Re[5]: [AWS] Non-lineal parsing time with Templates_parser

[Alejandro. R. Mosteo Chagoyen]

Dicho por Pascal Obry:

> Alejandro,

>> Umm... I've downloaded the latest Templates_Parser from Obry's page
>> (not AWS 1.4, but I think it's the same or better)

> Right.

>> and it shows only marginal improvements:

> Expected. The improvemen was for large tables. In your case I suspect
> that the problem comes from the Ada.Strings.Unbounded implementation.
> The 3.15 version is really slow compared to the new implementation
> found starting with 3.16.

I was about to post the following:

Ok, I've made a simple test and effectively the culprit is the
Unbounded_String implementation (though I don't know why gprof is
completely missing the track here). I'll search an alternative
implementation or, if nothing found, roll out my own. If I achieve
some noticeable improvement I'll post again my findings; maybe it
could be worth to be added to Templates_Parser.

>> So I suppose the mentioned optimizations aren't addressing this
>> particular situation.

> Right.

>> Any ideas?

> Jean Pierre Rosen has contributed an optimization to the templates engine
> by using a buffer. This avoids lot of costly reallocations. I had no time yet
> to integrate it. I'll try nevertheless to post this patch later today on this list.

Many thanks. I suppose I should try to get a better Unbounded_String
implementation while waiting for 3.16p O_o