoid having clients actually construct and modify
instances of this structure. This commit makes that change.
We could take it one step further and add accessor functions even
for reading out the structure, but in my opinion that would be
overkill for this structure.
Ben Pfaff [Sun, 13 Feb 2011 19:49:30 +0000 (11:49 -0800)]
i18n: Introduce C_ENCODING as replacement for LEGACY_NATIVE.
The LEGACY_NATIVE name seems a bit awkward for something that is just the
name of the encoding for strings in C source code, that is, the C locale,
so this commit renames it to C_ENCODING and moves it to i18n.h with the
rest of the encoding-related functions.
In addition, PSPP assumes in various places that the local system has
ASCII-based locales. I don't think there's much point in pretending to
support EBCDIC, so this commit removes that little bit of support.
Ben Pfaff [Mon, 14 Feb 2011 06:20:45 +0000 (22:20 -0800)]
Use new Gnulib function dtoastr() to format short, accurate real numbers.
%.*g with DBL_DIG + 1 as argument is simple but in rare cases it fails to
accurately format a real number. The recently added Gnulib routine
dtoastr() always formats a real number accurately, so switch to using it
for these cases.
Ben Pfaff [Sat, 12 Feb 2011 16:02:48 +0000 (08:02 -0800)]
PRINT SPACE: When an output file is specified, don't ignore expression.
When both OUTFILE= and an expression were specified on PRINT SPACE, the
expression was ignored (if it was only a single token) or an error would
occur (if it was was more than one token).
Ben Pfaff [Sun, 12 Dec 2010 22:00:28 +0000 (14:00 -0800)]
float-format: Eliminate tests' dependence on exact string encoding.
Until now, the float-format tests have depended on the PSPP syntax
accepting arbitrary byte values in strings, without treating them as part
of any particular encoding. The lexer is being rewritten so that this
assumption is no longer true, so this commit eliminates this assumption in
the float-format tests. After this commit, the tests only use ASCII
characters in strings.
Ben Pfaff [Sun, 2 Jan 2011 00:42:06 +0000 (16:42 -0800)]
por-file-reader: Remove dependency on VAR_NAME_LEN.
VAR_NAME_LEN wasn't really needed here because we knew that the name we
started from was no more than 8 bytes long. Also, we know that we can
come up with a unique name within ULONG_MAX tries since we'd run out of
memory before running out of values to try.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sun, 2 Jan 2011 00:26:25 +0000 (16:26 -0800)]
REGRESSION: Eliminate restriction to VAR_NAME_LEN in reg_get_name().
There's still an obvious problem here that the prefix isn't being trimmed
down so that the suffix will fit. Since an upcoming series of changes
would have to completely redo how this would be done, I'm not fixing that
now, only marking it with XXX.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sun, 2 Jan 2011 00:16:30 +0000 (16:16 -0800)]
FLIP: Eliminate false dependency on VAR_NAME_LEN.
The size of this buffer really shouldn't have anything to do with
VAR_NAME_LEN but with the string about to be put into it.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 23:54:54 +0000 (15:54 -0800)]
variable-parser: Drop VAR_NAME_LEN restriction from var_set_lookup_var_idx().
This restriction is purely artificial, as part of an assertion. Since
longer variable names are going to have to be supported, remove it.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 21:00:49 +0000 (13:00 -0800)]
variable-parser: Rewrite parse_DATA_LIST_vars().
This rewrite was prompted by getting rid of the VAR_NAME_LEN limit inside
parse_DATA_LIST_vars(), but then I noticed that the variable naming and
coding style was dated, and that duplicate variable names were only
detected for variables named using TO, not for individual names, so I
rewrote much of the code instead.
Ben Pfaff [Sat, 1 Jan 2011 19:15:33 +0000 (11:15 -0800)]
DEBUG EVALUATE: Eliminate VAR_NAME_LEN limit.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 19:11:07 +0000 (11:11 -0800)]
VECTOR: Eliminate VAR_NAME_LEN limit for variable names.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 19:09:01 +0000 (11:09 -0800)]
MODIFY VARIABLES: Eliminate VAR_NAME_LEN limit on variable names.
This is actually a general code improvement, since it eliminates memory
allocation and copying that was not actually necessary.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 18:59:03 +0000 (10:59 -0800)]
GET DATA /TYPE=TXT: Get rid of VAR_NAME_LEN limit on variable names.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 18:56:55 +0000 (10:56 -0800)]
FILE HANDLE: Get rid of VAR_NAME_LEN limit on handle name.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 18:52:14 +0000 (10:52 -0800)]
combine-files: Eliminate VAR_NAME_LEN restriction from combine_files().
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 18:47:56 +0000 (10:47 -0800)]
vector: Remove VAR_NAME_LEN limit for internal representation of name.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sat, 1 Jan 2011 18:45:11 +0000 (10:45 -0800)]
variable: Remove VAR_NAME_LEN limit for internal representation of name.
Most uses of VAR_NAME_LEN within PSPP are wrong due to encoding issues:
the limit applies to variable names in the encoding used by the data
set, but most uses of VAR_NAME_LEN actually limit the length of a name
in UTF-8. The UTF-8 representation of a name can be longer or shorter
than its representation in the data set encoding, so it seems best to
eliminate references to VAR_NAME_LEN entirely.
Ben Pfaff [Sun, 30 Jan 2011 23:33:30 +0000 (15:33 -0800)]
sys-file-reader: Fix double-free on read of partial record.
partial_record() does a longjmp() to code near the top of
sys_file_casereader_read() that unrefs the case, but this code at the
end of sys_file_casereader_read() would have already unref'ed it once in
that case, resulting in a double-free.
Ben Pfaff [Mon, 31 Jan 2011 04:34:32 +0000 (20:34 -0800)]
hash-functions: Use C instead of preprocessor to test size of double.
I don't see a reason that this has to be done as an #if directive. The
compiler should be easily able to optimize out the branch that will never
be taken. This allows us to get rid of a configure check, so it seems
worthwhile to me.
Ben Pfaff [Sun, 2 Jan 2011 16:37:31 +0000 (08:37 -0800)]
README: Describe meaning of copyright year ranges.
This is required by the GNU maintainers guide as a prerequisite for using
copyright year ranges. So far, PSPP itself doesn't use ranges, but Gnulib
(which PSPP uses) does.
Ben Pfaff [Sat, 1 Jan 2011 21:42:04 +0000 (13:42 -0800)]
expressions: Fix test failures due to new year.
Some of the date-related test cases assumed that year 41 would be
interpreted as 1941, but due to the 30-year century window, as soon as it
became 2011 this assumption failed. Write these years explicitly as 1941
instead to fix the problem.
Ben Pfaff [Tue, 21 Dec 2010 08:12:09 +0000 (00:12 -0800)]
NPAR TESTS: Fix two test failures.
Tests "NPAR TESTS CHISQUARE with DESCRIPTIVES" and "NPAR TESTS CHISQUARE,
listwise missing" were failing because of some unexpected behavior of
lex_put_back(). This fixes the problem by avoiding lex_put_back()
entirely.
Ben Pfaff [Mon, 6 Dec 2010 06:21:14 +0000 (22:21 -0800)]
NPAR TESTS: Prefer lex_match_id() over lex_match_hyphenated_word()
This code used lex_match_hyphenated_word() in many places where it was
necessary to match a plain identifier. But it is better to use
lex_match_id() when it can be, because it is simpler and faster, so this
commit does that.
Ben Pfaff [Mon, 6 Dec 2010 06:19:34 +0000 (22:19 -0800)]
q2c: Prefer lex_match_id() over lex_match_hyphenated_word() in emitted code.
Until now, q2c has used lex_match_hyphenated_word() wherever it was
necessary to match a plain identifier. But it is better to use
lex_match_id() when it can be, because it is simpler and faster, so this
commit does that.
Ben Pfaff [Sun, 12 Dec 2010 06:17:01 +0000 (22:17 -0800)]
DEBUG XFORM FAIL: Issue an error message as part of transformation.
PSPP only returns a failure exit status if an error message is issued, but
DEBUG XFORM FAIL doesn't issue an error message. This means that, if PSPP
has no other errors, then it's difficult to tell that the transformation
really did what it should have. This commit fixes the potential problem.
