The output is now limited to 80 characters; this is useful when cat'ing to a
file and terminal.get_size() is called. The previous behaviour was to always
output according to the terminal width, something that probably isn't what
someone usually wants when outputting to a file.
Filtering can be specified using an arbitrary number of -x (or --exclude)
parameters. The filtering works in a similar fashion to an inverted grep,
meaning that matched file names will be excluded from the generated
statistics.
To accommodate this modification, the changing of working directory (when
a repository is specified) has been moved from the system module to the
main gitinspector module. This enables us to change directory in one go
when gitinspector starts, thus removing the need for the previous system
module.
The following concurrency components control the implementation:
* A semaphore that limits the number of running blame instances to the
number of cores detected in the system.
* A lock that protects internal data structures inside the blame module.
The threading functionality is currently disabled because some additional
changes are needed in the system module before it can work correctly.
This was added because "git blame -C -C" really takes a lot of time and
something is needed to indicate that gitinspector is actually running during
this stage.
The progress text is not shown if stdout points to a non-terminal.
It should now be able to handle multi-line comments that are only on one line.
Furthermore, it should also handle multi-line comments from languages such as
python (where the identifiers for comment begining/end are identical).
The metrics module currently only includes a very simple ELOC (estimated lines
of code) metric, but will be expanded eventually with some other metrics.
If any missing files are detected during analysis, gitinspector reports them
in a list. If, on the other hand, the --checkout-missing option is supplied,
gitinspector will try to checkout any missing files instead.
