Re: [boost] [Soc suggestions] TR1 implementation

From: John Maddock (john_at_[hidden])
Date: 2007-03-07 04:44:45

AlisdairM wrote:
> How close are we to having a releasable Boost implementation of TR1?
>
> There are also a few interface changes for some Boost components to
> catch up with, most of which seem implemented in the HEAD already.

It's a good idea IMO.

> I know we are missing the following libs:
>  unordered containers

Mostly just needs docs and tests, which might not make it a very interesting 
SOC project.  However, we could split this project in two:

Phase 1: bring what we have up to review standard and get it in the queue, 
hopefully sheppard it through before the end of the SOC period.

Phase 2: other hashing related projects: a fixed size closed-hash table (cf 
Boost.Array).  Maybe other hash algorithms.

>  special math functions

I've been tempted to suggest this for obvious reasons, but I'm a bit leary 
of suggesting a student work on something that hasn't been reviewed yet. 
Any thought on this?

There are still four special functions to do for TR1: zeta, exponential 
integral, and the two hypergeometric functions.  The first two are 
straightforward, while the hypergeometrics are verging on impossible (IMO 
they have no right being there, at the very least they are untestable in 
practice).  At present I'm ignoring these until the review.

>  C99 library

This is harder: a lot of it involves changes to existing functions (the 
printf family for example) that we can't implement.  We should do a quick 
audit and see what can be done though.

> Is this a project we could imagine completing with a SoC student or
> two?

Maybe, thoughts?

John.