Re: [Libreoffice] [REVIEWED 3-4-2] [PUSHED 3-4-2] Update mdds to 0.5.3

On Tue, 2011-07-19 at 12:05 +0100, Michael Meeks wrote:
> On Mon, 2011-07-18 at 16:57 -0400, Kohei Yoshida wrote:
> > Just to tone down this statistics a bit.  The vast majority of changes
> > are in fact code relocations with no changes
>
>       So - I've pushed it to -3-4-2 - after doing my best at reviewing the
> changes; but I'm really unhappy with this.
>
>       I would like to understand:
>
>       * why mdds is maintained in a separate place - can we not
>         whack it into the calc/ git module ?

Because I'd like mdds to be product-neutral, general purpose library.
That's the primary motivation on why I started this project, and I
intend to keep it that way.  So, no, I don't want this to be a part of
the repository of one product that uses it.

>       * why it is not branched and frozen with the same cadence - so
>         we have small fixes back-ported, rather than having to review
>         large code movements / renaming / etc.

It is because the repository was a Mercurial repository up until I
converted it to git *only a few days ago*, and I was extremely
uncomfortable using branches with Mercurial.

Now that the repo is git, I can use branches more aggressively.

>               + I'm well up for API cleaning and big beautification
>                 patches etc. but prefer them on the unstable branch.
>
>       Also - some UI pointers for QA to do some more hard-core testing of its
> various uses cases here would be much appreciated.

Sure.  The code mostly affects matrix calculations in formulas.  This
can be either a function that takes in-line formula expressions, or an
expression whose result is an array which spans over multiple cells.

Even when you don't pass or receive matrix data to or from a function,
the code itself may still use the matrix objects, so those functions are
affected as well.  Many statistical functions fall under this category.

Kohei

-- 
Kohei Yoshida, LibreOffice hacker, Calc
<kyoshida@novell.com>