Changes in the "2010-01" release (2009-01-13 thru 2010-01-10):

* aggregate
  * Add support for min/max output.
	* Bug fix: avoid seg-fault when labels are not found.
	* Performance enhancements.

* fieldsplit
  * Bug fix: When running with -b, pad bucket numbers in output filenames
	  with zeros instead of spaces.
	* Rewritten in C for improved performance.
	* NOTE: This now depends on PCRE (dev package with library and header) in
	  order to be fully backward-compatible.

* filterkeys
  * New: a tool for filtering data streams using an external list of field
	  values to either keep or discard.

* mergekeys
  * Remove in-order requirement for auto-detecting keys.

* pivot
  * Bug fix: Do not reference invalid dbfr_t in pivot's output phase.

* Global
  * Distribution filename changed from "crush_<release>.tar.gz" to
	  "crush-tools_<release>.tar.gz".
  * Better error-checking for memory allocation.
	* General code cleanup.

* translate_field
	* Add a -k option for keeping the header row intact.
  * Add missing documentation for the -F option.


Changes in the "2009-01" release (2008-10-05 thru 2009-01-13):

* Field-label support: anywhere that you could use a field index as a command-
  line argument, there is now the option to use that field's label from
	the first line instead.

* Generate and install man pages.

* fieldsplit:  Add a "-b/--buckets" option, which limits the number of ouptut
  files which may be produced, while guaranteeing that field values are
  grouped together.  Performance enhancements.

* aggregate & aggregate2: -l/--labels and -L/--auto-labels options to
	control the headers for aggregated output fields.

* mergekeys: -D/--default option to specify the value to put into a
  mergeable field when there is no match in an outer join.

* Support stdin as a trailing filename option specified as "-" in several apps.

* translate_field: -x option to specify a substitution regular
  expression instead of a hard-coded mapping.


Changes in the "2008-10" release (2008-07-15 thru 2008-10-04):

* Make sure m4 is available at configure-time.  Previously it was assumed that
  the install host had m4 installed, but then the Perl scripts would fail to
  build.

* Check for presence of non-standard Perl modules, and only build scripts with
  those dependencies if the modules are installed.

* Allow escape sequences to be specified as delimiter arguments in utilities
  written in Perl.

* Add a "contrib" directory for non-core utilities, etc.

* Add options to mergekeys to allow user-specified key fields as an alternative
  to auto-detection of keys.  This currently does NOT remove the requirement
  that both files have a header line.

* Prepend all error messages with the name of the executable.

* Improve testing, and include test files in the distribution package.

* Bug-fix in find_not_processed - printed file names were not space-separated
  when -p option was specified.

* Avoid Perl warnings in csvformat when there is no linebreak at EOF, and fix
  commandline option specifications to avoid ambiguity.

* Make cutfield correctly handle empty fields.  It previously failed to output
  fields which were empty in the input.

* Fix a bug in reorder which causes a seg-fault when large ranges are 
  specified in the -f option.

* Bug fix in aggregate2 - print the header correctly line when the last field
  is one of the keys.  This bug was introducing a line break in the header line.

* Bug fix in aggregate2 - handle empty key fields correctly.

* Standardize and centralize the version reported by the applications.

* Add --Header option to dbstream to print the sql header fields if specified.