2025-04-05  Timothy Rice  <trice@posteo.net>

	version 1.9
	* NEWS: Record release date.

2025-03-29  Timothy Rice  <trice@posteo.net>

	Make various changes suggested by make syntax-check
	- Add copyright notice to README
	- Ensure README refers to COPYING and INSTALL
	- #include "error.h" -> #include <error.h>
	- Use GNU link instead of postal address
	- Portability: Grep: use "> /dev/null" instead of "-q"
	- Break a long line

	gnulib: update to latest

2025-03-22  Timothy Rice  <trice@posteo.net>

	m4: Update gitignore for latest gnulib

	gnulib: update to latest

	cfg.mk: Set UPDATE_COPYRIGHT_HOLDER to myself
	This data is consumed by `make update-copyright`, so it should be set to
	whoever's copyright notice will be updated to the latest year.

	Add myself to a bunch of copyright notices

	Put NEWS in correct format for do-release-commit-and-tag

2025-03-22  Timothy Rice  <trice@posteo.net>

	Revert "rand: Add new program"
	This reverts commit 0ed0482e4fb05c99ed6fdf6d3c20864cbe61acfd.

	Adding rand(1) seemed desirable from the point of view of wanting a way
	to benchmark datamash(1). However, a comprehensive pseudorandom number
	generation engine would requires a non-trivial amount of statistical
	library support. There is currently little maintainer will for such
	effort.

	We can "double-revert" in future if there is interest, but for the
	moment it seems cleanest to not have rand(1) enter the upcoming stable
	release 1.9.

2025-01-28  Georg Sauthoff  <mail@gms.tf>

	improve const correctness

2025-01-28  Georg Sauthoff  <mail@gms.tf>

	datamash: decrease visibility of local symbols
	i.e. symbols that are only used inside their translation unit and don't
	have a prototype declaration in a header should be hidden with static.

	the main motiviation behind such a change basically is:

	1) minimize the likelihood of possible (future) symbol clashes between
	   different translation units or libraries
	2) speed up linking
	3) eliminate noise when using `-Wmissing-prototypes`, i.e. if your local
	   symbols are 'clean', then all remaining missing-prototypes warnings
	   point to missing includes or obsolete functions

	Note that a global symbol definition that shadows another doesn't necessarily
	yield a link error, in general.

	And even in the constellations where it does, for fixing, it's way more
	convenient to get a compile error, instead.

	Dima Kogan also added in the email discussion:

	Keeping your symbols localized is general good hygiene, and has even
	more upsides:

	- The robot reader of the code (the compiler) gains a higher level of
	  understanding: it gets knowledge of the FULL list of callers of such a
	  function, and is free to make deep optimizations, like inlining the
	  function

	- The human reader of the code also gains a higher level of
	  understanding. With local functions I also know about exactly what
	  paths are possible for calling them by simply doing a text search in
	  one file.

	Relatedly, the linker visibility settings are controllable also (gcc
	-fvisibility or attribute((visibility))), but that's important for
	shared libraries, and less so for executables.

2025-01-27  Georg Sauthoff  <mail@gms.tf>

	datamash: add missing include for validate_double_format
	Inconsistencies with prototypes yield a compile error

2025-01-25  Timothy Rice  <trice@posteo.net>

	AUTHORS: Add Georg Sauthoff

2025-01-25  Georg Sauthoff  <mail@gms.tf>

	datamash: Add explicit arguments to crosstab_print prototype
	In C23, a prototype like `crosstab_print ()` implies that the function
	will take no arguments. This conflicts with the definition
	`crosstab_print (const struct crosstab* ct)`.

	We can demonstrate the issue by running:

	$ CFLAGS+="-std=c2x" ./configure && make

	That resulted in the following error:

	> src/crosstab.c:152:1: error: conflicting types for 'crosstab_print'; have 'void(const struct crosstab *)'
	>   152 | crosstab_print (const struct crosstab* ct)
	>       | ^~~~~~~~~~~~~~
	> In file included from src/crosstab.c:34:
	> src/crosstab.h:50:1: note: previous declaration of 'crosstab_print' with type 'void(void)'
	>    50 | crosstab_print ();
	>       | ^~~~~~~~~~~~~~

2025-01-25  Timothy Rice  <trice@posteo.net>

	datamash: Address pure attribute error

2023-12-18  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: keep "getnum" operation inside fields
	Before, the "getnum" operation could scan into adjacent fields,
	because it used string operations on fields terminated by a
	field separator, not a NUL byte.

	* NEWS: Mention bug fix.
	* src/utils.c (extract_number): Copy field contents to
	  NUL-terminated string to ensure string operations stay
	  inside the field.
	* tests/datamash-tests-2.pl: Add new, and activate existing,
	  tests for the corrected "getnum" behavior.
	* tests/datamash-vnlog.pl: Add new tests for the corrected
	  "getnum" behavior.

2023-12-17  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: bug fix for decimal sep. as field sep.
	When the locale's decimal separator is used as field separator,
	reading a numeric value would continue over the field separator,
	resulting in an "invalid numeric value" error message.  This is
	now prevented by re-trying to read the numeric value after the
	field contents have been copied to a NUL-terminated buffer.

	* NEWS: Mention bug fix.
	* THANKS: Add bug reporters.
	* src/field-ops.c (field_op_collect): If strtold reads past the
	  field contents, copy field contents to temporary buffer and
	  try strtold again.
	* tests/datamash-i18n-de.pl: Activate two previously failing
	  tests, and add a test case from the recent bug report.
	* tests/datamash-tests-2.pl: Activate two previously failing
	  tests.

2023-12-11  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	maint: fix "make syntax-check" errors
	* cfg.mk: Except file doc/fdl.texi from line length check.
	* src/utils.c (update_best_seq): Add space before opening parenthesis.

2023-12-10  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tweak NEWS entry for datamash antimode bug fix
	A "set of numbers" implies that no number is repeated.  Both
	the mode and antimode operations look at repeated numbers.
	Thus use "sequence" instead of "set" in the description.

	* NEWS: Replace "set" with "sequence" in latest bug fix entry.

2023-12-10  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: fix "antimode" operation
	Both "mode" and "antimode" operations are implemented in the same
	function "mode_value()" in file "src/utils.c".  The algorithm
	traverses a list of sequences of numbers and attempts to remember
	the longest (for mode) or shortest (for antimode).

	Before, the sequence information was updated for every new number
	encountered, i.e., in general before the actual sequence length
	was known.  This approach works for mode, but not for antimode.

	Now, the sequence information is updated only after a sequence
	has ended.  There are two ways a sequence can end: (1) a different
	number is read, (2) there are no more numbers to read.  In both
	cases, this new sequence information (length and value) needs
	to be checked against the previous longest/shortest length and
	possibly updated.

	* NEWS: Mention bug fix.
	* src/utils.c (mode_value): Update best sequence information only
	  after end of sequence.  Move update code to new function, because
	  it is now used twice.
	  (update_best_seq): New function.
	* tests/datamash-tests-2.pl: Activate formerly broken now working
	  "antimode" tests.

2023-12-10  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: new "mode" and inactive "antimode" test
	* tests/datamash-tests-2.pl: Add new test for both "mode" and "antimode".
	  The "antimode" test currently fails and is commented out.

2023-12-09  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: four more "mode" and "antimode" tests
	* tests/datamash-tests-2.pl: Duplicate regression test for "mode"
	  to also test "antimode".  Add normal test with single input value
	  to both operations.

2023-12-08  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: more "mode" and "antimode" tests
	Kingsley G. Morse Jr. has reported a bug regarding the
	"antimode" operation in
	<https://lists.gnu.org/archive/html/bug-datamash/2023-12/msg00003.html>.

	Both the "mode" and "antimode" operations are implemented
	in the function "mode_value()" in the file "src/utils.c".

	Currently, there are only a few tests for each of the two
	operations.  Add more tests to have working examples.
	Also add commented out tests that trigger "antimode" bugs.

	* tests/datamash-tests-2.pl: Add 11 tests for "mode", 11
	  working tests for "antimode", and 5 commented out tests
	  for "antimode" that are currently broken.

2022-11-07  Shawn Wagner  <shawnw.mobile@gmail.com>

	Remove obsolete autoconf AC_PROG_CC_STDC macro
	* configure.ac: Remove AC_PROG_CC_STDC

	Remove deprecated gnulib fdl module
	* bootstrap.conf: Remove fdl from the module list
	* doc/fdl.texi: Explicitly include instead of relying on gnulib to
	                generate it.

	Remove deprecated gnulib non-recursive-gnulib-prefix-hack module
	* Makefile.am: Add AUTOMAKE_OPTIONS needed for the new approach
	* bootstrap.conf: Remove old module, enable suggested replacement for it.
	* m4/.gitignore: Update for removed module

	Update bootstrap notes for OpenBSD.
	* HACKING.md: Updates for OpenBSD 7.2, related typo fixes.

2022-11-07  Shawn Wagner  <shawnw.mobile@gmail.com>

	Add gnulib checks for sys/random.h and getrandom().
	Needed to compile rand on non-GNU OSes.

	 * Makefile.am: Add $(LIB_GETRANDOM) to rand libraries
	 * bootstrap.conf: Add getrandom and sys_random modules
	 * m4/.gitignore: Updates for the new modules

2022-09-03  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: more --vnlog tests (legend corner cases)
	The vnlog format is quite permissive regarding field names.
	Field names can comprise non-alphanumeric characters, and
	field names may be duplicated.  Add tests to check that GNU
	Datamash can handle (at least some) such vnlog input.

	The test with a duplicate field name is intended to help avoid
	accidental changes to GNU Datamash behavior, not to prescribe
	that the first matching field name should be used.

	* tests/datamash-vnlog.pl: Add two tests, the first with non-
	  alphanumeric characters, the second with a duplicate field
	  name.

2022-08-27  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: --vnlog "legend" parsing fixes
	* src/text-lines.c (line_record_fread): Skip prefix matching
	  regex '^\s*#\s*' instead of regex '^[\s#]+'.
	* tests/datamash-vnlog.pl: Additional vnlog legend tests.

2022-08-17  Shawn Wagner  <shawnw.mobile@gmail.com>

	Make -g and groupby take ranges of fields.
	    So that 1-4 works instead of needin

	  * src/op-parser.c: Accept ranges for -g and groupby
	  * tests/datamash-tests.pl, datamash-error-msgs.pl, datamash-crosstab.pl:
	    Some tests for the above.
	  * doc/datamash.text: Document the above.
	  * NEWS: Changelog entry.

2022-08-17  Timothy Rice  <trice@posteo.net>

	rand: Add new program

2022-08-14  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	docs: adjust damatash --vnlog texinfo description
	* doc/datamash.texi: Add that both input and output are affected.
	  Mark the link to the format description as a URL.

2022-08-13  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: comments and empty lines
	Without any options, GNU Datamash does not support comments,
	and does not treat empty lines as different from any other
	line.

	With -C/--skip-comments, it ignores complete lines starting
	with either '#' or ';' as first non-whitespace character.

	With --vnlog, it ignores empty lines, complete lines where the
	first non-whitespace character is a '#' (followed by either
	'!' or '#' in the vnlog prologue), and trailing comments,
	i.e., the part of a non-empty, non-comment, and non-header
	line started with optional whitespace and a '#' character.

	* tests/datamash-check.pl: Add tests with empty and comment
	  lines in the input with different options.
	* tests/datamash-vnlog.pl: Add trailing comment started with
	  whitespace to existing input.  Add test to verify that a
	  Semicolon (';') does not start a trailing comment.  Add a
	  test that the header line does not support comments.

2022-08-13  Timothy Rice  <trice@posteo.net>

	maint: Add vnlog to bash completion

	maint: Add Dima Kogan to AUTHORS/THANKS files

2022-08-13  Dima Kogan  <datamash@dima.secretsauce.net>

	Add vnlog support
	Dima Kogan's vnlog data format is explained at
	https://github.com/dkogan/vnlog

	This support is experimental in GNU Datamash.

2022-08-08  Timothy Rice  <trice@posteo.net>

	maint: Tidy up NEWS item for dotprod

	datamash: add new option -S/--seed to set random seed

2022-08-07  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: datamash -C has no inline comments
	* tests/datamash-tests-2.pl: Add two tests for errors when
	  trying to use inline comments.

	maint: wrap two lines to make syntax-check happy

2022-08-07  Timothy Rice  <trice@posteo.net>

	maint: Fix small typo 'data' -> 'date' in NEWS

	datamash: new operation: dotprod

2022-08-06  Timothy Rice  <trice@posteo.net>

	Add NEWS item for getrandom change

	Switch to getrandom for seed source

	Move init_random function into new randutils source

2022-08-06  Timothy Rice  <trice@posteo.net>

	Remove sc_indent from syntax checks
	The sc_indent check was added to gnulib in 8f043c6 on 2021-09-03. By default it
	forces code to conform to `indent -ppi 1`. This contradicts the list of format
	constraints suggested at [1], in particular the GNU preference of having
	two spaces for each indent level.

	[1] https://www.gnu.org/prep/standards/standards.html#Formatting

2022-08-06  Timothy Rice  <trice@posteo.net>

	datamash, decorate: add -h/-V for --help/--version

	maint: Fix typo 'syntaax'

2022-08-03  Timothy Rice  <trice@posteo.net>

	Update gnulib to latest
	Includes changing or casting a couple of size_t variables/outputs to idx_t.

2022-07-23  Timothy Rice  <trice@posteo.net>

	maint: Ignore vc-diffs

2022-07-23  Timothy Rice  <trice@posteo.net>

	maint: post-release administrivia
	Automatically done by `make release RELEASE='1.8 stable'`

	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* cfg.mk (old_NEWS_hash): Auto-update.

2022-07-23  Timothy Rice  <trice@posteo.net>

	version 1.8
	* NEWS: Record release date.

2022-07-17  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: tests: decimal point as field separator
	* tests/datamash-tests-2.pl: Add a commented out test for the
	  "getnum" operation with a decimal point as field separator.
	  Adjust alignment of two existing commented out tests.

2022-07-17  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: tests: commented out getnum i18n tests
	The "getnum" operation is neither properly locale-aware nor does
	it ignore the locale setting.  As a result floating point numbers
	are not extracted correctly unless the locale-specific decimal
	separator is a period.

	* tests/datamash-i18n-de.pl: Add commented out getnum tests.

2022-07-17  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: numbers with two decimal points in input
	* tests/datamash-error-msgs.pl: Add tests for error message when
	  a number with two decimal points is given as input data.

	tests: decimal separator as field separator
	* tests/datamash-i18n-de.pl: Add tests using the locale's decimal
	  separator (a comma) as field separator.
	* tests/datamash-tests-2.pl: Add tests using the locale's decimal
	  separator (a period) as field separator.

	tests: adjust commented out i18n test
	* tests/datamash-i18n-de.pl: Use a unique ID to allow commenting
	  in without further adjustments.  Align test definition fields.

	maint: regression test consolidation
	* tests/datamash-tests-2.pl: Move regression tests to have them
	  next to each other in one group.

