Re: Please list all the Ada libraries you know

[Adam Beneschan <adam@irvine.com>]

On Wednesday, March 6, 2013 1:01:27 PM UTC-8, J-P. Rosen wrote:
> Le 06/03/2013 18:08, Adam Beneschan a écrit :

> > Anyway, though, my feeling is that once one has chosen to build a
> > "source model" compiler--which means, if I understand it correctly,
> > that if you WITH a package then the compiler has to read the original
> > source of the WITH'ed package specification--then one has already
> > decided that disk access time isn't an issue.  (If it were, the
> > compiler could do something to create a smaller, compressed version
> > of the specification that it would read in instead of reading the
> > original source.)  So whatever good reasons there might be for GNAT
> > to default to a particular case convention, saving disk read time
> > probably isn't a valid reason.
> 
> It may well be the other way round. In a library model, you keep the
> same information (about specs) in the form of tables, and it is not
> obvious that they use less space than the source form.

If you keep *all* information about spec declarations in the library file, including information that isn't explicitly in the source file but that the compiler fills in itself, you may be right.  But that isn't what I said.  Clearly, the compiler could create a representation of a source specification that includes only the important information that already exists in the source file, and store it in a way that's much smaller than the original source.  I suppose that you could still call this a "source model", since the file would still be in some sense a representation of the original source, without any additional semantic information.  But as far as I know, GNAT doesn't do anything like that.  As far as I know, it rereads the original source of every package you've WITH'ed, including all the comments.  (Again, someone please correct me if I'm wrong; but if GNAT creates any kind of 
stripped-down version of the original source, I haven't seen it.)  This doesn't sound like an implementation that makes decisions based on saving disk access time.

                         -- Adam