Re: Semantic difference of source files

John wrote:


> [If you're only going to compare the ASTs to see if they're the same,
> all you need to save is a hash.


Ah, but then you lose the ability to selectively compare, which is the
biggest advantage of the method I propose. The programmer could write
a whole class (or various global functions), including implementation,
in the same file. Then it can still be automatically determined if
only the implementation has changed, or if the interface has changed
as well. (All private members of the class belong to the
implementation. All public members belong to the interface.)


However, I suppose you could save a separate hash for the interface
and for the implementation. Also a separate pair for each class and/or
each function in the file, maybe. It's a good idea.


> The question of why nobody seems to
> care about compilation speed any more. -John]


Are you saying nobody cares about compilation speed anymore? (Your
sentence seems to be gramatically ambiguous.)


It's one of the big annoyances during development, I'd say. Especially
for the really big projects. The compiler I'm working on is relatively
small. Still, I have to wait a minute or two if I change an
implementation detail or even a comment in an often used header file.


If people didn't care about compilation speed, why do they use the PIMPL
idiom? I see plenty of people using it for exactly that reason.


--
Michiel Helvensteijn

Return to the comp.compilers page.
Search the comp.compilers archives again.