Ben Pfaff [Sun, 12 Dec 2010 06:07:14 +0000 (22:07 -0800)]
categoricals: Make parameter to categoricals_done() const.
From a high-level point of view, categoricals_done() does not actually
mutate any data in a categoricals object. Instead, it just computes a
cache. So it seems reasonable to make it take a "const" parameter, and
this fixes a warning:
oneway.c: In function ‘run_oneway’:
oneway.c:502: warning: initialization discards qualifiers from pointer target type
Ben Pfaff [Sun, 12 Dec 2010 05:29:55 +0000 (21:29 -0800)]
covariance: Fix const-ness of covariance_calculate[_unnormalized] retval.
The covariance_calculate() and covariance_calculate_unnormalized()
functions documented that their return values were owned by the covariance
object itself and thus should not be freed by the caller. However, this
documentation was incorrect, because in fact the covariance object did not
retain the pointer at all. This commit fixes the comments, updates
the return values to be non-"const", and changes the callers to free the
returned "gsl_matrix" objects.
Ben Pfaff [Sun, 12 Dec 2010 05:09:17 +0000 (21:09 -0800)]
friedman: Fix GCC warning about uninitialized fr.w.
On my machine GCC 4.4 complains:
friedman.c: In function ‘friedman_execute’:
friedman.c:106: warning: ‘fr.w’ may be used uninitialized in this function
This appears to be because GCC can't see that the assignment to fr.w and
its later use (in show_sig_box()) both have the same condition. At any
rate, this commit suppresses the warning by always initializing fr.w.
Ben Pfaff [Sun, 5 Dec 2010 14:43:51 +0000 (06:43 -0800)]
gui: Factor out duplicated code for executing and pasting syntax.
Reducing code duplication is good on its own. This will also make it
easier in upcoming patches to swap out "getl_interface" with a new
structure, by eliminating most of the references to getl_interface.
Ben Pfaff [Sun, 21 Nov 2010 20:00:32 +0000 (12:00 -0800)]
Remove the PROMPT, CPROMPT, and DPROMPT settings.
These settings will be insufficient to configure prompting for the new
lexer, which is able to detect more special cases in syntax and report
them through the prompt. (However, if users actually want this feature,
then we could add it; it's not difficult.)
Ben Pfaff [Sat, 20 Nov 2010 05:05:04 +0000 (21:05 -0800)]
lexer: Change the functions for retrieving token strings.
Until now, lex_tokid() has been for T_ID tokens only and lex_tokstr() has
been for T_ID and T_STRING tokens. For T_ID tokens, lex_tokid() and
lex_tokstr() had slightly different semantics.
This doesn't entirely make sense, and these particular functions are not
the ones wanted most by clients, so this commit removes these functions
in favor of lex_tokcstr() and lex_tokss(), which are both applicable
to both T_ID and T_STRING tokens, with the same semantics in each case.
These functions are also easier for the upcoming reimplementation of the
lexer.
Ben Pfaff [Tue, 20 Jul 2010 05:52:18 +0000 (22:52 -0700)]
lexer: Get rid of lex_negative_to_dash().
This function stands in the way of replacing the lexer in a
straightforward way, because it disrupts the stream of tokens: the
lexer can't tell in advance whether, e.g., "-4" is one token or
two. Fortunately, it is not used in very many places and is
relatively easy to remove.
Ben Pfaff [Mon, 20 Sep 2010 00:37:30 +0000 (17:37 -0700)]
lexer: New type enum token_type.
It seems useful to have a type that indicates that a value is a token.
Furthermore, being able to enumerate all of the possible tokens in a
straightforward way seems worthwhile. It also makes it possible for
GCC to check "switch" statements on token types and to build arrays
indexed by token.
Ben Pfaff [Fri, 19 Nov 2010 06:19:58 +0000 (22:19 -0800)]
command: Factor command name matching out of command.c.
Making command parsing into a library will allow other code to use
the same functionality, which will be useful later in the lexer.
It also simplifies parsing command names and allows us to easily
add tests for command name parsing.
The new command name parsing code supports UTF-8. This is not useful yet,
because its only client does not feed it UTF-8 strings, but it will be
useful later when the rest of the lexer is rewritten, where it will gain
another client that does feed it UTF-8.