2022-07-11  Timothy Rice  <trice@posteo.net>

	datamash: tests: Add test for crosstab with header-in, no header-out

	datamash: tests: Rename c3 header-in test for pcov consistent with scov

	datamash: tests: Add more paired-param tests with header in/out

	docs: Note non-standard --header-out with crosstab

	docs: Remove sentence about groupby/crosstab not understanding header names.

2022-07-09  Timothy Rice  <trice@posteo.net>

	datamash: maint: Fix some indentation

	datamash: bugfix: Ensure rmdup respects --output-delimiter
	Fixes bug reported by Dima Kogan.

	datamash: bugfix: Allow crosstab to be called by header field name.
	Ensure `--header-in crosstab x,y` does not crash.
	Fixes bug reported by Dima Kogan.

	datamash: maint: fix long line

	maint: ignore side effects of make syntax-check

	maint: Fix minor typo

	datamash: maint: Add debug helper macro WHEREAMI()

	tests: Add the test for pcov header in/out

	datamash: Print all operation columns in output header
	Ensure `--header-out pcov x:y` shows `pcov(x,y)` in header.
	Fixes bug reported by Dima Kogan.

2022-07-05  Timothy Rice  <trice@posteo.net>

	maint: Break long lines

	maint: Update version notices.

2022-07-04  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	maint: align datamash binning tests
	* tests/datamash-tests-2.pl: Align binning tests.

2022-07-04  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: more datamash binning tests
	Floating point numbers as operation parameters can be specified
	using scientific notation (e.g., 1.2e3 for 1200.0).

	* tests/datamash-tests-2.pl: Add scientific notation bin sizes.

2022-07-04  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: test parser corner cases
	Add tests of corner cases regarding whitespace in the operation
	parsing of GNU Datamash in order to avoid introducing unintended
	changes of behavior.

	* tests/datamash-parser.pl: Add tests with additional whitespace.

2022-07-03  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: add more datamash parser tests
	* tests/datamash-parser.pl: Additional testing of correct and
	  incorrect use of optional operation parameters.

2022-07-03  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	maint: more unique test identifiers for datamash
	Having unique test identifiers helps in locating the problems
	with test failures.  It seems as if the intention is to have
	unique test identifiers, at least per tested binary.  Fix a
	case of duplicated test identifiers in the two test files
	datamash-tests.pl and datamash-parser.pl for binning related
	test cases.

	* tests/datamash-parser.pl: Rename test identifier 'b1' to '31'
	  and 'b2' to 'b32'.

2022-06-26  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: add third datamash i18n test case
	Depending on tokenizer changes to support comma as decimal
	separator, three comma separated fields might trigger a
	problem that is avoided with two comma separated fields.

	* tests/datamash-i18n-de.pl: Add test case with three fields
	  separated with commas.

2022-06-26  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: avoid Perl warning in datamash-i18n-de.pl
	The string comparison `$lc_de eq undef` results in up to two
	warnings.  If `$lc_de` is defined, the single warning

	    Use of uninitialized value in string eq at
	     ./tests/datamash-i18n-de.pl line 39.

	is emitted.  This is caused by comparing with `undef` in the
	string equality test `eq`.

	If the locale `de_DE.utf8` is not found, `$lc_de` is undefined.
	If `$lc_de` is undefined, two warnings are emitted:

	    Use of uninitialized value $lc_de in string eq at
	     ./tests/datamash-i18n-de.pl line 39.

	    Use of uninitialized value in string eq at
	     ./tests/datamash-i18n-de.pl line 39.

	Using `defined()` to test if `$lc_de` is defined avoids this.

	* tests/datamash-i18n-de.pl: Use defined() to check if a
	  variable is defined.

2022-06-26  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: add second datamash i18n test case
	Supporting a comma as decimal separator for numeric arguments
	to GNU Datamash operations risks confusing a comma separated
	list of fields with a floating point number.

	* tests/datamash-i18n-de.pl: Add test case with a comma
	  separated list of field numbers.

2022-06-25  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	maint: fix a typo in a comment

2022-06-25  Timothy Rice  <trice@posteo.net>

	maint: Remove incorrect comment about LC_NUMERIC

	maint: Skip German test if de_DE.utf8 locale not found

	Test decimal separator in de_DE.UTF-8 locale

2022-06-25  Shawn Wagner  <shawnw.mobile@gmail.com>

	src/decorate.c: Fix a NetBSD-specific seg fault.

2022-06-24  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: re-write binning for negative numbers
	Make the binning code more explicit regarding handling of
	negative binning values:

	 - changing a negative zero into a positive zero can only
	   be needed for negative values;
	 - if the fractional part of a negative value is zero:
	   - the number is the lower bound of the bucket interval,
	   - and the number could be a negative zero;
	 - if the fractional part of a negative value is non-zero:
	   - the number falls into the preceding bucket interval.

	When testing with both negative and non-negative numbers,
	the new code was not slower.  It even seemed to be a tiny
	bit faster on average.

	* src/field-ops.c (field_op_collect): re-write binning code
	  for negative numbers.

2022-06-23  Timothy Rice  <trice@posteo.net>

	datamash: Fix binning of negative numbers.

2022-06-18  Shawn Wagner  <shawnw.mobile@gmail.com>

	Add framework for installing hooks into cloned git repositories.
	Includes a pre-commit hook that runs make syntax-check

2022-06-18  Timothy Rice  <trice@posteo.net>

	maint: Convert sort+header tests from shell to perl

	maint: fix long lines

	maint: Make test indentation more consistent

	maint: fix long lines

2022-06-17  Timothy Rice  <trice@posteo.net>

	Rename deprecated tests

	Deprecate -f/--full for non-linewise operations

2022-06-12  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: add cheap I/O error test
	The existing tests/datamash-io-errors.sh is marked as expensive
	and requires two file system images prepared to provoke I/O
	errors for GNU Datamash to detect and report.  As a result it
	is executed less frequently than most other tests.

	The new tests/datamash-io-errors-cheap.sh requires just the
	availability of the special file "/dev/full" to immediately
	provoke an I/O error on output.  This test requires minimal
	input data and is cheap to run.

	* Makefile.am (TESTS): Add new test file.
	* tests/datamash-io-errors-cheap.sh: New file with one test.

2022-06-07  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: fix segmentation fault
	As reported by Catalin Patulea on bug-datamash@gnu.org in
	<https://lists.gnu.org/archive/html/bug-datamash/2020-11/msg00001.html>,
	GNU Datamash could crash with a segmentation fault if the
	unique or countunique operations were used with input data
	containing NUL bytes.

	The problem was that the field_op_get_string_ptrs() function
	could create more pointers than it allocated memory for if
	the input data contained NUL bytes.  The solution is to add a
	check to avoid writing past the end of the "ptrs" buffer.

	* NEWS: Mention bug fix.
	* src/field-ops.c (field_op_get_string_ptrs): Do not write past
	  the end of the "ptrs" buffer.
	* tests/datamash-tests-2.pl: Add tests to verify bug fix.

2022-06-06  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: mention ms and rms in --help and man page
	The recent operations ms (mean square) and rms (root mean
	square) are listed only in the texinfo manual.  Add them
	to both the 'datamash --help' output and the datamash(1)
	man page.

	* src/datamash.c (usage): Add ms and rms to Statistical
	  Grouping operations.
	* man/datamash.x: Likewise.

2022-06-05  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: more leading and trailing whitespace tests
	* tests/datamash-tests.pl: Add tests with leading whitespace,
	  trailing whitespace, and both leading and trailing whitespace.

	tests: leading and trailing whitespace behavior
	* tests/datamash-tests.pl: Add whitespace-only tests.

	doc: texinfo manual adjustments
	* doc/datamash.texi: Mention that "cut" operation uses given field
	  ordering. Expand on leading and trailing whitespace description.
	  Ask for unified instead of context diff.

2022-06-04  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	maint: fix "make syntax-check" errors
	* .gitignore: Use "file system" instead of "filesystem".
	* tests/datamash-tests-2.pl: Wrap lines longer than 80 characters.

	maint: fix typo in a comment

2022-06-04  Timothy Rice  <trice@posteo.net>

	Remove commented-out code

	Fix typo

	datamash: Align field_operations columns
	It was difficult to visually group entries with the columns zig-zagging.

	datamash: Alias echo -> cut and unique -> uniq

2022-06-03  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	tests: enable valgrind test to pass
	The sub-test "custom-format" of the expensive test
	datamash-valgrind.sh would always fail, because the
	datamash was supposed to sum all input numbers, but
	some of those are too big for 80-bit extended floating
	point numbers as used on x86.  Thus datamash would
	emit an error message and return an exit code of 1.
	That was interpreted as a test failure.

	* tests/datamash-valgrind.sh: Use different valgrind error
	  exit code to distinguish between valgrind detecting a
	  memory leak and datamash reporting an error.

2022-06-02  Timothy Rice  <trice@posteo.net>

	maint: Impose '-T small' for fullfs check

	maint: Ignore side-effects make check-expensive

2022-06-01  Shawn Wagner  <shawnw.mobile@gmail.com>

	Merge branch 'master' of ssh://git.sv.gnu.org/srv/git/datamash

	Fix memory leaks in decorate

	Fix memory leaks with custom numeric precisions and formats

	Fix some memory leaks in crosstab code

	Fix a memory leak in parsing column names.

2022-06-01  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: fix a few typos
	* HACKING.md: Fix typos.
	* build-aux/create_corrupted_file_system.sh: Fix typos in comments.
	* build-aux/create_small_file_system.sh: Fix typo in a comment.

2022-05-31  Shawn  <shawnw.mobile@gmail.com>

	Add bootstrap instructions for NetBSD and OpenBSD to HACKING.md

2022-05-31  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: fix typo and quoting in man page template
	This is based on a Debian patch applied to GNU Datamash during
	packaging.  The Debian patch from Alejandro Garrido Mota
	<alejandro@debian.org> modifies the generated datamash man page.
	The Debian patch can be found at:
	https://sources.debian.org/src/datamash/1.7-2/debian/patches/fix-manpage.diff/

	* man/datamash.x: Spelling fix varient->variant.  Consistently
	  use ' instead of \' as single quote character.

2022-05-30  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: tweak help output and man pages
	Mention the LC_NUMERIC environment variable in datamash --help
	output and the generated man page.

	Add an "Options:" header line to the output of datamash --help
	and decorate --help so that help2man detects the OPTIONS section.

	Use DESCRIPTION instead of OVERVIEW to add text from the man page
	template to the DESCRIPTION section of the decorate(1) man page.

	Move the examples from the decorate(1) man page template into the
	EXAMPLES section of the generated man page.

	* man/decorate.x: Rename OVERVIEW to DESCRIPTION.  Move EXAMPLES.
	* src/datamash.c (usage): Add "Options:" header and "Environment:"
	  section to help output.
	* src/decorate.c (usage): Add "Options:" header to help output.

2022-05-29  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	datamash: fix short option -c X
	The short option -c X is documented as a short alternative to
	--collapse-delimiter=X, but it is not accepted.  Fix this.

	* src/datamash.c (short_options): Add "c:".
	* tests/datamash-tests.pl: Test use of -c.

2022-05-29  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: leading and trailing whitespace with -W
	With the -W, --whitespace option, datamash(1) ignores leading
	whitespace, but not trailing whitespace.  Trailing whitespace
	results in an additional empty field at the end of the line.
	This is different to field splitting in Awk, Bash, or Python.

	This behavior is tested for, but not documented.  Thus mention
	it in the texinfo documentation.

	* doc/datamash.texi: Add a senetence to the -W, --whitespace
	  option description.

2022-05-29  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: tweak NEWS entries for new features
	There are new features for both datamash(1) and decorate(1).
	Ensure that every new feature description mentions the program.
	Additionally, move the new decorate(1) feature entry below the
	new datamash(1) features to keep chronological order of entries.

	* NEWS: Always mention the changed program.  Move new entry
	  below existing entries.

2022-05-29  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	decorate: new conversions ipv6v4map and ipv6v4comp
	The two new decorate(1) conversion methods ipv6v4map and
	ipv6v4comp allow to sort on a field that may contain either
	an IPv6 or an IPv4 address.

	IPv4 and IPv6 addresses can be seen as IP addresses.  Logs from
	a "dual-stack" application, e.g., a web server, may contain
	either an IPv4 or IPv6 address at a given position in each line.
	Thus if one wants to sort the log file on the IP address,
	both IPv4 and IPv6 addresses need to be accepted as sort key
	and sorted consistently.  One approach is to transform one
	address type into the other before sorting.  IPv6 supports
	the transformation of IPv4 addresses into IPv6 addresses.

	There are two common methods for accomodating IPv4 addresses in
	IPv6: IPv4-Mapped addresses and the deprecated IPv4-Compatible
	addresses.  Both can be used to convert a given IPv4 address
	to an IPv6 address.  Both IPv4-Mapped and IPv4-Compatible IPv6
	address ranges are reserved by IANA and always represent IPv4
	addresses in a dual-stack enabled application.

	IPv4-Compatible addresses just add 96 leading zero bits to
	the 32 bit IPv4 address to create a 128 bit IPv6 address.
	This results in an ambiguity for the unspecified address
	(all-zero in both IPv4 and IPv6) and the IPv6 localhost address
	::1 with the first host address of "this" network in IPv4
	(0.0.0.1).  IPv4-Mapped addresses avoid this ambiguity.

	But since IPv4-Compatible IPv6 addresses can be seen as
	treating the IP address (both version 4 and version 6) as
	a specific way to represent an integer value I think it is
	useful to support this transformation as well.

	Both conversions logically convert an IPv4 address to an IPv6
	address, but the code actually creates a textual representation
	of an 128 bit integer from either an IPv4 or IPv6 address.

	Functionality like this was requested for sort(1) from GNU
	Coreutils:
	  https://lists.gnu.org/archive/html/coreutils/2011-06/msg00078.html
	  https://lists.gnu.org/r/bug-coreutils/2015-06/msg00039.html

	It was rejected for sort(1) from GNU Coreutils:
	  https://www.gnu.org/software/coreutils/rejected_requests.html#sort
	  https://lists.gnu.org/r/bug-coreutils/2015-06/msg00041.html

	Thus it seems appropriate for decorate(1), complementing the
	existing conversion methods for either IPv4 or IPv6 addresses,
	but not both.

	* NEWS: Mention new decorate conversion methods.
	* src/decorate-functions.c: Implement new conversion methods.
	  (decorate_ipv6_ipv4): New function.
	  (decorate_ipv6_ipv4_mapped): New function.
	  (decorate_ipv6_ipv4_compat): New function.
	  (builtin_conversions): Add new operations.
	* tests/decorate-errors.pl,
	  tests/decorate-sort-tests.pl,
	  tests/decorate-tests.pl: Add tests for new conversion methodss.

2022-05-29  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: adjust short description in decorate man page
	The decorate man page used the datamash short description.
	Replace it with a short description of decorate.

	* man/decorate.x: Adjust short description.

2022-05-29  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	maint: fix syntax-checks
	* AUTHORS: Break line at 80 characters.
	* THANKS: Break line at 80 characters.

	doc: mention -C, --skip-comments in texinfo manual
	* doc/datamash.texi: Mention -C, --skip-comments

	doc: fix man page name in help output
	* src/datamash.c: Use PROGRAM_NAME as man page name.
	* src/decorate.c: Likewise.

