Re: NSCharacterSet bloat?

[Richard Frith-Macdonald]

On 23 Jan 2006, at 20:20, Derek Zhou wrote:

> Hi all,
> It looks like now NSCharacterSet in base (1.11.2) contains more than
> 600k of const static data (all the NSCharacterSet bitmaps). To put
> thing in persective, it is more than 25% of the total TEXT size of the
> compiled libgnustep-base.so file. I check base-1.9.1 I had from a
> while ago, base used to store them in files (they still exist, just
> out of date) and read them in on demand. IMO, the old behavior is much
> saner. Embedding large data structure in the source code just don't
> look right to me. Why do we made this change? If the concern is to
> share the bitmaps among multiple app instances, we can mmap them
> instead of read them anyway.

This is because some people (windows users primarily) voiced  
objections to the dependencies on other files and wanted to minimise  
the number of files installed.  Many times I got the impression that  
objections are aesthetic, but the idea of minimising the number of  
external resources that can be misplaced and break things seems  
reasonable.  Now it sounds like you have your own aesthetic  
objections to the new way of doing things, I guess you can't please  
everyone.

> Also the file NSCharacterSetData.h seems like auomatically generated
> by some program and I cannot find the program in the base
> tarball. Inconvienence aside, it is also against common practices of
> distributing free software.

The charsets are provided both in the header files in the source code  
and (less readably) in the old data files.
The location of the utilities used to generate the files is given in  
the readme file in the NSCharacterSets directory containing those  
data files.