El dom, 08-03-2009 a las 09:15 -0700, Adam Winer escribió:
> On Sat, Mar 7, 2009 at 5:56 AM, Santiago Gala <[email protected]> wrote:
> > El vie, 06-03-2009 a las 08:13 -0800, Adam Winer escribió:
> >> What is the advantage of removing trailing whitespace?  I can't see a
> >> reason to do it at all.  Unlike removing unnecessary imports (say), it
> >> doesn't improve the codebase in any objective manner.
> >>
> >
> > The typical reason for removing whitespace is that whitespace changes
> > cause puzzling spurious diffs, where nothing changes but some whitespace
> > change gets committed. So, while it does not improve the codebase viewed
> > statically, useless whitespace are prone to change in time, and those
> > changes pollute the sequence of changesets that make the history of the
> > codebase. Hence the reason why Ian asked for a separate commit for
> > whitespace cleanup.
> 
> What creates spurious diffs isn't the presence of trailing whitespace,
> it's tools that remove trailing whitespace on lines you're not
> directly editing.  Eclipse, IntelliJ, and Emacs can all be configured
> not to do so.  (As best I can tell, they're all configured that way by
> default, too.)
> 

Lots of times it is also created by people inadvertently deleting or
adding invisible spaces, or having files with mix of tabs/spaces
processed automatically by editors. 

> > As for benefits of not having them, the simplest one is that codereview
> > and other diff tools won't flag trailing whitespace and whitespace
> > between tabs in red in ulterior diffs involving lines with them.
> 
> Code review tools should only flag diffs on lines that have been
> modified, so again we're back to "don't remove trailing whitespace on
> lines you're not otherwise touching".
> 

Most if not all diff tools are currently flagging "useless" whitespace
in red everywhere in the context shown, so I think you will need a lot
of evangelism to get this "should only" working.


> > But
> > whitespace cleanups are only worthwhile if new commits are disciplined
> > to not bring them in again.
> 
> Indeed, and a lot of the changes in these whitespace-removal commits
> are just removing "trailing" whitespace which is really just entirely
> blank lines between code.  That's much harder to stop, and IMO not
> worth trying to stop.
> 
> I really don't think we need any whitespace cleanup effort.
> 

I think whitespace cleanup + tight specs for whitespace is useful as a
whole. Quite often it is a nice byproduct of efforts to understand code
or other refactorings. We clearly disagree here. But I'm about to stop
being a mentor for Shindig, due to lack of time + lack of professional
incentive for gadgets, so my opinion should not be weighted too much
here.

Regards
Santiago

> -- Adam
> 
> 
> >
> > Regards
> > Santiago
> >
> >> Cheers,
> >> Adam
> >>
> >> On Fri, Mar 6, 2009 at 3:11 AM, Vincent Siveton <[email protected]> 
> >> wrote:
> >> > The majority of the source code had already removed trailing
> >> > whitespace. Some statistics:
> >> > - around 600 java files
> >> > - only around 140 java files don't removed trailing whitespace
> >> >
> >> > So, IMHO removing trailing whitespace is the way to go.
> >> >
> >> > WDYT?
> >> >
> >> > Cheers,
> >> >
> >> > Vincent
> >> >
> >> > 2009/3/5 Adam Winer <[email protected]>:
> >> >> Instead of 2 commits, could you change it to not remove trailing
> >> >> whitespaces?  It unnecessarily obfuscates history, especially when
> >> >> there's trailing whitespace on non-blank lines.  This is typically an
> >> >> option that can be adjusted in an IDE.
> >> >>
> >> >> -- Adam
> >> >>
> >> >>
> >> >> On Thu, Mar 5, 2009 at 3:21 PM, Vincent Siveton <[email protected]> 
> >> >> wrote:
> >> >>> Hi Adam,
> >> >>>
> >> >>> 2009/3/5 Adam Winer <[email protected]>:
> >> >>>> I see a lot of whitespace changes with each submission, apparently
> >> >>>> around blank lines.  Any way you could avoid making these changes?
> >> >>>> It's much harder to read diffs like this.
> >> >>>
> >> >>> Yeah, my IDE removes trailing whitespaces. I will do 2 commits next 
> >> >>> time.
> >> >>>
> >> >>> Cheers,
> >> >>>
> >> >>> Vincent
> >> >>>
> >> >>
> >> >
> >
> >

Reply via email to