More than two years after the 2.0.0 release we are releasing a new
version, it is 2.1.0. It includes a lot of bug fixes provided by you
guys during last 1K days.
Again, thanks to the people that have contributed during this time!
The .deb and .tgz files are available at:
https://projects.libresoft.es/projects/cvsanaly/files
================
cvsanaly 2.1.0
================
This is the CVSAnalY which bring together a lot of commits done during
2009, 2010 and 2011.
Bug fixes:
* Fix --version command line option
* [metrics] Fix file extension extraction from paths
* [metrics] Fix path building in cvs when module is '.'
* Encode strings to utf-8 before writing to stdout/stderr
* Fix prefix branch for paths in multi-paths actions. In SVN it's
possible to copy/move/replace a path from one branch into
another. In this case the action branch is different from the
branch of the second path. This usually happens when creating a
branch: A /branches/foo (from /trunk: 1234)
* Remove leading / from path before querying the cache.
For repositories like git where paths don't start with / we have to
remove the leading / from the built path before inserting in into
the cache, but also before querying the cache
* Fix compatibility in SQL statement LIMIT.
The previous declaration of the LIMIT statement in the class
ICursor is incompatible with Postgresql. However, both SQLite
and MySQL supports the statement LIMIT ... OFFSET which also
support Postgresql.
* Fix SQL portability issue in view action_files.
The syntax used when the view action_files is created is not
compatible with Postgresql, which requires the clause 'AS' when
an alias is defined. The clause 'AS' is optional in MySQL and
SQLite. Fixes bug #28.
* Add conditional use of args when executing a query.
SQLite cursor's does not handle extra parameters in a query that
does not have paraments. Instead of passing None, it should not
pass any argument at all. Fix Bug #30.
* [git] Fix a crash when there's a commit in a svn tag. Fixes #75
* [cvs] Remove additional trailing \n added to commit messages.
Problem spotted by David Gray.
* Fixes incorrect licence header in most of the files. Fixes #240
* [extensions] Missing self before class attribute. Fixes #248
New features:
* Add a flag command line option to ignore errors in metrics
extension
* Add ICursor class to split up queries.
The result of a query is always stored in memory, which means that
queries that return a large amount of data might fill up the
memory.
ICursor is a class that uses a real cursor emulating the execute +
fetchmany API, but splitting up the query by using limit, so that
the query is executed with different limit values everytime a
fetchmany is required.
* [extensions] Add Blame extension.
It adds a new table with information about the amount of lines
owned by every author of every file for every revision
* [extensions] Add Patches extension to save diffs of changes.
It adds a new table to store the patch of every commit
* Add a configuration sample file. Fixes bug #27
Performance fixes and improvements:
* Do not save caches to disk on every database commit since it's a
very heavy operation, it is done only once at the end of the
parsing process
* Paths in different branches are now considered as different
files. That was obvious with SVN because branches are directories,
but in CVS and Git exactly the same path can be in different
branches.
* Use cat instead of checkout/update to get the contents of files
in metrics extension. Multiple cat commands can run in parallel
which improves performance quite a lot.
* Proccess finished jobs every 100 jobs to make sure there are jobs
in the done queue and removing the timeout the first time
job_pool.get_next_done is called to make sure we always get a job.
* Update the adjacency matrix sequentially instead of re-create it
for every revision. It drastically improves the performance of the
Metrics extension
* Add debug messages to git parser
* [git] Rework git parser to save memory with large repos
* [metrics] Remove unnecessary code
* Add get_filename() to FilePaths to get just the file name of a
given file_id
* Detect and ignore empty branches in git
* [git] Ignore commits done on svn tags directory. This is a
workaround for a bug in the GNOME Git migration
* DBContentHandler.py refactored
* Do not call DBContentHandler methods until proxy has finished.
DBContentHandler creates a database connection in begin() and
keeps it opened until proxy finishes. For repos with a very long
history it might take a long time until proxy finishes and mysql
resets the connection of the DBContentHandler because of
inactivity timeout.
* Improved Debian files
Documentation:
* Rename ChangeLog to ChangeLog.pre-git and autogenerate it
* Improved README file
Thanks to all contributors:
Santiago Dueñas
Jesus M. Gonzalez-Barahona
Carlos Garcia Campos
Germán Póo-Caamaño
Juan Francisco Gato Luis
David Gray
Luis Cañas Díaz