Blair Zajac wrote:
> Hyrum K. Wright wrote:
>> Mark Phippard wrote:
>>> On 3/16/07, Hyrum K. Wright <firstname.lastname@example.org> wrote:
>>>> I've been mucking around in the JavaHL bindings the past few days,
>>>> and I
>>>> have the following suggestions:
>>>> 1) Update the .cpp and .java files to use the same coding style that
>>>> main codebase uses.
>>> I am not sure it makes sense to force the .java files to follow a style
>>> in C/C++, but to be honest I am not familiar with the scope of what that
>>> would mean. Maybe not much. For me, since I use tools like Eclipse, I
>>> would just want the style to be something reasonable that I can set in
>>> Eclipse and let it handle for me.
>> That's understandable. The .java code is different enough from the .cpp
>> code that it shouldn't matter too much. The problem I'm trying to solve
>> is two-fold: the .cpp files have a style all their own, and it isn't
>> consistent; and, C++ is similar enough to C that when working in them,
>> it would be easier if they just had the same style.
>> Would moving function parameters to individual lines be incongruent with
>> prevailing Java styles?
>>> 2) Rearrange methods within classes in alphabetical order, with older
>>>> methods first in the case of overloaded methods.
>>>> 3) Implement older APIs as wrappers around newer APIs in the same
>>>> (Currently, some are implemented as wrappers around newer APIs in
>>>> other classes.)
>>>> I think this would improve maintainability substantially, but I wanted
>>>> to know what others thought. Comments? Objections?
>>> In general I am fine with the concept. I do not know what #3 means at a
>>> technical level so cannot comment. Obviously the main issue is that the
>>> changes do not make it too hard to backport future fixes, but I would
>>> that the number of those are likely to be a number approaching zero. So
>>> better to do this now, well in advance of a 1.5 branch.
>> Other than giving the compiler/optimizer a bit more information, #3,
>> doesn't have much technical impact. This is mainly limited to the
>> SVNClientSynchronized class, which can implement old APIs using newer
>> ones in the same class, or by calling the new API in SVNClient. It just
>> seems Right to use the API in the same class whenever it is available.
> Can you give an example with the actual method names? I'd like to see
> what you're referring to.
Look at the two remove() methods in SVNClientSynchronized. The old
version calls the new version directly.
Now, look at the two add() methods in SVNClientSynchronized. They both
call the corresponding versions in SVNClient.
Functionally, the two behaviors are the same: they both end up calling
the correct native code. It just seems like there should just be one
call to SVNClient, and that SVNClientSynchronized APIs should wrap each
other, where possible.
Hope this helps.
Received on Fri Mar 16 18:48:32 2007