Re: [Boost-build] CVS: boost/tools/build/new generators.jam,1.44,1.45

From: David Abrahams (gclbb-jamboost_at_[hidden])
Date: 2003-05-10 22:49:28

Vladimir Prus <ghost_at_[hidden]> writes:

> David Abrahams wrote:
>
>> Modified Files:
>> generators.jam
>> Log Message:
>> Improved commenting, naming, formatting. Began refactoring.
>
> I've a problem with this log message. Suppose you look at it a month ago. Can 
> you glean anything --- what was refactored, how, why? I think no, so this 
> message is as good as no message. Also, to understand what changed I need to 
> browse the entire diff, which includes not so interesting comment/naming 
> changes. 
>
> Could we agree to use a bit more verbose log messages. For example, if message 
> for this commit read:
>
> * new/generators.jam: Improved commenting, naming, formatting
> (generator.match-rank): Count also requirements when determining the rank, 
> not only optional properties.
> (generator.run): Always set 'multiple' to true.
>
> I would immediately understand the important changes. 

I'll try. I usually find log comments to be useless, so I don't give
them much attention, but I can understand that not everyone feels that
way.

>> - return [ sequence.length [ set.intersection
>> - [ optional-properties ] : $(properties) ] ] ;
>> + # We're only used to rank matches based on the number of
>> + # optional properties that appear in the property set.
>> +
>> + # That seemed a little weak to me, so I changed it: now we
>> + # account for the number of properties and features that
>> + # were matched as well. -- dwa 5/6/2003
>
> I'm pretty sure the current behaviour was prompted by some real use case from 
> my work usage of V2. Once I'm able to update from CVS, I'll try to find if 
> your changes break that usage.

Please do let me know.

>> {
>> + multiple = true ; # The tests seem to tolerate this; will
>> + # remove the parameter altogether in the
>> + # next revision to see what I learn -- DWA
>
> I actually wonder if this means "multiple" is unneeded, or that simply we 
> don't have tests for that. The last paragraph of 'generators.jam' comment 
> still makes sense to me --- but I don't have real examples to present.

It is still clearly needed, but perhaps no variability is needed at
this point in the code. Anyway, I'll be posting a proposal for
eliminating multiple shortly...

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com