2022-05-29  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	doc: mention LC_NUMERIC in texinfo manual
	The LC_NUMERIC environment variable affects both the decimal and
	thousands separator used by GNU Datamash.

	* doc/datamash.texi: Mention LC_NUMERIC.

2022-05-28  Erik Auerswald  <auerswal@unix-ag.uni-kl.de>

	fix floor/ceil/up/down confusion in man page
	* man/datamash.x: floor is down, ceil is up.

2022-05-28  Timothy Rice  <trice@posteo.net>

	Change tabs to spaces for consistency

	Update bug report address to bug-datamash

	Promote Erik to the 'members' area in AUTHORS (i.e. git push access.)

	Clean up trailing end-of-line whitespaces.
	Partial revert of 9347234 which incorrectly deleted entire lines due to
	poor choice of sed command.

	Clean up trailing end-of-line whitespaces.

2022-05-27  Shawn Wagner  <shawnw.mobile@gmail.com>

	Squashed commit of the following:
	commit ee69c4d1c27e683d91163129ca1f5e43adf34243
	Author: Shawn Wagner <shawnw.mobile@gmail.com>
	Date:   Thu May 26 05:51:19 2022 -0700

	    Add tests for mean square and root mean square

	    Also update the R code in the file used to verify results to include
	    newer operations.

	commit 2036f7259a371420c8a90b8af31187f77a044290
	Author: Shawn <shawnw.mobile@gmail.com>
	Date:   Fri Jun 19 15:38:07 2020 -0700

	    Add mean square/root mean square commands.

	    doc/datamash.texi: Documentation
	    src/field-ops.c,src/op-defs.c,src/op-defs.h: New ms and rms commands.

2022-05-27  Timothy Rice  <trice@posteo.net>

	Expand AUTHORS and THANKS

2022-05-18  Shawn Wagner  <shawnw.mobile@gmail.com>

	Rebuild mr/.gitignore

2022-05-18  Shawn  <shawnw.mobile@gmail.com>

	Update bash completion module
	contrib/bash-completion/datamash: Add new commands and options

2022-05-18  Shawn Wagner  <shawnw.mobile@gmail.com>

	Add -c/--collapse-delimiter argument

	Tweaks to satisfy make syntax-check

2022-05-18  Shawn  <shawnw.mobile@gmail.com>

	Squashed commit of the following:
	commit 700e3ab38e19d5f53eb332ea11fa6001505fc026
	Author: Shawn Wagner <shawnw.mobile@gmail.com>
	Date:   Sat Jun 20 01:53:04 2020 -0700

	    Properly escape sort commands executed by sh

	    Also removes the arbitrary 1024 character limit on sort command.

	    * bootstrap.conf,Makefile.am: Add sh-quote gnulib module
	    * src/decorate.c: --print-sort-args quotes arguments when needed.
	    * src/datamash.c: Shell escape sort command used by popen(), get rid of
	      max length of it.
	    * tests/datamash-sort-errors.sh,tests/decorate-tests.sh: Handle new
	      output.

	commit ae84f5dddf6761d80cd08e441358ca23c96f13d6
	Author: Shawn Wagner <shawnw.mobile@gmail.com>
	Date:   Sat Jun 20 01:05:09 2020 -0700

	    Use an unstable sort with NetBSD /usr/bin/sort in decorate

	    It's stable by default, unlike any other sort(1) I checked, so
	    when using it, pass an option to make it unstable to pass tests cases.

	    * src/decorate.c: Pass -S to /usr/bin/sort on NetBSD
	    * tests/decorate-sort-tests.pl tests/decorate-tests.pl: Force using
	      /usr/bin/sort on NetBSD, and account for new output in test cases.

	commit 267c8974a5eed4844d9ee17e3f0aa9c45fe5d95f
	Author: Shawn Wagner <shawnw.mobile@gmail.com>
	Date:   Fri Jun 19 23:14:53 2020 -0700

	    Add a --sort-cmd argument to datamash and decorate.

	    Allows overriding the version of sort(1) found by configure.

	    src/datamash.c,src/decorate.c: Add --sort-cmd argument.
	    tests/datamash-sort-errors.sh: Use it in tests designed to fail.
	    doc/datamash.texi: Document it.

	commit 82fb2366e500533f91171465c511ddc3e600a53b
	Author: Shawn Wagner <shawnw.mobile@gmail.com>
	Date:   Fri Jun 19 23:00:39 2020 -0700

	    Make configure check for sort(1)

	    First looks for gsort, then sort, so that non-GNU userland OSes with
	    GNU coreutils installed will prefer GNU sort.

	    * configure.ac: test for gsort and sort
	    * src/datamash.c,src/decorate.c: Use the path to sort found by configure.
	    * src/datamash.c: Make it harder to overflow the sort command string.
	    * tests/decorate-tests.pl: Fix test cases for using full path to sort

2022-05-18  Shawn Wagner  <shawnw.mobile@gmail.com>

	Fix --help not including default value for --filler argument.
	Patch from Erik Auerswald.

2022-05-14  Shawn Wagner  <shawnw.mobile@gmail.com>

	Some auditing of gnulib modules
	Added some missing ones for things used by datamash, removed some
	unused ones.

	Squashed commit of the following:

	commit ce8fd789a0d60f0592ec79567f84c09b05e2b4aa
	Author: Shawn Wagner <shawnw.mobile@gmail.com>
	Date:   Sat May 14 05:43:05 2022 -0700

	    Update m4/ax_c_long_long.m4 from autoconf-archive latest

	commit d5cc572bc52d02169959361aca3f0e0823643d79
	Author: Shawn Wagner <shawnw.mobile@gmail.com>
	Date:   Sat May 14 05:34:48 2022 -0700

	    Update m4/.gitignore

	commit ef0ec5ca533067948800e3f8c10292b553cfb4eb
	Author: Shawn <shawnw.mobile@gmail.com>
	Date:   Thu Apr 23 15:57:02 2020 -0700

	    Add missing gnulib modules used in decorate

	commit 67cbe0b3462dab739a3bd81a89e1e91b2d128519
	Author: Shawn <shawnw.mobile@gmail.com>
	Date:   Mon Apr 20 01:49:03 2020 -0700

	    Add missing gnulib modules

2022-05-14  Shawn Wagner  <shawnw.mobile@gmail.com>

	Typo in manpage -h -> -H
	Patch by Erik Auerswald

	Fix spelling errors and other issues in texi documentation.
	Patch from Erik Auerswald.

	Fix an incorrect format in a decorate(1) message.
	Reported by Jiri Hladky, fixed by Erik Auerswald.

	Spelling fixes in documentation

2021-01-06  Assaf Gordon  <assafgordon@gmail.com>

	maint: update copyright year
	Run "make update-copyright" and then:
	* gnulib: Update to latest.
	* tests/init.sh, bootstrap: Update to latest from gnulib.
	* m4/.gitignore: Add gnulib updates.

2020-04-30  Shawn  <shawnw.mobile@gmail.com>

	datamash: fix missing -z when calling sort(1)
	When sorting NUL terminated lines, the needed -z option wasn't being
	passed to sort(1), causing errors due to improperly sorted data. Fixed,
	plus raises an error if the system sort(1) doesn't support -z.

	* NEWS: Mention bugfix.
	* configure.ac: Test to see if sort(1) understands -z
	* datamash.c (open_input): If "-z" and "-s" are used, add "-z" to the
	sort(1) command line; Exit with an error if sort(1) does not support "-z".

2020-04-30  Shawn  <shawnw.mobile@gmail.com>

	datamash,decorate: use pledge(2) on OpenBSD to restrict capabilities
	pledge is an OpenBSD specific mechanism for restricting what syscalls
	a process can execute, or limiting how they behave. It's considered
	polite on OpenBSD to limit a program to just the bare minimum it
	needs to run. See https://man.openbsd.org/pledge.2 for details.

	* configure.ac: Add check for pledge(2).
	* src/system.h (openbsd_pledge): New static function, calling pledge(2).
	* src/datamash.c (main), src/decorate.c (main): Call new function on startup.

2020-04-30  Shawn  <shawnw.mobile@gmail.com>

	tests: use jot as seq on OpenBSD instead of skipping tests
	* init.cfg(openbsd_seq_replacement_): New function.
	  Work around OpenBSD's lack of seq(1) using jot(1).
	* tests/datamash-rand.sh, tests/datamash-sort-errors.sh,
	  tests/datamash-strbin.sh: Use new function.

	tests: handle getopt_long(3) differences on OpenBSD
	* tests/datamash-error-msgs.pl, tests/decorate-errors.pl: Work around
	OpenBSD's getopt_long(3) producing different error messages from GNU
	version.

2020-04-24  Assaf Gordon  <assafgordon@gmail.com>

	maint: update NEWS with version 1.7 and date

2020-04-24  Assaf Gordon  <assafgordon@gmail.com>

	decorate: new program (experimental)
	decorate(1) works in tandem with sort(1) to enable sorting by
	non-standard ordering, e.g. IPv4, IPv6, roman numerals.

	Suggested by Pádraig Brady in:
	https://lists.gnu.org/r/bug-coreutils/2015-06/msg00076.html

	* NEWS: Mention new program.
	* bootstrap.conf: Add sys_socket, netinet_in gnulib modules.
	* .gitignore: Ignore decorate related files.
	* Makefile.am (bin_PROGRAMS): Add decorate.
	  (decorate_SOURCES/CFLAGS/CPPFLAGS/LDFLAGS): New items.
	  (man_MANS): Add decorate.
	  (TESTS): Add new tests.
	* src/decorate.c, src/key-compare.{c,h}: New program, adapted from
	  https://lists.gnu.org/archive/html/coreutils/2019-03/msg00056.html
	* src/decorate-functions.{c,h}: New module.
	* man/decorate.x: New file.
	* src/system.h: Add quotef,quotef_n,FALLTHROUGH,SORT_FAILURE.
	* tests/decorate-errors.pl, tests/decorate-sort-tests.pl,
	  tests/decorate-tests.pl: New tests.
	* po/POTFILES.in: Add new .c files.

2020-04-24  Assaf Gordon  <assafgordon@gmail.com>

	maint: fix syntax-checks following sha256,geomean additions
	* src/datamash.c (usage),
	  tests/datamash-sha.pl,
	  tests/datamash-stats.pl: Break line at 80 characters.
	* src/fields-ops.c: Fix space-before-parentheses.

2020-04-23  Assaf Gordon  <assafgordon@gmail.com>

	maint: update help2man

2020-04-23  Shawn  <shawnw.mobile@gmail.com>

	datamash: new operations: geomean,harmmean
	They compute the geometric and harmonic means, thus adding support
	for all three classic Pythagorean means.

	  $ seq 5 | datamash harmmean 1
	  2.1897810218978
	  $ seq 5 | datamash geomean 1
	  2.6051710846974

	Equivalent to the following R code:

	  > library("psych")
	  > x = seq(5)
	  > harmonic.mean(x)
	  [1] 2.189781
	  > geometric.mean(x)
	  [1] 2.605171

	* NEWS,  doc/datamash.texi,
	  man/datamash.x: Mention new operations
	* bootstrap.conf: Use gnulib's logl module.
	* Makefile.am (datamash_LDADD): Add more math libraries.
	* m4/.gitignore: Add logl files.
	* src/op-defs.{h,c}: Define new operations.
	* src/field-ops.c: Implement new operations
	* src/datamash.c (usage): List new operations.
	* tests/datamash-stats.pl: Add tests.

2020-04-10  Shawn  <shawnw.mobile@gmail.com>

	datamash: new line operations: sha224 and sha384
	Added for the sake of completeness; the implementations are already present.

	  $ printf "%s\n" 1 2 3 | datamash sha224 1
	  e25388fde8290dc286a6164fa2d97e551b53498dcbf7bc378eb1f178
	  58b2aaa0bfae7acc021b3260e941117b529b2e69de878fd7d45c61a9
	  4cfc3a1811fe40afa401b25ef7fa0379f1f7c1930a04f8755d678474

	  $ printf a | datamash sha384 1
	  54a59b9f22b0b80880d8427e548b7c23abd873486e1f035dce9cd697e85175033caa88e6d57bc35efae0b5afd3145f31

	Equivalent to:

	  $ printf 1 | sha224sum
	  e25388fde8290dc286a6164fa2d97e551b53498dcbf7bc378eb1f178  -
	  $ printf 2 | sha224sum
	  58b2aaa0bfae7acc021b3260e941117b529b2e69de878fd7d45c61a9  -
	  $ printf 3 | sha224sum
	  4cfc3a1811fe40afa401b25ef7fa0379f1f7c1930a04f8755d678474  -

	  $ printf a | sha384sum
	  54a59b9f22b0b80880d8427e548b7c23abd873486e1f035dce9cd697e85175033caa88e6d57bc35efae0b5afd3145f31

	* NEWS,
	  doc/datamash.texi,
	  man/datamash.x: Mention new operations.
	* src/datamash.c (usage): List new options.
	* src/op-defs.{h,c}: Define new operations.
	* src/field-ops.c: Implement operations.
	* tests/datamash-sha.pl: Add tests.

2020-02-24  Assaf Gordon  <assafgordon@gmail.com>

	doc: update NEWS with version 1.6 and date

	build: update bootstrap.conf with xstrtol-error
	* bootstrap.conf: add xstrtol-error (following gnulib update, without
	it "make dist" fails at "update-po" stage).

	gnulib: update to latest
	* m4/.gitignore: Auto-updated by 'bootstrap'.

	doc: mention getnum bugfix in NEWS
	* NEWS: Mention bugfix

2020-02-13  Assaf Gordon  <assafgordon@gmail.com>

	datamash: bug fix 'getnum' operation ignoring 8,9 chars
	Digits 8,9 were wrongly ignored on in default numeric format (decimal):

	   $ echo foo411586 | datamash getnum 1
	   4115   [ wrong result ]
	   411586 [ fixed result ]

	* src/utils.c (extract_number_types): Add missing digits to string
	pattern.
	* tests/datamash-tests-2.pl: Add tests to catch this error.

2020-01-02  Assaf Gordon  <assafgordon@gmail.com>

	maint: update copyright date to 2020
	* all files: Run "make update-copyright".

2019-09-17  Assaf Gordon  <assafgordon@gmail.com>

	doc: update NEWS with version 1.5 and date

2019-09-03  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operation: cut
	Similar to cut(1), it copies the input field to the output as-is.
	The advantage over cut(1) is that combined with datamash's other features,
	input fields can be specified by name instead of column number,
	and output fields can be re-ordered and duplicated.

	  Example:
	    $ printf "a b c\n1 X 6\n" | datamash -W -H cut c,a,c
	    cut(c)  cut(a)  cut(c)
	    6       1       6

	Suggested by Torsten Seemann <torsten.seemann@gmail.com> in
	https://lists.gnu.org/archive/html/bug-datamash/2019-09/msg00000.html

	* NEWS: Mention new operation.
	* src/datamash.c (usage): Add new op.
	* src/op-defs.{h,c}: Define new OP_CUT.
	* src/field-ops.{c}: Implement new OP_CUT.
	* contrib/bash-completion/datamash: Add new operation.
	* doc/datamash.texi, man/datamash.x: Mention new operation.
	* tests/datamash-tests-2.pl: Add tests.

