Re: scm_num2float() ?

[Dirk Herrmann]

On 4 Sep 2001, Michael Livshin wrote:

> Dirk Herrmann <address@hidden> writes:
> 
> > However, this would still give 42 functions (instead of 21 if we use
> > the reduced set of conversion functions).
> 
> it looks like you think that many functions is bad.  I disagree:

Well, let's say that I'm in favor of a minimalistic approach.  IMO, we
should make Guile as small as possible - but not smaller.  I understand
that reducing the set of conversion functions to the set that I have
suggested is controversial.  Whether this reduction crosses the border to
making guile 'smaller than possible' is something we have to decide.  Both
solutions are possible.

> * in terms of space overhead, 21 small functions is nothing.

I have not tried it, but it I would find it interesting to know how much
'nothing' really is.  And, many 'nothings' add up to infinity.  However, I
agree with you about the following points:

> * maintenance overhead is nonexistent, if we generate these functions
>   from macro "templates", like now.
> 
> * cognitive overhead is also nonexistent -- you just say "for every
>   standard C integral type foo, there are scm_num2foo_ext, scm_num2foo
>   and scm_foo2num".

Best regards
Dirk Herrmann