meta-comment on typing Per Bothner (05 Oct 2005 17:35 UTC)
Re: meta-comment on typing John.Cowan (05 Oct 2005 22:00 UTC)
Re: meta-comment on typing Per Bothner (05 Oct 2005 22:14 UTC)
Re: meta-comment on typing John.Cowan (06 Oct 2005 04:55 UTC)
Re: meta-comment on typing Michael Sperber (06 Oct 2005 06:03 UTC)
Re: meta-comment on typing Per Bothner (06 Oct 2005 15:35 UTC)
[SRFI 77] integer-length and integer-sqrt Jens Axel Søgaard (06 Oct 2005 15:54 UTC)
Re: meta-comment on typing Michael Sperber (06 Oct 2005 16:17 UTC)

Re: meta-comment on typing Per Bothner 06 Oct 2005 15:35 UTC

Michael Sperber wrote:
> I personally have no objection to type declaration, but disagree that
> the code becomes more readable.  In fact, I've seen plenty of evidence
> that the exact opposite is the case, both in the C world (with type
> declarations) and in the R5RS/CL world (without).

Type declarations, especially of procedure parameters, is an essential
part of the documentation and specification of a procedure.  As a
compact, easily-understood, machine-checkable specifiction that as
a side benefit (often) improves performance it seems a no-brainer.

The R5RS pervasively uses type declarations, in the form of
conventions for variable names.

I agree that for short functions adding type declarations make the
code longer and hence harder to read - but the key issue is whether
it makes the code easier to understand.  I'm convinced it usually does.

> Some anecdotal evidence can be found in the paper by Egner et al.
 > cited at the bottom of the SRFI.

I found the paper, but no anectodal evidence about type declarations.
Rather the opposite: The unexpectedly slow behavior mentioned in some
programs due to decimal points wouldn't have happened if they had
used type declarations.
--
	--Per Bothner
xxxxxx@bothner.com   http://per.bothner.com/