Re: Unicode support pre alpha

Joel de Guzman wrote:

> Hi Mathias,
> 
> Yes, I am aware of your project. There are some things that kept me from
> using it. First, I started working on unicode support at a time before
> your library was taking shape. I was waiting for it to get more mature.
> One thing I noticed is that your library is not optimal with the use
> of tables, when I had a chance to look, it was not using optimal 2 stage
> tables and there is no way to pick only the features that I need.

It does have a two-stage table, albeit it is not really optimal. The 
first table contains pointers to the second-level table and not indexes, 
which makes it somewhat bigger than necessary.

Duplication is only avoided when all properties are the same for a given 
per-code-point entry, but if you did that per property you could avoid 
much more, of course.

Most of this I didn't write myself, I inherited that from some code 
Rogier van Dalen wrote.


> You
> get to have all the databases wholesale. Correct me if I'm wrong.

There is indeed no way to pick only the features you need. I am not even 
sure myself how much I want to put into that table.

The way I see it, it should be a shared library. I suppose there could 
be variants with different sizes and information (and actually, there 
are already two variants albeit it's not included in the build scripts 
yet), but I didn't really plan on making a small subset you can use 
header-only.


> The
> library seemed rather immature for me at the time and I couldn't see
> active post GSOC activity.

I have had a busier year than I anticipated, but getting it reviewed and 
accepted is certainly my intention.


> The "future work" section does not even
> mention of an eventual Boost review (or is it in the review queue
> already?).

I'm addressing the future work plus some other things I didn't mention, 
such as changing the names of the concepts and adding codecvt facets, 
and then I'm submitting it for review after good documentation, example 
and test updates.

I also wanted to do some evaluations about performance and 
specializations for contiguous and well aligned memory before doing so, 
but I've been asked to get it reviewed as quickly as possible.


> I'd love to use your library if these concerns are allayed.
> Most importantly, I'd like to know what the plans are for Boost inclusion.
> Spirit could very well be an incubator of your project as it has done in
> the past with Fusion and Phoenix. Tell me your thoughts.

If you tell me what you need to use it in Spirit, I can make the changes 
accordingly.

I would certainly appreciate ideas on how to make the database or the 
parsing/generating scripts better.

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev