[PATCH 2 of 3] move store walking from streamclone to filelog

Adrian Buehlmann adrian at cadifra.com
Thu Jul 17 13:09:54 CDT 2008


On 17.07.2008 18:24, Matt Mackall wrote:
> On Thu, 2008-07-17 at 16:23 +0200, Adrian Buehlmann wrote:
>> # HG changeset patch
>> # User Adrian Buehlmann <adrian at cadifra.com>
>> # Date 1216299442 -7200
>> # Node ID d29d8f832d09a8570a25a68c964baf8572454b75
>> # Parent  64d51f7facf9804d3b004f4eae7844ae7a1937ca
>> move store walking from streamclone to filelog
>>
>> * add method localrepo.localrepository.storefiles
>> * add function filelog.store
>> * add class filelog._store
> 
> This could all be a lot simpler.

If I knew how, I would try doing so.

>> +def _dirwalk(path, recurse, strip_count, decodefn):
>> +    '''yields (*.d or *.i filename, size)'''
>> +    for e, kind, st in osutil.listdir(path, stat=True):
>> +        pe = os.path.join(path, e)
>> +        if kind == stat.S_IFDIR:
>> +            if recurse:
>> +                for x in _dirwalk(pe, True, strip_count, decodefn):
>> +                    yield x
>> +        else:
>> +            if kind != stat.S_IFREG or len(e) < 2:
>> +                continue
>> +            sfx = e[-2:]
>> +            if sfx in ('.d', '.i'):
>> +                f = decodefn(util.pconvert(pe[strip_count:]))
>> +                yield (f, st.st_size)
> 
> This function is too smart. Decoding, stripping and filtering can all be
> done at a higher level.

This is nearly verbatim moved from streamclone. I wouldn't have
expected to rewrite streamclone.

>> +def store(requirements, spath, createmode):
>> +    efn = encodefn(requirements)
>> +    if 'store' not in requirements:
>> +        return _store(efn, lambda x: x, spath, createmode)
>> +    else:
>> +        return _store(efn, decodefilename, spath, createmode)
> 
> Layering violation: filelog's not supposed to know about "requirements".

Could you please tell me how to do it instead?

>> +class _store:
> 
> If you're going to use a class, you should use inheritance to derive the
> variations of your class. 

I'm lost here. I think I did use inheritence? _store is a base
class.

>> +    def __init__(self, encodefn, decodefn, spath, createmode):
>> +        self.spath = spath
>> +        self.encodefn = encodefn
>> +        self.__decodefn = decodefn
>> +        so = util.opener(self.spath)
>> +        so.createmode = createmode
>> +        self.opener = encodedopener(so, self.encodefn)
> 
> For instance, if you used inheritance, you could simply use a normal
> opener here.

I did use inheritance. fnlogstore (patch 3) is derived from _store.

>> +    def _walkfiledata(self):
>> +        strip_count = len(self.spath) + len(os.sep)
>> +        dpath = os.path.join(self.spath, 'data')
>> +        for x in _dirwalk(dpath, True, strip_count, self.__decodefn):
>> +            yield x
>> +
>> +    def walk(self):
>> +        '''yields: direncoded filename, size of file'''
>> +        # yield file data first
>> +        for x in self._walkfiledata():
>> +            yield x
>> +        # yield manifest before changelog
> 
> Layering violation: filelog.py is only concerned with what lives inside
> data/. Perhaps we want a store.py.

What should be in store.py?

>> +    def storefiles(self):
>> +        '''get all *.i and *.d files in the store
>> +
>> +        Returns (ok, list of (filename, size), total_bytes)'''
>> +
>> +        # get consistent snapshot of repo, lock during scan
>> +        repolock = None
> 
> We usually just call this lock.

This code was moved nearly verbatim from streamclone.py. Of course
I have no problem with renaming anything.

>> +    def storefiles(self, getsizes):
>> +        return (False, [], 0)
> 
> Is this actually necessary? We could just test whether the repo is
> local, no?

I was not sure here. I think it *is* already tested that the repo
is not local. I will try to see if this can be removed.

Some general comments:

If I still happen to have the energy to continue trying.
Currently, I'm a bit tired.

I just couldn't believe that the aux and filename length problem can't
be solved. I still think it should be solvable. But maybe not by me.



More information about the Mercurial-devel mailing list