2019-08-27  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operation: getnum
	Extract numeric values from a string:

	  $ echo foo-42.0-bar | datamash getnum 1
	  42.0

	Also accept single-letter option for different numeric types:

	  $ echo zoo-42.0-bar | datamash getnum:p 1  # positive decimal (default)
	  -42.0
	  $ echo zoo-42.0-bar | datamash getnum:n 1  # Natural numbers
	  42
	  $ echo zoo-42.0-bar | datamash getnum:i 1  # Integers
	  -42
	  $ echo zoo-42.0-bar | datamash getnum:d 1  # Decimal
	  -42.0
	  $ echo zoo-42.0-bar | datamash getnum:h 1  # Hex
	  66
	  $ echo zoo-42.0-bar | datamash getnum:o 1  # Oct
	  34

	* NEWS: Mention new operation.
	* src/datamash.c (usage): Add new op.
	* src/op-defs.{h,c}: Define new OP_GETNUM.
	*src/op-parser.c (set_op_params): Parse single-letter options.
	(parse_operation_params): Accept character parameters (not just numbers).
	* src/field-ops.{c}: Implement new OP_GETNUM.
	* src/utils.{h,c} (extract_number): New function.
	* contrib/bash-completion/datamash: Add new operation.
	* doc/datamash.texi, man/datamash.x: Mention new operation.
	* tests/datamash-tests-2.pl: Add tests.

2019-08-27  Assaf Gordon  <assafgordon@gmail.com>

	build: work-around automake 1.16 regression with non-gnu make
	Discussed in https://lists.gnu.org/r/automake/2019-08/msg00000.html
	Solved by Bruno Haible:
	https://lists.gnu.org/r/automake/2019-08/msg00005.html
	https://lists.gnu.org/r/bug-gnulib/2019-08/msg00064.html

	* Makefile.am (LDADD): Remove $(top_builddir) from library path.

2019-08-27  Assaf Gordon  <assafgordon@gmail.com>

	gnulib: update to latest (gettext 0.20 fix)

2019-08-23  Assaf Gordon  <assafgordon@gmail.com>

	maint: don't syntax-check for trailing spaces on a unit test file

	maint: appease space_before_parens syntax-check

	maint: ignore space_before_parens syntax-check on STREQ_LEN

2019-08-23  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operations: extname, barename
	Extract the file extension, and the basename without the extension:

	   $ echo foo.tar.gz | datamash barename 1 extname 1
	   foo     tar.gz

	   $ echo /home/foo/bar.txt \
	       | datamash dirname 1 basename 1 barename 1 extname 1
	   /home/foo    bar.txt    bar     txt

	* NEWS: Mention new operation.
	* src/datamash.c (usage): Add new op.
	* src/op-defs.{h,c}: Define new OP_BARENAME,OP_EXTNAME.
	* src/field-ops.{c}: Implement new OP_BARENAME,OP_EXTNAME.
	* src/utils.{h,c} (is_add_on_extension,+guess_file_extension): New
	helper functions.
	* contrib/bash-completion/datamash: Add new operation.
	* doc/datamash.texi, man/datamash.x: Mention new operation.
	* tests/datamash-tests-2.pl: Add tests.

2019-08-23  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operations: dirname, basename
	Similar to dirname(1) and basename(1), but operate on a file.

	   $ echo /home/foo/bar.txt | datamash dirname 1 basename 1
	   /home/foo    bar.txt

	* NEWS: Mention new operation.
	* src/datamash.c (usage): Add new op.
	* src/op-defs.{h,c}: Define new OP_DIRNAME,OP_BASENAME.
	* src/field-ops.{c}: Implement new OP_DIRNAME,OP_BASENAME.
	* contrib/bash-completion/datamash: Add new operation.
	* doc/datamash.texi, man/datamash.x: Mention new operation.
	* tests/datamash-tests-2.pl: Add tests.

2019-08-07  Assaf Gordon  <assafgordon@gmail.com>

	datamash: accept backslash-escaped characters in field names
	Suggested by Renato Alves <renato.alves@embl.de> in
	https://lists.gnu.org/archive/html/bug-datamash/2019-08/msg00000.html

	The following are now possible:
	      datamash -H sum FOO\\-BAR < input.txt
	      datamash -H sum 'FOO\-BAR' < input.txt
	      datamash -H sum "FOO\\-BAR" < input.txt

	* NEWS: Mention the feature.
	* doc.datamsah.texi, man/datamash.x: Document the feature.
	* src/op-scanner.h(MAX_IDENTIFIER_LENGTH): New limit.
	* src/op-scanner.c(scanner_get_token): Improve identifier scanning to
	allow backslashes.
	* tests/datamash-error-msgs.pl, tests/datamash-tests-2.pl: Add tests.

2019-08-07  Assaf Gordon  <assafgordon@gmail.com>

	datamash: fix mode/antimode incorrect results for negative values
	In version 1.4 and earlier, this resulted in incorrect values:

	  $ echo -1 | datamash mode 1
	  1.844674407371e+19

	Reported by Renan Valieris <rvalieris@gmail.com> in
	https://lists.gnu.org/archive/html/bug-datamash/2019-06/msg00001.html

	* src/utils.c (mode_value): Fix type from size_t to long double.
	* tests/datamash-tests-2.pl: Add test.
	* NEWS: Mention fix.

2019-01-02  Assaf Gordon  <assafgordon@gmail.com>

	maint: update gnulib to latest
	* gnulib: Update to latest.
	* bootstrap, tests/init.sh: Sync from gnulib.

	maint: update copyright date to 2019
	* all files: Run "make update-copyright".

2018-12-23  Assaf Gordon  <assafgordon@gmail.com>

	doc: update release date for version 1.4
	* NEWS: Update date.

2018-12-22  Assaf Gordon  <assafgordon@gmail.com>

	gnulib: update to latest

2018-12-15  Assaf Gordon  <assafgordon@gmail.com>

	maint: update .gitignore
	Following gnulib recent update.

	* m4/.gitignore: Update file list.

2018-12-15  Assaf Gordon  <assafgordon@gmail.com>

	build: update autoconf version requirements
	Require version 2.69 to bootstrap.
	After recent gnulib update, gnulib-tool would complain:

	  $ ./gnulib/gnulib-tool
	  ./gnulib/gnulib-tool: *** minimum supported autoconf version is 2.63. \
	      Try adding AC_PREREQ([2.63]) to your configure.ac.
	  ./gnulib/gnulib-tool: *** Stop.

	Hence this update.

	* configure.ac, bootstrap.conf: Require autoconf version 2.69 or later.

2018-12-14  Assaf Gordon  <assafgordon@gmail.com>

	build: add special CPPFLAGS for windows binaries
	FORCE_C_LOCALE: Force C/POSIX locale, instead of assuming it is
	available as an environment variable (not easily changeable in cmd.exe).

	SORT_WITHOUT_LOCALE: don't add "LC_ALL=C" to the sort command-line
	execution string - the windows interpreter (cmd.exe) can't handle that.

	* src/datamash.c (main): Add preprocessor conditionals.

2018-12-08  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new option -C/--skip-comments
	Suggested by Mark van Rossum in
	https://lists.gnu.org/r/bug-datamash/2018-11/msg00001.html
	https://lists.gnu.org/r/bug-datamash/2018-12/msg00003.html

	* NEWS: Mention new option.
	* src/datamash.c (usage, main): Update getopt parsing and help screen.
	* src/text-options.{c,h} (skip_comments): New varaible.
	* src/text-lines.{c,h} (line_record_fread): New option to skip comments.
	(line_record_is_comment): New function.
	* contrib/bash-completion/datamash: Update bash auto-completion options.
	* man/datamash.x: Add "skipping comment lines" section.
	* tests/datamash-tests-2.pl: Add new test cases.

2018-12-08  Assaf Gordon  <assafgordon@gmail.com>

	gnulib: update to latest

2018-08-22  Assaf Gordon  <assafgordon@gmail.com>

	doc: fix man-page typo
	Reported by Steve Ward in
	https://lists.gnu.org/r/bug-datamash/2018-08/msg00004.html .

	* man/datamash.x: Fix missing formatting letter B.

2018-03-23  Assaf Gordon  <assafgordon@gmail.com>

	tests: disable --format="%a" test
	Valid output can differ (e.g. 0x8.000p-3 vs 0x1.000p+0).

	* tests/datamash-output-format.pl: Disable 'a1' test.

2018-03-23  Assaf Gordon  <assafgordon@gmail.com>

	tests: fix --format='%4000f' expected output
	Can be 1.000009... or 1.000008999, depending on representation.

	* tests/datamash-output-format.pl: Check only the first 5 digits.

2018-03-17  Assaf Gordon  <assafgordon@gmail.com>

	doc: update NEWS with new version and date
	* NEWS: Set version 1.3 and date.

2018-01-25  Assaf Gordon  <assafgordon@gmail.com>

	maint: fix syntax-check issues
	* doc/datamash-texinfo.css: Limit to 80 characters.
	* po/POTFILES.in: Add missing files.
	* src/double-format.c: Remove unused "ignore-value.h".
	* src/text-options.{c,h}: Space after parens.

2018-01-23  Assaf Gordon  <assafgordon@gmail.com>

	maint: change URLs from http to https
	* all files: Change to https URLs.

	datamash: add --format=FORMAT option
	* NEWS: Mention this.
	* Maefile.am:
	* doc/datamash.texi: Mention new option.
	* src/datamash.c
	* src/double-format.{c,h}:
	* src/text-options.{c,h}:
	* tests/datamash-error-msgs.pl,
	  tests/datamash-output-format.pl: Test new option.
	* tests/datamash-valgrind.sh: Test large output buffer under valgrind.

2018-01-18  Assaf Gordon  <assafgordon@gmail.com>

	datamash: add -R/--round=N option
	Print numeric values with N decimal places.
	Example:
	  $ echo 1.1 | datamash --round=5 sum 1
	  1.10000

	Requested in https://github.com/agordon/datamash/issues/5 .

	* NEWS: Mention this.
	* src/datamash.c (usage): Mention new option.
	* configure.ac: Remove nonliteral-printf warning.
	* doc/datamash.texi: Mention new option.
	* src/text-options.c,src/text-options.h (numeric_output_format,
	  numeric_output_bufsize): New variables.
	  (set_numeric_output_precision,finalize_numeric_output_buffer): New
	  functions to set the printf style based on requested precision.
	* src/fields-ops.c: Print numeric output using new format variables.
	* tests/datamash-error-msgs.pl: Test invalid --round usage.
	* tests/datamash-output-format.pl: Test --round usage.
	* Makefile.am: Add new test file.

2018-01-04  Assaf Gordon  <assafgordon@gmail.com>

	maint: update copyright year to 2018
	* all files: Update with 'make update-copyright'.

	gnulib: update to latest

2017-12-11  Assaf Gordon  <assafgordon@gmail.com>

	maint: add github ISSUE/PULL-REQUEST templates
	Encourage contributors to use bug-datamash@gnu.org instead of github.

	* .github/ISSUE_TEMPLATE.txt,
	  .github/PULL_REQUEST_TEMPLATE.txt: New files.

2017-11-15  Assaf Gordon  <assafgordon@gmail.com>

	gnulib: update to latest
	* m4/.gitignore: Automatically updated git newest gnulib bootstrap.

2017-11-15  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operation 'trimmean'
	To calculate 20% trimmed mean:
	   $ printf "%s\n" 13 3 7 33 3 9 | datamash  trimmean:0.2  1
	   8

	Using 'trimmean:0' is equivalent to 'mean'.
	Using 'trimmean:0.5' is equivalent to 'median' (for R compatability).

	Suggested by Khavish Bhundoo in
	https://lists.gnu.org/archive/html/bug-datamash/2017-10/msg00004.html .

	* NEWS: Mention new operation.
	* src/datamash.c (usage): Add new op.
	  (print_column_headers): Print percent value in headers.
	* src/op-defs.{h,c}: Define new OP_TRIMMED_MEAN.
	* src/field-ops.{h,c}: Implement new OP_TRIMMED_MEAN.
	* src/op-parser.c: (set_op_params): Handle 'trimmean:N' paramater.
	* src/utils.{h,c} (trimmed_mean_value): New function.
	* contrib/bash-completion/datamash: Add new operation.
	* doc/datamash.texi, man/datamash.x: Mention new operation.
	* tests/datamash-error-msgs.pl: Test 'trimmean' parameter error messages.
	* tests/datamash-stats.pl: Add tests, compare against R results.

2017-09-07  Assaf Gordon  <assafgordon@gmail.com>

	build: indicate which hash functions are used at end of ./configure
	* NEWS: Mention the change (and the link fix from previous commit).
	* configure.ac: Print 'internal' or 'external' based on $LIB_CRYPTO.

2017-09-07  Jeroen Roovers  <jer@gentoo.org>

	build: link with external crypto libraries if needed
	Datamash uses gnulib's sha* modules, which allow linking with OpenSSL
	for increased performance (using ./configure --with-openssl=yes).
	Adjust the LDADD flags to link with the external libraries.

	Reported in
	http://lists.gnu.org/archive/html/bug-datamash/2017-08/msg00007.html .

	* Makafile.am (datamash_LDADD): Add $(LIB_CRYPTO)

2017-09-06  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new --output-delimiter=X option
	Suggested by Dave Myron in
	https://lists.gnu.org/archive/html/bug-datamash/2017-08/msg00008.html .

	* src/datamash.c (explicit_output_delimiter): New variable.
	(main): Handle new option.
	(usagE): Mention new option.
	* NEWS: Mention feature.
	* contrib/bash-completion/datamash: Add option.
	* doc/datamash.texi: Mention new option.
	* tests/datamash-tests-2.pl,
	  tests/datamash-error-msgs.pl: Test new option.

2017-09-06  Yu Fu  <yfu@yfu.me>

	doc: fixed two typos
	Reported in: https://github.com/agordon/datamash/pull/2 .

	* examples/readme.md: Fixed transcribes -> transcribed.

2017-08-23  Assaf Gordon  <assafgordon@gmail.com>

	version 1.2
	* NEWS: Record release date.

	maint: remove missing files from Makefile.am
	* Makefile.am: Remove files (which were deleted in previous commit).

	maint: remove outdated build-aux scripts
	* build-aux/check-remote-make-all.sh,
	  build-aux/check-remote-make-extra.sh,
	  build-aux/check-remote-make-git.sh,
	  build-aux/check-remote-make.sh: Precursors the http://pretest.nongnu.org,
	  no need for standalone scripts any more.
	* build-aux/gen-coverage-report.sh,
	  build-aux/rebuild-coverage.sh: obsoleted by 'make coverage'.
	* build-aux/tag-new-version.sh: Removed.

	maint: update prerelease checks script
	* build-aix/prerelease-checks.sh: Remove outdated checks, add
	non-root-installation check.

	maint: fix 'syntax-check' errors
	* src/crosstab.c: space-before-parens.
	* src/datamash.c: double-word.

	gnulib: update to latest

2017-08-21  Assaf Gordon  <assafgordon@gmail.com>

	tests: skip tests if paste(1) is not found
	On Alpine linux, paste(1) is not installed by default - skip instead
	of failing.

