[m-rev.] for post-commit review: getters and setters in deep profiler

[Paul Bone]

On Tue, Jan 06, 2009 at 05:53:13PM +1100, Zoltan Somogyi wrote:
> On 06-Jan-2009, Paul Bone <pbone at csse.unimelb.edu.au> wrote:
> > Is it possible to have a getter or setter named "_get_FIELD" or
> > "_set_FIELD", that is with a blank data structure name?
> 
> Yes, though a programmer would have to be crazy to write such code.
> The code should handle it anyway.

Oh!  Now I understand.  I thought this was for procedures created by the
field access code such as:

   Struct ^ field = Value.

Which don't get compiled as procedures anyway.  But instead for the
human-written functions and procedures like hlds_info_get_detism

> > If not this predicate should be split in two such that a datastructure
> > name of non-zero length is enforced.  I beleive that this would be more
> > readable.
> 
> I don't see how "not there" is more readable than "there even though in
> a less than ideal format".

I don't know what you mean.  But my comment here isn't relevant anymore.

> > In any case it's probably best to simply split the string on the '_'
> > delimiter and ensure that each of the three components matches the
> > strings in is_getter_or_setter 
> 
> What THREE components? Underscores can appear in names ANY NUMBER of times.

This was due to my misunderstanding.

I understand how is_getter_or_setter_2 works now and why it is written
this way (without splitting the string on the '_' delimiter).

Thanks.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20090107/ceaa8f3e/attachment.sig>