[Teammetrics-discuss] Git repositories parsing is ready.

[Andreas Tille]

On Wed, Jul 13, 2011 at 06:12:28PM +0530, Sukhbir Singh wrote:
> 
>         Table "public.gitstat"
>      Column     |  Type   | Modifiers
> ----------------+---------+-----------
>  project        | text    |
>  name           | text    |
>  package        | text    |
>  changes        | integer |
>  lines_inserted | integer |
>  lines_deleted  | integer |
> 
> I had initially decided to put everything under one relation called
> 'commitstat'. But as you see the schema for 'gitstat' above, let's
> suppose we have this for one relation called 'commitstat' that will
> have all VCS. But then for SVN parsing, we are not counting the
> 'lines_inserted', 'lines_deleted' or 'changes' even!

Ahh, OK, so you want to tell me the SVN API is weaker and not able to
uncover those values?  That's a shame!  I would really like to regard
the VCSes equally in the team statistics.  If this is not possible I
have no clue how to compare those teams like Debian Med and Debian
Science where people (sometimes the same people) are commiting to SVN
*and* Git.

> That is why I
> thought it to make the VCS distinct. When we have another VCS like hg
> for example, we will have a relation _specific_ to that.
> 
> Don't worry about management, I am always ready to write extra code if
> it keeps it clean :-) But I will do as you say. If you still want to
> separate it, let me know but then for some VCS the records will be
> blank as is obvious.

Well, if there is no good chance to get those extra data from SVN the
separate tables do make sense.  However, I think I remember that this
Java thingy svnstat (? or so - it was mentioned in the application phase
about the "do not reinvent existing tools" discussion) just had quite
detailed numbers.  I'm not perfectly sure and to busy to check again.

> As far as the upstream changes are concerned, I am trying to find the
> best possible way and will have something soon.

Fine.

Kind regards

       Andreas.

-- 
http://fam-tille.de