	* init.cfg (require_paste_): New shell function.
	* tests/datamash-rand.sh,
	  tests/datamash-sort-errors.sh: Skip if paste(1) is not found.

2017-08-21  Assaf Gordon  <assafgordon@gmail.com>

	tests: quote '^' in shell commands
	The '^' character has special meaning in older shells (e.g. OpenSolaris
	10/sparc). Quote it to prevent test failures.

2017-08-21  Assaf Gordon  <assafgordon@gmail.com>

	tests: accept '-nan' under OpenBSD
	OpenBSD systems return '-nan' instead of 'nan' for two tests.
	Ignore the '-' prefix.

	* tests/datamash-tests-2.pl (narm12,narm15): Discard '-' prefix.

2017-08-21  Assaf Gordon  <assafgordon@gmail.com>

	datamash: fix incorrect int size for percetiles
	On SunOS 5.11/i86pc, sizeof(size_t)==4 while sizeof(uintmax_t)==8.
	This leads to incorrect printf output.
	Reported by Dagobert Michelseni in
	https://lists.gnu.org/archive/html/bug-datamash/2017-08/msg00004.html .

	A better long term solution is to switch entirely from size_t to uintmax_t.

	* src/op-parser.c,
	  src/datamash.c: use PRIuMAX for printf, and convert to uintmax_t.

2017-08-10  Assaf Gordon  <assafgordon@gmail.com>

	maint: improve bash-completion script installation (again)
	Reported by Bruno Haible in
	https://lists.gnu.org/archive/html/bug-datamash/2017-08/msg00002.html:
	the current installation defaults to global location
	(e.g /usr/share/bash-completion.d) and does not work with non-root
	installation. Change default to local installation, with option to
	auto-detect the system's global location.
	Usage:
	 ./configure --with-bash-completion-dir=[no|local|global|PATH]
	See README for more details.

	* configure.ac: Change --with-bash-copmletion-dir processing.
	* NEWS: Mention this.
	* README: Document this.

2017-08-08  Assaf Gordon  <assafgordon@gmail.com>

	bash-completion: add range,perc operations
	* contrib/bash-completion/datamash: Add range/perc (percentile)
	operations to list of suggested completions.

2017-06-08  Assaf Gordon  <assafgordon@gmail.com>

	datamash: add lines/fields option to 'check' operation
	Datamash will fail with non-zero exit code if the input does not
	have the expected number of lines/fields.

	Typical usage:

	    $ seq 10 | paste - - | datamash check 2 fields && echo ok
	    5 lines, 2 fields
	    ok

	    $ seq 10 | paste - - | datamash check 6 fields && echo ok
	    line 1 (2 fields):
	      1     2
	    datamash: check failed: line 1 has 2 fields (expecting 6)

	    $ seq 10 | datamash check 11 lines
	    datamash: check failed: input had 10 lines (expecting 11)

	* NEWS: Mention new options.
	* src/datamash.c (tabular_check_file): Implement additional checks.
	* src/op-parser.h (struct mode_check_params_t): New struct to hold
	  options. (struct datamash_ops): Add new struct.
	* src/op-parser.c (parse_check_line_or_field, parse_mode_check): New
	  functions to parse 'check' options.
	  (parse_mode): Parse option in 'check' mode.
	* tests/datamash-parser.pl: Add parsing tests.
	* tests/datamash-check.pl: Add 'check' tests.
	* Makefile.am: Add 'datamash-check.pl' test script.
	* man/datamash.x: Add 'check' examples.
	* doc/datamash.texi: Expand 'check' section.

2017-06-07  Assaf Gordon  <assafgordon@gmail.com>

	maint: mark more lines as LCOV_EXCL_LINE for more accurate coverage

2017-05-26  Assaf Gordon  <assafgordon@gmail.com>

	gnulib: update to latest

2017-05-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: free crosstab memory if LINT
	Detected by Covrity scan (CID 72245, 72246).

	* src/crosstab.c (print_crosstab): Call free() if lint is enabled.

2017-05-26  Assaf Gordon  <assafgordon@gmail.com>

	build: add lint option to configure script
	Usage: ./configure --enable-lint

	Use during development to supress non-critical warnings.

	* configure.ac: Add --enable-lint option.
	* HACKING: Mention this option.

2017-05-09  Assaf Gordon  <assafgordon@gmail.com>

	build: add src/die.h to source files list
	Omitted from previous commit v1.1.1-10-gedab279
	"datamash: use die() instead of error()".

	* Makefile.am (datamash_SOURCES): Add missing file.

2017-05-09  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operation 'range'
	Calculate the range of values in the group:

	  $ printf '%s\n' 5 3 14 6 | datamash range 1
	  11

	Using 'range' is equivalent to:
	  $ printf '%s\n' 5 3 14 6 | datamash min 1 max 1 | awk '{print $2-$1}'
	  11

	Suggested by Kingsley G. Morse Jr. in
	https://lists.gnu.org/archive/html/bug-datamash/2017-05/msg00000.html .

	* NEWS: Mention new operation.
	* src/datamash.c: (usage): Add new op.
	* src/op-defs.{h,c}: Define new OP_RANGE.
	* src/field-ops.{h,c}: Implement new OP_RANGE.
	* tests/datamash-stats.pl: Test new operation.
	* tests/datamash-tests-2.pl: Test new operation with various inputs.
	* doc/datamash.texi, man/datamash.x: Mention new operation.

2017-05-09  Assaf Gordon  <assafgordon@gmail.com>

	build: add 'fallthrough' comments to supress gcc-7.1 warnings
	gcc-7.1 warns about:
	   src/op-parser.c:491:10: error: this statement may fall through
	      [-Werror=implicit-fallthrough=]
	   if (scan_val_int>0)
	       ^

	The fallthrough is intentional - add an explicit comment that will
	detected by the compiler and supress the warnings.

	* src/op-parser.c: Add explicit 'fallthrough' comment in two intentional
	fall through cases.

2017-05-09  Assaf Gordon  <assafgordon@gmail.com>

	datamash: use die() instead of error()
	Copied from GNU coreutils v8.25-79-g492dcb2,
	http://git.savannah.gnu.org/cgit/coreutils.git/commit/?id=492dcb2eb1:
	  "die() has the advantage of being apparent to the compiler
	   that it doesn't return, which will avoid warnings in some cases,
	   and possibly generate better code."

	* src/die.h: Copied from GNU coreutils.
	* src/op-parser.c,
	  src/op-scanner.c,
	  src/field-ops.c,
	  src/datamash.c: Changed 'error(EXIT_FAILURE,...)' to 'die(...)'.

2017-05-09  Assaf Gordon  <assafgordon@gmail.com>

	build: require pkg-config>=0.28 when building from git
	configure.ac uses the PKG_CHECK_VALUE macro from the pkg-config' pkg.m4
	file for bash-completion directory detection. This macro is not
	available in older pkg-config versions.
	Does is only required when building from git and running ./bootstrap.
	Building from tarball does not require pkg-config at all.

	* bootstrap.conf: Specify minimum version for pkg-config.

2017-05-03  Assaf Gordon  <assafgordon@gmail.com>

	build: remove bashcompdir from Makefile.am
	It is already defined in configure.ac (warning generated by autoreconf).

	* Makefile.am: Remove bashcompdir.

2017-04-19  Assaf Gordon  <assafgordon@gmail.com>

	doc: expand build instructions in README and HACKING.md
	Explain how to build from git.

2017-04-19  Assaf Gordon  <assafgordon@gmail.com>

	build: require 'pkg-config' when bootstrapping
	pkg-config is used to detect the bash-completion directory.
	It is optional during ./configure (and ./configure will succeed even if
	pkg-config is not installed), but pkg.m4 is required by ./bootstrap
	for the PKG_CHECK_VAR in 'configure.ac'.

	* bootstrap.conf: Add pkg-config requirement.

2017-04-06  Assaf Gordon  <assafgordon@gmail.com>

	gnulib: update to latest

2017-03-24  Barry Nisly  <barry.nisly@gmail.com>

	datamash: new operation 'perc' (percentiles)
	Usage:
	    $ seq 100 | datamash --header-out perc:95 1
	    perc:95(field-1)
	    95.05

	See https://lists.gnu.org/archive/html/bug-datamash/2017-03/msg00000.html .

	* NEWS: Mention new operation.
	* src/datamash.c: (usage): Add new op.
	  (print_column_headers): Print percentile value in headers.
	* src/op-defs.{h,c}: Define new OP_PERCENTILE.
	* src/field-ops.{h,c}: Implement new OP_PERCENTILE.
	* src/op-parser.c: (set_op_params): Handle 'perc:N' paramater.
	* tests/datamash-stats.pl: Test new operation.
	* tests/datamash-tests.pl: Test 'perc' headers.
	* doc/datamash.texi, man/datamash.x: Mention new operation.
	* tests/datamash-error-msgs.pl: Test 'perc' parameter error messages.

2017-03-17  Assaf Gordon  <assafgordon@gmail.com>

	build: install bash-completion script by default (if possible)
	If the 'bash-completion' package is instaleld with pkg-config support,
	detect the default bash-completion directory and install the datamash
	completion script there (typically: /usr/share/bash-completion/completions).
	Otherwise, install in /usr/local/share/datamash/bash-completion.d (as
	before).

	Disable with `./configure ---with-bash-completion-dir=no`.
	Set custom path with `./configure ---with-bash-completion-dir=PATH`.

	* README: Mention bash-completion options to './configure'.
	* configure.ac: Add --with-bash-completion-dir=DIR option, and
	  autodetect bash-completion directory using pkg-config.
	* Makefile.am: Add bash-completion target (if enabled).

2017-03-17  Assaf Gordon  <assafgordon@gmail.com>

	maint: update .gitignore

	gnulib: update to latest

2017-01-20  Assaf Gordon  <assafgordon@gmail.com>

	maint: update bootstrap script to latest
	* bootstrap: Copy latest version from gnulib/build-aux/bootstrap.

	maint: update NEWS

	gnulib: update to latest

2017-01-11  Assaf Gordon  <assafgordon@gmail.com>

	build: fix out-of-tree builds without dep-tracking
	Create src,lib,doc,tests subdirectories at the end of
	'configure', preventing build problems when building out of tree
	with --disable-dependency-tracking.
	Inspired by sed's https://bugs.gnu.org/25371 .

	* configure.ac: Call AS_MKDIR_P() to create subdirectories.

2017-01-11  Assaf Gordon  <assafgordon@gmail.com>

	maint: update copyright year to 2017
	Automatic update with 'make update-copyright'.

	maint: add gnulib's 'update-copyright' module
	* bootstrap.conf: Add 'update-copyright'.
	* cfg.mk: Define 'update-copyright-env' with relevant values for
	  'make update-copyright'

	maint: update gnulib to latest
	* m4/.gitignore: Ignore few more m4 files.

2016-09-23  Assaf Gordon  <assafgordon@gmail.com>

	doc: fix man page escapes
	Suggested by Dave Love in
	http://lists.gnu.org/archive/html/bug-datamash/2016-02/msg00002.html

2016-09-23  Assaf Gordon  <assafgordon@gmail.com>

	datamash: bugfix: trailing delimiter should count as an empty field
	In version 1.1.0 and earlier, trailing delimiter at the end of the line
	did not count as an empty field:

	    $  printf "a:b:\n" | datamash -t: check
	    1 line, 2 fields

	After bugfix:

	    $  printf "a:b:\n" | datamash -t: check
	    1 line, 3 fields

	Related to bug report by Sanjeev Kumar Sharma in
	http://lists.gnu.org/archive/html/bug-datamash/2016-09/msg00000.html

2016-09-23  Assaf Gordon  <assafgordon@gmail.com>

	datamash: 'transpose' bugfix (missing values in last line)
	Missing values in the last line would cause fields to be silently dropped:

	    $ printf "1\t2\na\n" | datamash --no-strict transpose
	    1       a

	After bugfix:

	    $ printf "1\t2\na\n" | datamash --no-strict transpose
	    1       a
	    2       N/A

2016-01-17  Assaf Gordon  <assafgordon@gmail.com>

	maint: update date in NEWS
	* NEWS: update release version, metion version bump.

	doc: add 'field delimiters' section to manual
	* doc/datamash.texi: add new 'field delimiters' section.

2016-01-03  Assaf Gordon  <assafgordon@gmail.com>

	maint: update gnulib to latest version
	* gnulib: update to latest version with copyright year adjusted
	* tests/init.sh,
	  bootstrap: sync with latest gnulib.

	maint: update all copyright year number ranges

2016-01-01  Assaf Gordon  <assafgordon@gmail.com>

	maint: rename pmccabe makefile target name
	* Makefile.am: rename 'CYCLO_SOURCES' to 'cyclo_sources', to avoid
	  confusing autoreconf about missing library 'CYCLO'.

	doc: add rounding,binning sections to manual

	doc: add 'crosstab' section to manual

	doc: add 'check' usage example

	doc: add 'groupby on /etc/passwd' usage example section

	doc: merge reverse+transpose nodes

	doc: added 'Usage Examples' chapter

	doc: reword 'overview' section

	doc: add custom CSS to the texinfo HTML output
	* doc/datamash-texinfo.css: new custom styles
	* doc/local.mk: use new CSS file in 'make html'
	* Makefile.am: use new CSS file in 'make web-manual'

	doc: update manual with new operations

2015-12-26  Benno Schulenberg  <bensberg@justemail.net>

	datamash: improve an error message

	datamash: gettextize an option description as a whole

	datamash: remove redundant header, and refer correctly to the man page

	datamash: improve the punctuation and grammar of the usage text

2015-12-20  Assaf Gordon  <assafgordon@gmail.com>

	doc: update NEWS for upcoming version

	maint: update gnulib submodule to latest version

2015-12-18  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operation: strbin (binning strings to numeric value)
	hashes a text string into a number, returns values from 0 to #bins
	(default 10). Can be used to bin many free-text input values into a
	finite (small) set of bins. Identical strings will be hashed to same
	numeric values.

	Examples:
	assign a bin (0 to 9) to a list of strings (bin number added
	as last field):

	    seq 100 | xargs printf "user-%d\n" \
	       | datamash --full strbin 1

	identical input values will be assigned identical bin number:

	    ( seq 100 ; seq 20 ) | xargs printf "user-%d\n" \
	       | datamash --full strbin 1

	Splitting input values into 10 output files, while keeping
	identical keys in the same file:

	    ( seq 100 ; seq 20 ) | xargs printf "user-%d\n" \
	       | datamash --full strbin 1 \
	       | awk '{print $0 > $NF ".txt"}'

	* bootstrap.conf: add gnulib's hash-pjw-bare module
	* src/op-defs.{c,h}: add new op constant
	* src/field-ops.{c,h}: implement new operation
	* src/op-parser.c: strbin can take a parameter (# bins)
	* src/datamash.c: update usage()
	* man/datamash.x: update man page
	* contrib/bash-completion/datamash: add new op
	* tests/datamash-error-msgs.pl,
	  tests/datanash-strbin: test new operation
	* Makefile.am: add new tests

2015-12-18  Assaf Gordon  <assafgordon@gmail.com>

	maint: use gnulib's pmccabe2html module
	Generate report with:
	    make cyclo-datamash.html

	* bootstrap.conf: add the pmccabe2html module
	* Makefile.am: add new target
	* .gitignore: ignore generated html file

2015-11-13  Assaf Gordon  <assafgordon@gmail.com>

	maint: remove superfluous 'with' from doc's "invariant section"
	pointed by Debian/Ubuntu package maintainer (alejandro@debian.org).

