Mercurial > evolve
changeset 2329:3bb8382fc4cb
obscache: update some documentation
Update some data and do some proof reading.
| author | Pierre-Yves David <pierre-yves.david@octobus.net> |
|---|---|
| date | Wed, 03 May 2017 23:07:01 +0200 |
| parents | 7ccacaa38782 |
| children | d72c8c1f09e2 |
| files | hgext3rd/evolve/obscache.py |
| diffstat | 1 files changed, 16 insertions(+), 15 deletions(-) [+] |
line wrap: on
line diff
--- a/hgext3rd/evolve/obscache.py Wed May 03 21:47:06 2017 +0200 +++ b/hgext3rd/evolve/obscache.py Wed May 03 23:07:01 2017 +0200 @@ -298,13 +298,14 @@ """update the cache with new revisions Newly added changeset might be affected by obsolescence markers - we already have locally. So we needs to have soem global - knowledge about the markers to handle that question. Right this - requires parsing all markers in the obsstore. However, we could - imagine using various optimisation (eg: bloom filter, other on - disk cache) to remove this full parsing. + we already have locally. So we needs to have some global + knowledge about the markers to handle that question. - For now we stick to the simpler approach or paying the + Right now this requires parsing all markers in the obsstore. We could + imagine using various optimisation (eg: another cache, network + exchange, etc). + + For now we stick to the simpler approach of paying the performance cost on new changesets. """ node = repo.changelog.node @@ -363,7 +364,7 @@ if notpublic: obscache = repo.obsstore.obscache # Since we warm the cache at the end of every transaction, the cache - # should be up to date. However a non-enabled client might have touced + # should be up to date. However a non-enabled client might have touched # the repository. # # Updating the cache without a lock is sloppy, so we fallback to the @@ -372,7 +373,7 @@ # # With the current implementation updating the cache will requires to # load the obsstore anyway. Once loaded, hitting the obsstore directly - # will be about as fast.. + # will be about as fast... if not obscache.uptodate(repo): if repo.currenttransaction() is None: repo.ui.log('evoext-obscache', @@ -381,8 +382,9 @@ repo.ui.debug('obscache is out of date') return orig(repo) else: - # If a transaction is open, it is worthwhile to update and use the - # cache as it will be written on disk when the transaction close. + # If a transaction is open, it is worthwhile to update and use + # the cache, the lock prevent race and it will be written on + # disk when the transaction close. obscache.update(repo) isobs = obscache.get for r in notpublic: @@ -415,11 +417,10 @@ if repo is None: return repo = repo.unfiltered() - # As pointed in 'obscache.update', we could have the - # changelog and the obsstore in charge of updating the - # cache when new items goes it. The tranaction logic would - # then only be involved for the 'pending' and final saving - # logic. + # As pointed in 'obscache.update', we could have the changelog + # and the obsstore in charge of updating the cache when new + # items goes it. The tranaction logic would then only be + # involved for the 'pending' and final writing on disk. self.obsstore.obscache.update(repo) self.obsstore.obscache.save(repo)