|
|
Ublas : |
From: Neal Becker (ndbecker2_at_[hidden])
Date: 2007-04-25 06:45:35
Gunter Winkler wrote:
> Am Dienstag, 24. April 2007 20:11 schrieb Neal Becker:
>> The enclosed patch tests the feasibility of resize change. It
>> appears to not break any of my code.
>
> I applied the patch to my local copy and started testing. It looks very
> promising.
>
> about vector<bool>: I don't think uBLAS is prepared for vector<bool> in
> any way. How would the enable_if version of resize() look like?
>
> Do you have a small example at hand that checks if all variants of
> resize do what they are supposed to do?
>
Thank you for considering this.
1. The patch I sent is intended as proof of concept. The purpose is to
demonstrate that we can introduce the desired resize into storage, vector,
and matrix without breaking existing code. So far, the patch introduces
changes to storage::unbounded_array, and vector. To complete the job, we
need to look at other storage classes and other containters.
2. The testing I've done is only that my current batch of code, which uses
ublas quite a bit, still compiles and seems to produce the same results. I
don't have any real regression test, or specific test of the the new
functionality.
3. OTOH, I think this change is simple enough that we can mostly convince
ourselves that it is correct by inspection, providing it doesn't break
anything.
4. My comment about enable_if is probably wrong - but if we accept that
vector<bool> is not important - then maybe we can just ignore it.