	* doc/datamash.texi: remove 'with' from 'invariant section' of GFDL
	  to comply with official wording.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	build: prefer %PRIuMAX to %zu in printf
	%zu doesn't play well with mingw cross-compilers.

	* src/column-headers.c, src/op-parser.c: include <inttypes.h>,
	   use %PRIuMAX instead of '%zu' in printf .

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	build: add gnulib's stpcpy module (for mingw)
	* bootstrap.conf: add stpcpy.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	datamash: fail on invalid cov/pearson input
	Unequal number of items can occur if ignoring N/A values.

	* src/field-ops.c: fail if the two fields have different number of
	    items.
	* tests/datamash-pair-tests.pl: tests such inputs.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	datamash: fail on incorrect usage
	* src/op-parser.c: fail when primary operations are given fields
	    ranges or pairs (e.g. 'groupby 1:2'). fail when crosstab is
	    given more than one operation.
	* tests/datamash-{crosstab,error-msgs,parser}.pl: add new tests.

	doc: improve man page (new ops, examples)
	* man/datamash.x: mention new operations,
	   use 'field' instead of 'column',
	   add more examples, improve formatting.

	datamash: improve usage()
	* src/datamash.c: improve usage(), mention new operations,
	    use 'field' instead of 'column'.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	new operations: round,floor,ceil,trunc,frac
	typical usage:

	    $ (echo X ; seq -1 0.5 1 ) \
	        | datamash -H --full round 1 floor 1 ceil 1 trunc 1 frac 1
	        X  round(X)  floor(X)  ceil(X)  trunc(X)  frac(X)
	     -1.0  -1        -1        -1       -1         0
	     -0.5  -1        -1         0        0        -0.5
	      0.0   0         0         0        0         0
	      0.5   1         0         1        0         0.5
	      1.0   1         1         1        1         0

	* bootstrap.conf: add gnulib's floorl,ceill,roundl,modfl modules.
	* src/op-defs.{c,h}: add new operations constants.
	* src/field-ops.c: handle new operations.
	* tests/datamash-tests-2.pl: add tests.
	* contrib/bash-completion/datamash: add new operations.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	utils: add functions for long-double zero handling
	* src/utils.c: is_zero(), is_signed_zero() - retruns true/false.
	   pos_zero(): negates a negative zero, nop otherwise.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	new operation: bin (binning into buckets)
	typical usage (bin into buckets of size 5):

	    $ seq 0 10 | ./datamash bin:5 1
	    0
	    0
	    0
	    0
	    0
	    5
	    5
	    5
	    5
	    5
	    10

	* src/op-defs.{c,h}: add new operation constant and name.
	* src/op-parser.c:   special handling for operation with an optional
	                     parameter.
	* src/field-ops.h:   add optional parameter to struct fieldop.
	* src/field-ops.c:   implement binning operation.
	* tests/*.pl:        add tests

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	datamash: reword error messages
	* src/op-parser.c: use 'field' instead 'column', standarize error
	    messages.
	* tests/*.pl: adapt tests to new wording.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	build: fix long-double precision under NetBSD
	Under NetBSD (and others?) default precision is set to 'double'
	even for 'long double' variables. This results in incorrect
	results for 'roundl' and crahes for 'expl' (when using
	gnulib's implementations.
	Force 'long double' precision. See gnulib's 'lib/fpucw.h' file
	for many more details.

	* bootstrap.conf: add fpucw module.
	* src/datamash.c: force 'long double' precision.

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	tests: add rmdup expensive tests
	* src/datamash.c: add undocumented option to reduce initial size of
	   rmdup structures (forcing more frequent reallocations).
	* tests/datamash-valgrind.sh: use new option, validate results.

	tests: add pcov with header
	* tests/datamash-pair-tests.pl: tests pcov (taking two fields) with
	   output headers (print 1 field, 1 header).

2015-07-14  Assaf Gordon  <assafgordon@gmail.com>

	refactor: create new scanner module
	Use proper tokenizing instead of relying on argc/argv 'tokens' with
	ad-hoc string manipulation. In the future, this will allow accepting the
	script as on string (like sed's -e).

	* bootstrap.conf: add gnulib's c-type module
	* src/op-scanner.{c,h}: new scanner module, improved tokenizing.
	* src/op-parser.c: use new scanner, support _STANDALONE_ mode for
	    quicker testing (without autotools,gnulib).
	* src/op-defs.c: prefer name 'groupby' to 'grouping' when reporting errors.
	* src/system.h: support_STANDALONE_ mode for
	    quicker testing (without autotools,gnulib).
	* tests/datamash-{tests,parser,crosstab}.pl: adapt error message text
	    to new parser (no change in functionality).
	* Makefile.am: include new scanner.
	* po/POTFILES.in: include new scanner.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	build: move 'assert.h' to 'system.h'
	* src/op-defs.c: move 'assert.h' from here...
	* src/system.h: to here.
	  avoids 'syntax-check' warings about including 'assert.h' without using
	  it (it is used indirectly with 'internal_error()' macro).

	build: conditionally include minmax.h
	Don't include "lib/minmax.h" if MAX is already defined.
	Seems redundant (per the comment in minmax.h), but tcc 0.9.26 still fails.

	maint: force -std=c99 with static analysis
	* cfg.mk: force -std=c99 when using clang's scan-build program, as it
	   does not to do so automatically, and compliation fails.

	maint: improve configure messages
	* configure.ac: simplify printing of examples directory;
	  print bash-completion directory.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: improve,install bash-completion script
	By default, script installed to program's datadir, e.g.
	  /usr/local/datamash/share/bash-completion.d/datamash
	and the user can manually source it from ~/.bashrc.

	Optionally, can be installed to another location, e.g.
	  ./configure --with-bash-completion-dir=/etc/bash_completion.d
	to be automatically loaded.

	* contrib/bash-completion/datamash: fix script, improve suggestions,
	  do not use '_init_completion' (not available in bash-comp<1.9)
	* configure.ac: add option --with-bash-completion-dir
	* Makefile.am: install script.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operation: Pearson Correlation
	usage (calculate pearson (population) correlation coefficient on fields 1,2):
	  datamash ppearson 1:2 < 1.txt

	* src/op-defs.{c,h}: new op enums.
	* src/op-parser.c:   handle new op.
	* src/utils.{c,h}:   implement pearson correlation function.
	* src/field-ops.c:   handle new op.
	* tests/datamash-pair-tests.pl: add tests

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operation: covariance
	usage (calculate population-covariance on fields 1,2):
	  datamash pcov 1:2 < 1.txt

	* src/op-defs.{c,h}: new op enums.
	* src/op-parser.c: handle new operation, with special parsing for column
	   pairs (e.g. '1:2').
	* src/field-ops.{c,h}: implement new operation, and add master/slave
	   field-op options (as covariance is implemented as two linked field
	   ops, one for each column).
	* src/datamash.c: skip 'slave' field-ops when printing.
	* tests/datamash-parser.pl: test column-pairs parsing
	* tests/datamash-pair-tests.pl: test column-pair operations.
	* tests/datamash-valgrind.sh: test covariance operation.
	* Makefile.am: include new tests.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: utils: add covariance implementation
	* src/utils.{c,h}: covariance_value() - new function.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: improve parser, allow field range
	allows: datamash sum 1-5,7-9

	* src/op-parser.c: improve field parsing
	* tests/datamash-parser.pl: add new tests

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new feature: check
	reports back number of lines,fields - or exit with failure if the input
	lines do have have same number of fields.

	usage:
	    $ seq 10 | paste - - | datamash check
	    5 lines, 2 fields

	    $ seq 9 | paste - - | datamash check
	    line 4 (2 fields):
	      7 8
	    line 5 (1 fields):
	      9
	    ./datamash: check failed: line 5 has 1 fields (previous line had 2)

	* src/datamash.c: tabular_check_file() - new function.
	* src/op-defs.{h,c}: new mode enum.
	* src/op-parser.c: handle new mode.
	* tests/datamash-check-tabular.pl: add tests.
	* Makefile.am: add new test file.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	maint: git-ignore more files

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new feature: crosstab
	usage:
	    datamash crosstab 1,2 sum 3

	* src/crosstab.{c,h} - implementation
	* src/datamash.c     - handle new crosstab processing mode.
	* src/op-defs.{c,h}  - additional crossab constants.
	* src/op-parser.c    - special handling for crosstab mode.
	* tests/datamash-crosstab.pl - new tests.
	* Makefile.am        - add new files.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	maint: add GL_ATTRIBUTE_PURE where needed

	maint: fix NEWS syntax-check
	* NEWS: add missing empty line.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	maint: add coverage hints
	1. Add LCOV_EXCL_BR_LINE to skip few switch cases where the default case
	   should never be reached.
	2. Add LCOV_EXCL_LINE.

	LCOV_EXCL_BR_LINE requires geninfo from lcov>=1.11 .

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	maint: remove unused strcnt module

	datamash: improve parser module, add tests
	* src/datamash.c: remove redundant left-over code.
	* src/op-parser.c: improve token processing
	* src/op-defs.c: rename 'grp' alias to 'gb' (=groupby)
	* tests/datamash-parser.pl: new tests
	* Makefile.am: add new tests.

	maint: fix syntax-check errors

	datamash: move config vars to text-options
	* src/datamash.c,src/field-ops.c: move several configuration
	   parameters from here ...
	* src/text-options.{c,h}: ... to here.

	maint: fix empty variable init
	* src/field-ops.c: CLANG complained about '{0}'. replace with memset().

	refactor: move field-operation list.
	* src/field-ops.{c,h}: refactored to handle just one struct at a time.
	   private functions removed from the header file.
	* src/op-parser.{c,h}: the list of field-ops is handled here.
	* src/datamash.c: adapted as needed.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	refactor: create parsing module
	1. Better handling of processing modes (e.g. group-by/transpose/reverse/rmdup)
	   vs operations (e.g sum/count/min/max/mean).
	2. Support new syntax:
	     datamash groupby 1,2 sum 3
	   equivalent to:
	     datamash -g 1,2 sum 3

	* src/op-defs.{c,h} - list of processing modes and operations.
	* src/op-parser.{c,h} - parsing module. returns structure with parsed
	   information (mode, grouping fields, operations, etc.).
	* src/op-fields.{c,h} - adapt to new parsing, remove old pasring functions.
	* src/datamash.c - adapt to new parsing module.
	* tests/*.pl - minor changes to error messages.

2015-06-26  Assaf Gordon  <assafgordon@gmail.com>

	refactor: extract group-processing to a function
	* src/datamash.c:
	  process_group(): new function to print/summarize each group.

	refactor: fields-ops printing
	* src/field.ops{h,c}:
	  field_op_summarize_empty(), field_op_summarize():
	    store results in 'op->out_buf', don't directly print to stdout.
	  summarize_field_ops(), field_op_print_empty_value():
	    print buffer to stdout.

	refactor: print empty values (for testing)
	* src/datamash.c: move empty value printing code from here ...
	* src/field-ops.{h,c}: ... to new function field_op_print_empty_value().

2015-06-22  Assaf Gordon  <assafgordon@gmail.com>

	maint: add NEWS about 1.0.7 features

2015-05-29  Assaf Gordon  <assafgordon@gmail.com>

	build: update gettext/autoconf/automake versions
	* bootstrap.conf: require gettext>=0.19.4, autoconf=>2.62,
	automake>=1.11.1.
	* configure.ac: require gettext=>=0.19.4
	* po/ChangeLog: updated to mention gettext 0.19.4 by 'gettextize -f'

2015-05-28  Assaf Gordon  <assafgordon@gmail.com>

	maint: update gnulib to latest version.
	specifically, for mingw fixes:
	http://lists.gnu.org/archive/html/bug-gnulib/2015-05/msg00026.html
	http://lists.gnu.org/archive/html/bug-gnulib/2015-05/msg00030.html
	http://lists.gnu.org/archive/html/bug-gnulib/2015-05/msg00031.html

2015-05-27  Assaf Gordon  <assafgordon@gmail.com>

	maint: improve printf portability, remove mingw hack
	* bootstrap.conf: use gnulib's inttypes,stdint,extensions modules.
	See http://lists.gnu.org/archive/html/bug-gnulib/2015-05/msg00021.html .
	* configure.ac: remove mingw hack. gnulib's extensions module should
	take care of making PRIuMAX work correctly.
	* src/datamash.c: use PRIuMAX as format speficier in printf, instead of
	previous ugly mingw hack (which also didn't work with translations, see
	http://lists.gnu.org/archive/html/bug-datamash/2015-05/msg00001.html ).

	maint: avoid compiler warning about no-return.
	* src/fields-ops.c: add 'return' which should never be reached.

2015-05-20  Assaf Gordon  <assafgordon@gmail.com>

	tests: improve portability of perl Digest::MD5
	On some systems (Centos7, Fedora21), Perl's Digest::MD5 is not installed
	by default (despite being official a core module).
	If so, skip the MD5 tests instead of failing.

	* tests/datamash-md5.pl: separate tests file for MD5.
	* tests/datamash-tests{,-2}.pl: don't use Digest::MD5 by default.
	* Makefile.am: add the new test file

2015-05-19  Assaf Gordon  <assafgordon@gmail.com>

	maint: improve tests portability of inf/nan
	Depending on the OS/libc, printf(Not-a-number) can be
	'nan' or 'NaN'. printf(Infinity) can be 'inf', 'Inf', 'Infinity'.
	Instead of guessing it, add undocumented option to print the
	value on the current system.

	* src/datamash.c: add undocumented options to print the values of
	  inf/nan/progname.
	* tests/datamash-{tests,tests-2,stats}.pl: use the undocumented
	  option as the expected result string of various tests.

2015-01-18  A. Gordon  <assafgordon@gmail.com>

	build: update to lastest gnulib
	* gnulib: Update to latest
	* tests/init.sh: update from gnulib

2015-01-18  Benno Schulenberg  <bensberg@justemail.net>

	doc: fix some inconsistencies

	doc: do not mention the nonexistent options -h and -v

2015-01-18  A. Gordon  <assafgordon@gmail.com>

	maint: update copyright year to 2015

	maint: enable po translation
	* bootstrap.conf: remove SKIP_PO
	* .gitignore: ignore fetched po/ files

2014-11-27  A. Gordon  <assafgordon@gmail.com>

	datamash: new option '--narm': ignore NA/NaN values
	Suggested by Brandon Invergo:
	http://lists.gnu.org/archive/html/bug-datamash/2014-11/msg00002.html

	* src/utils.{h,c}: is_na() new function
	* src/field-ops.{c,h}:
	    field_op_collect(): detect and skip NA/NaN values.
	    field_op_summarize_empty(): print results of empty values.
	    field_op_summarize(): call above function if no values.
	* src/datamash.c: update usage(), main()
	* doc/datamash.texi: mention new option
	* tests/datamash-tests-2.pl: add tests for new option.

