Emacs integration - an overview

[Dan Christensen]

Dan Christensen <jdc at uwo.ca> writes:

> Bryan O'Sullivan <bos at serpentine.com> writes:
>
>> On Wed, 2005-08-31 at 14:18 -0400, Dan Christensen wrote:
>>
>>> But it happens because python's sys.getcwd returns the canonical path,
>>> after expanding symlinks.  Maybe the solution is for hg to use
>>> os.path.realpath on all filenames?
>>
>> I don't think so.  It's pretty expensive.
>
> I think it would only be needed for absolute filenames specified on
> the command line, right?  So the speed should matter too much.

I think the only change required would be to this function in util.py:

def canonpath(root, cwd, myname):
    """return the canonical path of myname, given cwd and root"""
    rootsep = root + os.sep
    name = myname
-   if not name.startswith(os.sep):
-       name = os.path.join(root, cwd, name)
    name = os.path.normpath(name)
    if name.startswith(rootsep):
        return pconvert(name[len(rootsep):])
    elif name == root:
        return ''
    else:
        raise Abort('%s not under root' % myname)

The marked lines would be replaced with:

+   if name.startswith(os.sep):
+       name = os.path.realpath(name)
+   else:
+       name = os.path.join(root, cwd, name)

I'll give this a try locally.  Can someone tell me what operations I
should time to see if this slows anything down?  As far as I can see,
canonpath is only called for things specified on the command line,
but I'm not 100% sure.

(There's a chance that root needs realpath called on it, but I doubt
it.  I would guess that it is obtained by starting with getcwd and
going upwards, and getcwd returns a result with no symlinks.)

realpath might not work on windows, in which case a dummy wrapper
should be provided in the appropriate section of util.py.  Can someone
with a Windows machine test this for me?

...

By the way, under Windows, is root stored with '/' or '\'?  The above
code seems to assume that root is stored with '\', but then returns a
relative path containing '/'.  Is this correct?

Dan