Re: [Chicken-hackers] [PATCH 0/4] Some symbol and module-related patches

[Evan Hanson]

Hi Peter,

On 2017-05-13 11:58, Peter Bex wrote:
> Two quick questions: Given that C_namespaced_symbolp is inlineable and
> unsafe, shouldn't it be named C_u_i_namespaced_symbolp?
> 
> And, given that it's inlineable, I think it would be better to put it in
> tweaks.scm (for example) as a define-inline.  I think this would be much
> more efficient.  Perhaps this even dwarfs the scan time of memchr.

Good point, I agree. Here's an updated patch that fixes the name and
inlines the C call.

Unfortunately, it can't just go in tweaks.scm, since that file includes
declarations that are only suitable for use in the compiler (e.g.
"no-argc-checks"), whereas this is one of those odd procedures that
straddles the boundary between compiler and runtime, being used in
expand.scm and modules.scm which both include user-facing procedure
definitions and can't include tweaks.scm as it is.

I think we probably ought to split that file up into two, one for
declarations for compiler-only code and one for inlined procedures that
are OK to use anywhere in core. There are also a handful of duplicated
definitions that appear in both expand.scm and modules.scm (the inline
versions of "getp"/"putp", for example) which might be good candidates
to move there. For now, though, I've simply changed patch to include a
"define-inline" wherever the procedure is needed. This isn't ideal, but
I hope it'll do for now. Let me know what you think.

Cheers,

Evan