2014-10-06  A. Gordon  <assafgordon@gmail.com>

	maint: update syntax-check rule to allow digits
	* cfg.mk: change regex in 'sc_prohibit_test_backticks' rule to allow digits
	  in file names of tests. Also remove two unrelated file names
	  (left-overs from 'coreutils')

2014-10-06  A. Gordon  <assafgordon@gmail.com>

	datamash: keep correct lines with '--full'
	Discussed in https://github.com/agordon/datamash/issues/3 ,
	and also existed in (very) old revisions of 'calc'.

	When using a selection operation (first/last/min/max/etc) which
	selects one line out of the group, and combining with --full,
	this ensures the correct line is selected (not just the correct
	fields intemixed with the values of the first line).

	* src/datamash.c: keep the correct line
	* src/field-ops.{c,h}: for selection operation, determine whether to
	    keep the line or not.
	* src/datamash-tests.pl: adapt one test of 'lst' to new behavior.
	* src/datamash-tests-2.pl: tests new --full behavour

2014-10-06  A. Gordon  <assafgordon@gmail.com>

	maint: add 'static-analysis' make target
	Example:
	    make static-analysis
	or:
	    make static-analysis-configure
	    make static-analysis-make

	Uses clang's scan-build static analysis tool to analyze the code.

	Use the two-step method to change code and re-analyze without
	re-runnning './configure' again.

2014-09-12  A. Gordon  <assafgordon@gmail.com>

	build: Enable build using Mingw cross-compilers
	* configure.ac: add flag and auto-detection for mingw
	* Makefile.am: add compilation flag
	* src/datamash.c: Use conditional printf type for size_t ("%zu" for all
	  unixes, "%Iu" for mingw)
	* HACKING.md: mention this issue
	* build-aux/check-remote-make-extra.sh: add mingw checks

	maint: add gnulib 'strsep' and 'random' modules
	These modules were missing when cross-compiling with mingw.
	* bootstrap.conf: add missing modules
	* m4/.gitignore: add files

2014-08-22  A. Gordon  <assafgordon@gmail.com>

	maint: expand HACKING.md file
	* HACKING.md: add notes.
	* cfg.mk: skip syntax-check on HACKING.md for few rules.

	datamash: improve delimiter edge-case handling
	* src/datamash.c: reject NUL delimiter, prevent shell-quoting problems
	  when sorting with single-quote delimiter.
	* tests/datamash-tests.pl: add tests for these edge-cases.

	datamash: improve '--group' parameter parsing
	* src/strcnt.{c,h}: new utility function to count characters in string.
	* Makefile.am: add new module.
	* src/datamash.c: parse_group_spec(): improve parsing code.
	* tests/datamash-tests.pl: add more tests.

2014-08-12  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new feature: accept named columns
	Examples:
	    datamash -H sum price < input.txt
	       (If 'input.txt' has a column named 'price').

	    datamash -H -g id sum 2 < FILE
	       (assuming FILE has column named 'id')

	Suggested by Shaun Jackman (sjackman@gmail.com).

	* bootstrap.conf: use gnulib's xstrndup module.
	* src/field-ops.{h,c}: if column parameter is not a valid number,
	  keep parameter as string, and defer finding the number to later.
	* src/column-headers.{h,c}: add a function to get the column number by
	  its name.
	* src/datamash.c: process_file(): if some column are named, try to find
	    the columns' number based on the header line, or exit with an error.
	  parse_group_spec(): optionally use names.
	  process_input_header(), remove_dups_in_file(): if grouping/operations
	    use named columns, find column number after reading the header line.
	  usage(): mention named columns.
	* tests/datamash-tests.pl: adapt existing tests to new error messages,
	  add new tests for this feature.
	* man/datamash.x: add named-columns examples.

2014-08-12  Assaf Gordon  <assafgordon@gmail.com>

	maint: improve 'syntax-check'

2014-08-09  A. Gordon  <assafgordon@gmail.com>

	maint: new 'coverage-expensive' target
	* cfg.mk: add 'coverage-expensive' target, for better coverage.

	maint: improve build-check scripts
	* build-aux/check-remote-make-all.sh: accept -c/-b/-m/-e arguments and
	pass them to 'check-remote-make.sh' script.
	* build-aux/check-remote-make-extra.sh: check cross-compilation and
	other compilers
	* build-aux/check-remote-make-git.sh: check building from git
	repository.
	* build-aux/prerelease-checks.sh: use new script.

2014-08-09  Assaf Gordon  <assafgordon@gmail.com>

	tests: test more rmdup edge-cases

	datamash: fix header line edge-case
	* src/datamash.c: if the input contains a header line but no further lines,
	print the output header line correctly.
	* tests/datamash-tests.pl: add tests for this edge case.

2014-08-09  A. Gordon  <assafgordon@gmail.com>

	datamash: improve header handling in rmdup/reverse
	* src/datamash.c: handle input/output header combinations.
	* tests/datamash-tests.pl: add tests
	* tests/datamash-transpose.pl: add tests

	maint: add git-log-fix file
	* build-aux/git-log-fix: this file is used by gnulib's build module,
	but after gnulib ugprade it was removed. Add it back.

	datamash: improve debase64 error handling
	* src/field-ops.{c,h}: return error code on failed base64 decoding.
	* src/datamash.c: handle decoding errors.
	* tests/datamash-tests.pl: test failed base64 decoding.
	* tests/datamash-valgrind.sh: test base64 encoding/decoding on large
	input.

2014-08-09  Assaf Gordon  <assafgordon@gmail.com>

	maint: improve output of parallel-tests
	* Makefile.am, init.cfg: redirect stderr to file-descriptor 9,
	improving error reporting using automake's parallel-tests.
	Based on similar setttings in GNU coreutils.

	build: add PURE attribute to functions

	build: add compiler warnings, -Werror

	build: speedup with gnulib's unlocked-io module
	* bootstrap.conf: add gnulib's unlocked-io module, providing
	significant speed improvement with readlinebuffer_delim().

2014-08-09  A. Gordon  <assafgordon@gmail.com>

	datamash: improve field-parsing code
	* src/text-lines.c:
	  line_record_parse_fields(): revise code in case of single-character delimiter.

2014-08-09  A. Gordon  <assafgordon@gmail.com>

	datamash: new operations: rmdup, noop
	rmdup: remove lines based on duplicated keys. Similar to:
	   awk '!a[$1]++'
	noop: no-operation (optionally with --full: print file as-is). Used for
	testing and profiling.

	* bootstrap.conf: use gnulib's hash module.
	* src/field-ops.{c,h}: add new operations
	* src/utils.{h,c}:
	   hash_compare_strings(): helper function for gnulib's hash module.
	* src/datamash.c:
	   noop_file(), remove_dups_in_file(): functions for new operations;
	   main(): call new functions.
	   usage(): mention new operation.
	* tests/datamash-valgrind.sh: test 'rmdup' operation with large input.
	* man/datamash.x: mention new operation, add example.

2014-08-09  A. Gordon  <assafgordon@gmail.com>

	maint: improve syntax-check
	* src/field-ops.c: replace tabs with spaces.

	datamash: improve assert and code coverage
	Asserts impossible conditions, and exclude from code coverage.

	tests: test more edge-case

	datamash: refactor operation output type
	* src/field-ops.{h,c} - define the output type (numeric/string) in the
	operations table, instead of during runtime.

	maint: fix/improve 'make coverage'
	* cfg.mk: force deletion of '*.gcno/*.gcda' when using 'make coverage'.
	Otherwise, some failures occur with lcov reporting 'unexpected end of
	file'.

	maint: fix syntax-check, part 3

	maint: fix syntax-check issues, part 2

	maint: fix source files for 'syntax-check', part 1

	maint: add few more syntax-check rules
	* cfg.mk: copy syntax-check rules from GNU coreutils.

	build: update gnulib submodule to latest

	build: disable --debug option
	* configure.ac: remove --enable-debug option.
	* src/*.c: remove #ifdef debug code.

2014-08-09  Assaf Gordon  <assafgordon@gmail.com>

	datamash: refactor field-splitting
	Instead of using gnulib's linebuffer directly, handle line-reading and
	field-splitting with a new structure 'line_record_t' .
	Field-splitting by delimiter is performed once, automatically, when a
	line is read.
	Pointers to each field are stored in 'line_record_t'.

	* src/text-lines.{c,h}: implement 'line_record_t' functionality.
	* src/datamash.c: replace 'linebuffer' with 'line_record_t' .
	* src/column-headers.{c,h}: use new functionality.
	* tests/datamash-tests.pl: adapt tests for new error messages.
	* tests/datamash-valgrind.sh: fix field delimiter in tests.

2014-08-08  Assaf Gordon  <assafgordon@gmail.com>

	maint: add 'default' case to prevent compiler warning

	datamash: bugfix: detect invalid column values
	* src/field-ops.c: add more checks
	* tests/datamash-tests.pl: add more tests

	tests: tests few more edge-cases

	datamash: improve field-op output handling
	* src/field-ops.{c,h}: pre-allocate buffer for string output when
	initializing the field-op, instead of malloc/free on every print.
	   field_op_reserve_out_buf(): ensures the field-op's output buffer is
	      large enough to hold the resulting string;
	   field_op_to_hex(), unique_values(), count_unique_values(),
	   collapse_values(), field_op_summarize(): use new output buffer.

2014-08-08  A. Gordon  <assafgordon@gmail.com>

	datamash: new line operations: md5/sha*, base64
	Line-Operations work on each line on the file (no grouping by key).

	Example: md5 on the first column on a file:
	    datamash md5 1 < FILE
	Is similar to:
	    perl -MDigest::MD5=md5_hex -lane 'print md5_hex($F[0])' < FILE

	* bootstrap.conf: add gnulib's md5/sha*/base64 modules
	* src/field-ops.{c,h}: implement new operations
	* src/datamash.c:
	    process_file(): implement new 'line mode' processing;
	    usage() mention new options.
	* man/datamash.x: mention new operations
	* doc/datamash.texi: ditto
	* tests/datamash-tests.pl: test new operations
	* tests/datamash-sha.pl: test sha1/256/512 opertions (requiring recent
	perl module Digest::SHA, this test might be skipped on systems with old Perl).
	* Makefile.am: add new test file.

2014-08-03  A. Gordon  <assafgordon@gmail.com>

	maint: fix wrong path in previous commit

	maint: added bash-completion script
	* contrib/bash-completion/datamash - bash completion script
	* Makefile.am: include script in distribution (without installing it)
	* .gitignore: ignore the binary executable only

2014-07-30  A. Gordon  <assafgordon@gmail.com>

	maint: update NEWS

	maint: gitignore more files

	maint: add pre-release build-and-check script

	maint: improve check-and-build scripts
	* build-aux/check-remote-make.sh: report system inforation, accept
	custom git branch to check-out.
	* build-aux/check-remote-make-all.sh: add more hosts.

	tests: skip some tests if perl isn't found
	* configure.ac: test for Perl, setup PERL_FOUND automake varaible.
	* Makefile.am: enable parallel tests for pl,sh files, and use a stub to
	skip tests if Perl is not found on the system.

	tests: mark valgrind test as 'expensive'

2014-07-27  A. Gordon  <assafgordon@gmail.com>

	tests: improve portability under qemu/binfmt
	When building with cross-compiling AND running (non-native) binary
	locally with qemu/binfmt, the argv[0] (binary name) is changed by qemu.
	argv[0] is used for error reporting.

	Example with most systems, argv[0] is exactly the same as the user
	entered:
	    $ datamash --foobar
	    datamash: invalid option --foobar

	    $ /custom/path/datamash --foobar
	    /custom/path/datamash: invalid option --foobar

	But under qemu/binfmt, argv[0] becomes the full path:

	    $ datamash --foobar
	    /home/gordon/projects/datamash/datamash: invalid option --foobar

	The tests were modified to detect the actual path reported by the
	program, or fall back to 'datamash' (as it was before).

2014-07-27  A. Gordon  <assafgordon@gmail.com>

	maint: improve build-aux check scripts
	* renamed remote-make-check.sh => check-remote-make.sh
	* renamed remote-make-check-all.sh => check-remote-make-all.sh
	* check-remote-make.sh: added features:
	  1) more source: local/remote tarball, git
	  2) more compressions: gz/bz2/xz
	  3) building from git with './bootstrap'
	  4) command-line options to set parameters to env/configure/make

2014-07-27  A. Gordon  <assafgordon@gmail.com>

	maint: improve debian-hardening build method
	* configure.ac: move option from './configure' to ..
	* cfg.mk: make rules.

	Instead of:
	    ./configure --enable-debian-hardenning
	Now use:
	    make deb-hard

2014-07-27  A. Gordon  <assafgordon@gmail.com>

	man page: improve hyphen vs minus characters
	* man/datamash.x: escape minus characters where needed, to distinguish between
	hyphens and minus/dash.
	Suggested by Alejandro Garrido Mota <alejandro@debian.org>

	tests: test I/O errors
	* tests/datamash-io-errors.sh: tests I/O errors using specially pre-mounted
	filesystems.
	* Makefile.am: add test.
	* build-aux/create_corrupted_filesystem.sh: script to create corrupted
	ext3 filesystem image.
	* build-aux/create_small_filesystem.sh: script to create small ext3
	filesystem which can be filled easily.

2014-07-25  A. Gordon  <assafgordon@gmail.com>

	datamash: add 'internal_error' macro
	* src/system.h: define 'internal_error' macro.
	* src/field-ops.c: Replace 'error' call with 'internal_error' macro.
	* src/datamash.c: ditto.

	build: don't re-generate manpage from tarball
	* configure.ac: search for '.git' directory, implying building from git
	(if found) or tarball (if not found).
	* Makefile.am: If building from GIT, re-generate manpage. If building
	rom tarball, assume the manpage is already there, do not regenerate it,
	and do not clean it (with 'make clean').

2014-07-23  Assaf Gordon  <assafgordon@gmail.com>

	datamash: new operations: transpose, reverse
	* src/datamash.c: new functions: transpose_file(), reverse_fields_in_file().
	main(): parse new options
	usage(): update help screen
	* src/field-ops.{h,c}: new function: parse_operation_mode().
	parse_operations() renamed to parse_grouping_operations().
	* man/datamash.x: updated man-page information
	* doc/datamash.texi: updated texinfo manual
	* tests/datamash-transpose.pl - new tests
	* tests/datamash-valgrind.sh - new tests
	* Makefile.am: include new tests

2014-07-18  A. Gordon  <assafgordon@gmail.com>

	datamash: fix typo in help screen
	* src/datamash.c: usage(): fix 'pto' => 'dpo' typo.

