Log of the #nice channel on calvino.freenode.net

* bonniot joins09:52
* arjanb joins13:04
<bonniot>hello arjan13:41
<arjanb>hi
<bonniot>i remember you implemented sometime something that was similar to removing non-minimal methods for overloading resolution. remember?13:42
* arjan_b joins13:49
<bonniot>what's up? did you get my question?13:50
<arjan_b>connection probs
answer to question is yes13:51
* arjanb leaves
<bonniot>what was it exaclty?
<arjanb>wdym13:52
<bonniot>what was the feature?13:53
<arjanb>better handling of patterns with an additional tc to resolve ambiguity13:54
MethodBodyDefinition line 224 to 286
<bonniot>thanks. yes, i see the TODO comment ;-)13:57
did you analyse the bug about integer comparison?14:46
<arjanb>i know what the problem is and it only causes incorrect error messages14:47
<bonniot>good. is it difficult to fix?14:49
allo?16:02
<arjanb>sorry was away16:14
it's not difficult and i had planned to improve the implementation anyway16:15
<bonniot>ok16:28
i'm getting ready to commit the change from functions to methods with default implementation22:08
reviewing the diffs...
<arjanb>how does it work out for functions on primitives?22:48
<bonniot>like methods on primitives :-)22:51
there will often be only one implementation in that case22:53
<arjanb>but not for operators
<bonniot>but not always, since you can dispatch on values, thanks to you ;-)
you mean inlined operations?
<arjanb>yes
<bonniot>that's untouched22:54
they are not ToplevelFunctions
that class disappeared :-)
testing is very welcome, as this is a quite big change22:56
<arjanb>maybe a change in the other direction is useful too23:01
<bonniot>?23:02
<arjanb>if the compiler can split up method implementations in multiple method declarations than some of the dispatch can be done at compile time23:03
<bonniot>yes, but that's only an optimization
the first thing is to update the design, to have a nice and powerful semantics23:04
then, sure, it will be useful to do some dispatch optimizations
<arjanb>but that is what happens for covariant return types right?23:05
<bonniot>actually, the next semantic step can be to allow redefinition of the return type
yes, it will fit well together :-)
the dev version is uploaded, and the changes are commited, so testing is open :-)23:07
<arjanb>ok but i'm waiting for diffs23:08
<bonniot>cvs diff
<arjanb>why isn't it possible to merge defaultmethodimplementation and methodbodydefinition?23:28
<bonniot>they have a common superclass, where all that is common is
just the specific parts are in each subclass
this makes shorter and simpler classes23:29
the main difference is that for the Default one, the declaration is already known, so there is no need to look for it
and there is no separate list of patterns23:30
(btw, it would be good if Pattern was split in several subclasses too)
<arjanb>agreed23:31
<bonniot>ok, i'll go an rest now ;-)23:32
good night
<arjanb>good night
* bonniot leaves
* arjanb leaves23:34

Generated by Sualtam