Re: size of .po changesets

Excerpts from Tom Lane's message of jue ago 23 13:33:46 -0400 2012:
> Roger Leigh <rleigh@codelibre.net> writes:
> > On Thu, Aug 23, 2012 at 11:21:35AM -0400, Alvaro Herrera wrote:
> >> Yeah, IMHO .po files are handled pretty badly by SCMs.  I wonder if we
> >> could reduce the amount of git churn caused by those files by simply
> >> removing all comment lines from these files as they are exported from
> >> pgtranslation into postgresql.git?
>
> > Have you considered adding "--no-location" to XGETTEXT_OPTIONS in
> > po/Makevars?  This stops the massive churn through line renumbering
> > changes.
>
> I think the line numbers are actually useful to the translators --- at
> least, the theory behind having them is to make it easy to look at the
> message in context.

Yeah, and I, for one, do use them quite a bit to look up the code
context when the message is unclear.

> Alvaro's point is that the copies of the .po files
> in our SCM are pretty much write-only data, and could be stripped
> relative to what the translators work with.

Right.

--
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services