[PATCH 1 of 2] parsers: incrementally parse the revlog index in C

[Bryan O'Sullivan]

On Tue, Apr 3, 2012 at 2:15 PM, Matt Mackall <mpm at selenic.com> wrote:

> > Instead of using Py_BuildValue to create each element, we introduce
> another
> > extension type that tracks a rev without building a tuple, and lazily
> > returns the elements currently contained in the tuple (perhaps caching
> them
> > in case of reuse?).
>
> No, I'm suggesting something a bit more fundamental. Instead of this
> data structure:
>
>  [(x1, x2...), (y1, y2...)] # an N-element list of 8-tuples
>
> ..we want this one:
>
>  ([x1, y1...], [x2, y2...]) # an 8-tuple of N-element lists
>

Oh! Column order vs row order. Now I gotcha.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mercurial-scm.org/pipermail/mercurial-devel/attachments/20120403/01a7b380/attachment-0002.html>