Re: creating classes on the fly ??
Re: creating classes on the fly ??
- Subject: Re: creating classes on the fly ??
- From: "Stefan Arentz" <email@hidden>
- Date: Sat, 22 Sep 2007 17:18:55 +0200
On 9/22/07, Thomas Engelmeier <email@hidden> wrote:
>
> Am 22.09.2007 um 15:52 schrieb Daniel Child:
>
> > As the lexicons and corpus data get large (say up to half a million
> > or even a million records), are the "standard" data containers going
> > to pose performance problems?
>
> EVERY container will pose an performance problem with 1 Mega entries
> and an inappropriate algorithm.
>
> I'd recommend to write an tiny prototype that generates some
> arbitrary, random dataset and profiles your typical data access
> pattern (search 100 entries). Profile should include timing and
> memory, as soon your app starts swapping you have an immense
> performance hit.
I would actually recommend to do a prototype in Python just to
understand the algorithms involved. You will also be 10x more
productive in Python since text processing is so much easier.
> - Cocoa data structures -> easiest to use, might be inefficient
> - C++ std:: data structures -> no serialisation buildt in, beware
> std::string is not Unicode-Savy
You forgot to mention: crazy fast. Also, std::wstring will deal with
unicode just fine and Boost can provide serialization. If required.
> - Core Data -> Even with an far higher abstraction layer it might be
> faster due to under-the-hood paging out of unused entries.
I doubt it. The OPs use case sounds like a bad match for Core Data
actually. It is absolutely fine for doing structured queries but
analyzing parsed data that you have in memory in some handy data
structures is one of those things that maps terrible to that model.
S.
_______________________________________________
Cocoa-dev mailing list (email@hidden)
Please do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden