Don't add a separate class to iterate history

This commit is contained in:
Chase Sterling 2016-07-07 17:30:49 -04:00
parent 74452fe58f
commit 35471d3c3f

View file

@ -96,7 +96,7 @@ class WikiPage(object):
if cached: if cached:
return cached return cached
info = self.get_history(limit=1)[0] info = next(self.history)
cache.set(cache_key, info) cache.set(cache_key, info)
return info return info
@ -104,10 +104,54 @@ class WikiPage(object):
def history(self): def history(self):
"""Get page history. """Get page history.
:return: list -- List of dicts History can take a long time to generate for repositories with many commits.
This returns an iterator to avoid having to load them all at once.
:return: iter -- Iterator over dicts
""" """
return PageHistory(self, self._cache_key('history')) cache_complete = False
cached_revs = cache.get(self._cache_key('history')) or []
start_sha = None
if cached_revs:
if cached_revs[-1] == 'TAIL':
del cached_revs[-1]
cache_complete = True
else:
start_sha = cached_revs[-1]['sha']
for rev in cached_revs:
yield rev
if cache_complete:
return
if not len(self.wiki.repo.open_index()):
# Index is empty, no commits
return
walker = iter(self.wiki.repo.get_walker(paths=[self.filename], include=start_sha, follow=True))
if start_sha:
# If we are not starting from HEAD, we already have the start commit
next(walker)
filename = self.filename
try:
for entry in walker:
change_type = None
for change in entry.changes():
if change.new.path == filename:
filename = change.old.path
change_type = change.type
break
author_name, author_email = entry.commit.author.rstrip('>').split('<')
r = dict(author=author_name.strip(),
author_email=author_email,
time=entry.commit.author_time,
message=entry.commit.message,
sha=entry.commit.id,
type=change_type)
cached_revs.append(r)
yield r
cached_revs.append('TAIL')
finally:
cache.set(self._cache_key('history'), cached_revs)
@property @property
def partials(self): def partials(self):
@ -298,75 +342,3 @@ class WikiPage(object):
# We'll get a KeyError if self.sha isn't in the repo, or if self.filename isn't in the tree of our commit # We'll get a KeyError if self.sha isn't in the repo, or if self.filename isn't in the tree of our commit
return False return False
return True return True
class PageHistory(collections.Sequence):
"""Acts like a list, but dynamically loads and caches history revisions as requested."""
def __init__(self, page, cache_key):
self.page = page
self.cache_key = cache_key
self._store = cache.get(cache_key) or []
if not self._store:
self._iter_rest = self._get_rest()
elif self._store[-1] == 'TAIL':
self._iter_rest = None
else:
self._iter_rest = self._get_rest(self._store[-1]['sha'])
def __iter__(self):
# Iterate over the revisions already cached
for r in self._store:
if r == 'TAIL':
return
yield r
# Iterate over the revisions yet to be discovered
if self._iter_rest:
try:
for r in self._iter_rest:
self._store.append(r)
yield r
self._store.append('TAIL')
finally:
# Make sure we cache the results whether or not the iteration was completed
cache.set(self.cache_key, self._store)
def _get_rest(self, start_sha=None):
if not len(self.page.wiki.repo.open_index()):
# Index is empty, no commits
return
walker = iter(self.page.wiki.repo.get_walker(paths=[self.page.filename], include=start_sha, follow=True))
if start_sha:
# If we are not starting from HEAD, we already have the start commit
print(next(walker))
filename = self.page.filename
for entry in walker:
change_type = None
for change in entry.changes():
if change.new.path == filename:
filename = change.old.path
change_type = change.type
break
author_name, author_email = entry.commit.author.rstrip('>').split('<')
r = dict(author=author_name.strip(),
author_email=author_email,
time=entry.commit.author_time,
message=entry.commit.message,
sha=entry.commit.id,
type=change_type)
yield r
def __getitem__(self, index):
if isinstance(index, slice):
return list(itertools.islice(self, index.start, index.stop, index.step))
else:
try:
return next(itertools.islice(self, index, index+1))
except StopIteration:
raise IndexError
def __len__(self):
if not self._store or self._store[-1] != 'TAIL':
# Force generation of all revisions
list(self)
return len(self._store) - 1 # Don't include the TAIL sentinel