2014-07-17  A. Gordon  <assafgordon@gmail.com>

	build: fix configure's message of examples path
	* configure.ac: Report correct path of installed examples
	( PACKAGE_NAME = "GNU Datamash" while PACKAGE = "datamash" ).

	build: fix configure.ac parameter typo
	* configure.ac: fix param definition (otherwise --disable-debug would fail).

	build: add build prerequisites
	* bootstrap.conf: add gettext, gperf, makeinfo as prerequisites when
	building from git source.

	tests: test headers with whitespace delimiters
	* tests/datamash-tests.pl: add test for header line with whitespace
	delimiters.

	datamash: simplify line buffer handling
	* src/text-lines.c: linebuffer_nullify(): based on
	readlinebuffer_delim(), there is always a EOL-delimiter present in the
	buffer.

	tests: add Sample-Skewness test
	* tests/datamash-stats.pl: test sample-skewness with just 2 data points.

	maint: improve internal errors
	* src/field-ops.c: don't gettextize internal errors (no need to
	translate them), and exclude them from coverage reports.
	* src/utils.c: exclude internal error from coverage reports.

	datamash: improve gettext strings
	* src/datamash.c: usage(): improve wording, gettextize missed string,
	line up long options.
	Based on suggestion by Benno Schulenberg from the Translation Project.

2014-07-16  A. Gordon  <assafgordon@gmail.com>

	maint: update NEWS and README

	maint: syntax-check fix
	* Makefile.am: reduce spaces to appease 'make syntax-check' rule.

	build: Explicitly use perl for help2man
	* Makefile.am: Use 'perl' executable in $PATH for help2man,
	don't count on #! path in help2man - Hydra/NixOS doesn't have
	"/usr/bin/env", so it can't be used.
	NOTE: Using customized help2man (see commit 33c75d), so don't use
	system's installed help2man.

2014-07-15  A. Gordon  <assafgordon@gmail.com>

	Expand manual page
	* man/datamash.x: expand manual page with content from old help screen.

	datamash: shorten help screen
	src/datamash.c: usage(): shorten help screen.

	build: man page depends on .x template
	* Makefile.am: make manual page generation depend on datamash.x
	template.

	build: accept single example in help2man
	* man/help2man: Modify the EXAMPLES pattern detection to accept
	a single example.

2014-07-15  A. Gordon  <assafgordon@gmail.com>

	build: don't hard-code Perl path in scripts
	* man/help2man - Use "/usr/bin/env perl" instead of "/usr/bin/perl".
	* tests/datamash-stats.pl - ditto
	* tests/datamash-tests.pl - ditto

	Hard-coded Perl fails Hydra/NixOS builds
	(http://hydra.nixos.org/build/12516876), and is also generally bad form.

2014-07-06  A. Gordon  <assafgordon@gmail.com>

	GNU Datamash: remove files relating to external sites

	GNU Datamash: update README

	GNU Datamash: rename package

2014-05-20  A. Gordon  <assafgordon@gmail.com>

	maint: improve comments in tests

2014-05-13  A. Gordon  <assafgordon@gmail.com>

	maint: add 'texlive' package to Travis-CI

	maint: add 'texinfo' package to travis-CI

2014-05-13  A. Gordon  <agordon@wi.mit.edu>

	doc: add texinfo documentation

2014-04-30  A. Gordon  <agordon@wi.mit.edu>

	maint: update copyright/license in files

2014-04-19  A. Gordon  <agordon@wi.mit.edu>

	maint: update 'git-log-fix' due to gnulib upgrade.

	gnulib: upgrade to latest version

	maint: add more hosts to build-check script

	maint: .gitignore more files

	maint: fix 'make syntax-check' errors

	maint: replace old bcopy/bzero calls with memmove/memset

2014-04-15  A. Gordon  <agordon@wi.mit.edu>

	maint: add cygwin distribution packing script

	build: work-around strtold() bug on cygwin

	build: remove unneeded header from src/compute.c
	Compilation failed on Cygwin with this header.

	maint: improve automatic checks script

	Formalize rarely used debug option
	To build with debug option:
	   ./configure --enable-debug
	To use:
	   compute --debug [other options]

	build: fix compilation on Cygwin (missing nanl/expl)

2014-04-12  A. Gordon  <agordon@wi.mit.edu>

	maint: improve instructions in 'tag-new-version' script

	maint: .gitignore few more files

	maint: add script to build and test on multiple hosts.

	maint: Improve output of 'aws-upload' script

	portability: cater to systems without a stable sort

	tests: improve portability when detecting 'nan'
	It's 'nan' on most systems, but 'NaN' on some.

	build: portability: add missing <strings.h> header
	bzero() and bcopy() fail to compile on DilOS without it.

	tests: improve portability of mktemp
	Five X's fail on OpenBSD.

	compute: default to TAB instead of Whitespace as delimiter

	compute: fix typos

2014-04-11  A. Gordon  <agordon@wi.mit.edu>

	maint: add script to build & check on remote machine

	tests: improve portability of error message detection.

	tests: don't assume 'seq' is installed. SKIP if not found.

	build: improve 'inline' declaration for two functions.
	They failed to build on OpenBSD.

	compute: improve help screen

	compute: new operation: D'Agostino-Pearson omnibus (dpo) normality test

	src/utils: adjust math functions for consistant nan results

	compute: new operation: Jarque-Bera normality test

	tests: test mad/madraw with unsorted input

2014-04-10  A. Gordon  <agordon@wi.mit.edu>

	tests: tests more sequences

	compute: new operation: excess kurtosis

	compute: new operation: skewness (for pop./sample)

	src/utils: extract function for arithmetic mean.

	tests: document equivalent R code for stat tests + reorder.

	compute: new operations: mad,madraw

	tests: add tests for pop/sample stdev/variance

	compute: new operations: q1,q3,iqr

	compute: new operation: rand

2014-04-09  A. Gordon  <agordon@wi.mit.edu>

	build: add -Werror to ./src/ files.

	tests: fix comments

	compute: new operations: first,last

2014-04-08  A. Gordon  <agordon@wi.mit.edu>

	compute: set numeric output precision

	build: improve comment in configure.ac

2014-04-08  A. Gordon  <agordon@wi.mit.edu>

	build: Use compiler warnings from gl_WARN_ADD
	Those were wrongly dropped when changed to non-recursive makefile.

	The new flags are used only with the 'compute' sources,
	but not with the gnulib sources.

2014-04-08  A. Gordon  <agordon@wi.mit.edu>

	compute: detect invalid suffix in numeric input.

	fields-ops: remove 'keep_lines' feature.
	Perhaps will be added in future versions.

	compute: move numeric conversion into 'field-ops' module.

2014-03-22  A. Gordon  <agordon@wi.mit.edu>

	build: add helper scripts to EXTRA_DIST

	build: work-around for missing '.tarball-version'
	See:
	http://lists.gnu.org/archive/html/bug-hello/2014-03/msg00017.html

	maint: update helper scripts

	build: include 'help2man' and use it.
	Not all systems have "help2man" installed.

2014-03-20  A. Gordon  <agordon@wi.mit.edu>

	maint: add helper script to upload files to AWS S3

	build: update helper scripts to new name

2014-03-19  A. Gordon  <agordon@wi.mit.edu>

	RENAME: calc -> compute
	To avoid name conflicts with existing software.

2014-03-18  A. Gordon  <agordon@wi.mit.edu>

	TravisCI - no need for GNU Sed 4.2.2 anymore.

	calc: bugfix on --sort without grouping.
	'pipe_through_sort' was not cleared, caused errors on FreeBSD
	(all other systems seems not to mind calling pclose on a non-popen
	FILE).

	calc: sort+headers: don't use GNU sed.
	Implement unbuffered input to avoid GNU sed dependency.

2014-03-15  A. Gordon  <agordon@wi.mit.edu>

	calc: fix 'make syntax-check' errors

	build: improve ./configure messages

	gitignore: ignore more files
	Automatically generated by gnulib

2014-03-14  A. Gordon  <agordon@wi.mit.edu>

	Travis-CI: add 'make distcheck', disable encode binary
	No need to save the static binary from a Linux build - can be
	easily created anywhere.

	build: make-bin script: improve tar filename

	build: update gnulib

	build: add empty git-log-fix
	Will be needed if there are ever commit tyops/errors,
	see http://lists.gnu.org/archive/html/coreutils/2011-11/msg00004.html

	build: fix help2man issue on FreeBSD
	http://lists.gnu.org/archive/html/bug-hello/2014-03/msg00003.html

2014-03-13  A. Gordon  <agordon@wi.mit.edu>

	Travis-CI: don't get git submodules. 'bootstrap' should do it.

	Travis-CI: build with debian hardening flags

	src/calc: (temporarily) ignore the return value of fwrite
	using gnulib's 'clouseout' module will check for write
	errors when the program terminates.

	build: add '--enable-debian-hardening' to './configure'
	report CPPFLAGS at the end of './configure'

	tests: minor change to improve portability
	Previous syntax failed on Mac-OS X.

	Switch to non-recursive Makefile

2014-03-11  A. Gordon  <agordon@wi.mit.edu>

	Travis-CI: check coverage after successful build

	gitignore: ignore coverage files

	tests: tests sort-pipe failure

	coverage tests: exclude unreachable line

	build: remove unused module 'strnumcmp'

	src/field-ops.c: document unreachable code.

	src/field-ops.c: refactor free() code.

	tests: add tests, improve coverage

	src/column-headers.c: improve error message

	build: add scripts to generate coverage information

	build: define inline'd symbols in src/text-options.c
	Avoids 'undefined reference' errors when compiling with coverage
	instrumentation.

2014-03-10  A. Gordon  <agordon@wi.mit.edu>

	Travis: improve build, add scripts

2014-03-09  A. Gordon  <agordon@wi.mit.edu>

	build: add 'make-bin' script to distribution

	build: print message at the end of ./configure

	build: make GZ tarballs, not XZ.

	build: add required README file as a stub.

	.gitignore: ignore more files

	build: helper script to build static binaries

	README: update information, refer to website

	.gitignore: ignore build-related files

	calc: free column headers, add tests

	calc: bugfix for headers without grouping

	tests: don't use valgrind on static binaries

2014-03-08  A. Gordon  <agordon@wi.mit.edu>

	calc: new global case-sensitive option '-i' + tests

	build: fix minor errors

	refactor: extract common functions into modules

	maint: update copyright year, add GPL

	calc: conditionally compile '--debug' option.
	To enable it, Use
	    ./configure CFLAGS="-DENABLE_DEBUG"

	calc: improve help screen.

	examples: improve 'scores' example

2014-03-07  A. Gordon  <agordon@wi.mit.edu>

	build: remove 'make distcheck' from Travis-CI.
	It fails for unclear reason.

	calc: remove extraneous 'const'
	Caused compilation errors with clang.

	build: Add ".travis.yml" for Travis-CI

	calc: fix sort+headers, add tests

	examples: use '-s' instead of piping to 'sort|'

	calc: implement auto-sorting, add tests

	calc: add stub for '--sort' option.

	build: remove 'key-compare' from PO list

	calc: simplify code, don't use key-compare module.

	calc: simply parameters, remove '--key' option.

	calc: add '-T' as shortcut for tab separator.

	Merge branch 'examples'

	examples: add example files.

2014-03-06  A. Gordon  <agordon@wi.mit.edu>

	tests: fix valgrind test for 'make distcheck'

	tests: add valgrind tests

	calc: fix memory problems

2014-03-05  A. Gordon  <agordon@wi.mit.edu>

	examples: new sub-directory

	tests: add countunique tests

	new operation: countunique

2014-03-04  A. Gordon  <agordon@wi.mit.edu>

	tests: test single-line groups

2014-02-27  A. Gordon  <agordon@wi.mit.edu>

	tests: check empty input

	bugfix: '--full' and no groups would return the wrong line.
	Returned the last line of the last group, instead of the first line of
	the last group (as it returns the first line of every other group).

	tests: add --header-in tests

	headers: don't print group headers with --full.

	tests: test count on non-numeric fields

	remove unused cruft (from header feature)

	bugfix: Allow 'count' to count non-numeric fields

	key-compare: make 'blanks' public.

	implement headers support.

	Implement output headers (without --full).

	Command-line processing and Help for 'headers' options.

2013-05-04  Assaf Gordon  <assafgordon@gmail.com>

	build: touch missing files

2013-04-27  Assaf Gordon  <assafgordon@gmail.com>

	build: use gl_WARN_ADD for compiler warnings

	key-compare: fix compilation warnings
	-Wunused-parameter and -Wswitch-default will trigger warnings.
	change code to avoid them.

2013-04-26  Assaf Gordon  <assafgordon@gmail.com>

	Merge branch 'gitchangelog'

	build: add gnulib's gitlog-to-changelog

2013-04-12  Assaf Gordon  <assafgordon@gmail.com>

	calc: improve help message for help2man

	gettext-ify the package

	autoconf: switch from gz to xz

	system.h: minor improvements

2013-04-10  Assaf Gordon  <assafgordon@gmail.com>

	calc: use gnulib closeout module
	Protect against write errors at the end of te program,
	e.g. 'calc XXX > /dev/full' (which should fail).

	Merge branch 'better_autotools'

	GNU-ify: add auto-generated man page

	cleanups: remove 'scripts' directory
	There are no add-on scripts for calc (at the moment).

	GNU-ify: update src/Makefile.am

	GNU-ify: update tests/Makefile.am

	GNU-ify: add AUTHORS,THANKS

	GNU-ify based on GNU-hello, step 1 (configure.ac)

	code cleanups and more comments

	add GPLv3+ text to source files

2013-04-09  Assaf Gordon  <assafgordon@gmail.com>

	some more comments

	minor code cleanups

	minor code cleanups

	calc: groupping support (with proper output)

	calc: process groups.

	calc: add tests for groups

	calc: re-arrange headers

	calc: compare lines using --key

	calc: prepare code for key-compare.

	calc: import key-compare form coreutils.

2013-04-06  Assaf Gordon  <assafgordon@gmail.com>

	calc: expand '--help' section

	README: add simple use-case.

	tests: more tests (groups)

	tests: fix automake for out-of-tree builds

	calc: more tests

	calc: more tests

	calc: initial testing framework
	run with:  make check
	or: make check VERBOSE=yes
	or: make check VERBOSE=yes DEBUG=yes SAVE_TEMPS=yes

	calc: proper printing of output

	calc: field extractiong works

2013-04-05  Assaf Gordon  <assafgordon@gmail.com>

	calc: re-factor, prepare for field support.

	calc: bugfix for numeric values with groups

	calc: initial grouping, by empty line

	calc: re-implement unique/collapse

	calc: implement unique/unique-nocase

	calc: implement mode/antimode

	calc: implement stdev/variance, other fixes.

2013-04-04  Assaf Gordon  <assafgordon@gmail.com>

	calc: first shot at multi-valued numeric (median)

	calc: single-value operations working.

	calc: add stub input reading

2013-04-03  Assaf Gordon  <assafgordon@gmail.com>

	calc: create field/op structures

	calc: parse operation arguments

	calc: set GNU coding styles for VIM

	calc: process -z and --debug

2013-03-30  Assaf Gordon  <assafgordon@gmail.com>

	build system: fix auto generated version.c

	calc: add template code
	Add version, coyright, help, usage messages.

	gnulib: add modules.

	configure.ac: remove extra GCC checks
	gnulib doesn't compile with them.
	TODO: add GCC warnings to projec'ts build in src/Makefile.am.
	      (but only if using GCC).

	added gnulib submodule

	Initial Commit