def get(self, table_name, *keys, **kwargs):
num_uuids = self._table_names[table_name]
## A mapping between shards and columns to retrieve from each shard
shards = defaultdict(list)
## Track the keys that we find
found_keys = set()
## Determine all the shards that we need to contact
for key in keys:
assert len(key) == num_uuids
joined_key = join_uuids(*key)
row_names = [self._make_shard_name(table_name, joined_key)]
for row_name in row_names:
shards[row_name].append(joined_key)
## For each shard we contact, get all the keys it may hold
for row_name, columns in shards.iteritems():
try:
for key, val in self.tables[table_name].get(row_name, columns=columns,).iteritems():
key = split_uuids(key)
logger.critical('cassandra get(%r) yielding %r %d' % (table_name, key, len(val)))
## key is a list [(uuid, uuid)]
assert not any(k in found_keys for k in key)
found_keys.add(tuple(key))
yield tuple(key), val
except NotFoundException:
pass
## Raise an exception if we don't retrieve all the keys that were requested
missing_keys = set(keys) - found_keys
for k in missing_keys:
yield k, None
def _get_from_one_row(self, table_name, row_name, columns, start, finish, num_uuids):
logger.debug('c* get: table_name=%r row_name=%r columns=%r start=%r finish=%r' % (
table_name, row_name, columns, start, finish))
if not columns:
assert start is not None and finish is not None
assert start <= finish
num_yielded = 0
while True:
## if we have
prev_start = start
logger.debug('cassandra get(%r...)' % row_name)
## if table_name == 'inbound':
## import ipdb
## ipdb.set_trace()
for key, val in self.tables[table_name].get(
row_name,
columns=columns,
column_start=start,
column_finish=finish,
column_count=1,
).iteritems():
key = split_uuids(key)
logger.critical('cassandra get(%r) yielding %r %d' % (table_name, key, len(val)))
yield key, val
num_yielded += 1
logger.debug('c* get: table_name=%r row_name=%r columns=%r start=%r finish=%r' % (
table_name, row_name, columns, start, finish))
## prepare to page ahead to next batch
if columns:
break
start = list(key)
start[-1] = uuid.UUID(int=key[-1].int+1)
assert len(start) == num_uuids
start = join_uuids(*start)
if start == prev_start or start > finish:
break
logger.debug('paging forward from %r to %r' % (prev_start, start))
## We need to raise a not found exception if the caller asked for
## a specific column and we didn't yield any results
if not columns and num_yielded == 0:
raise pycassa.NotFoundException
