def _diff_doc_worker(args):
# b1_target_collection, b2_es_index, ids, _path = args
_b1, _b2, ids, _path = args
import sys
if _path not in sys.path:
sys.path.append(_path)
import utils.diff
reload(utils.diff)
from utils.diff import _diff_doc_inner_worker, get_backend
b1 = get_backend(*_b1)
b2 = get_backend(*_b2)
_updates = _diff_doc_inner_worker(b1, b2, ids)
return _updates
def _diff_parallel_worker(old_collection_name, new_collection_name,
common_ids):
b1 = get_backend(old_collection_name, 'mongodb')
b2 = get_backend(new_collection_name, 'mongodb')
_updates = []
for doc1, doc2 in two_docs_iterator(b1, b2, common_ids):
assert doc1['_id'] == doc2['_id'], repr((common_ids, len(common_ids)))
_patch = jsondiff.make(doc1, doc2)
if _patch:
_diff = {}
_diff['patch'] = _patch
_diff['_id'] = doc1['_id']
_updates.append(_diff)
return _updates
def _diff_doc_worker(args):
#b1_target_collection, b2_es_index, ids, _path = args
_b1, _b2, ids, _path = args
import sys
if _path not in sys.path:
sys.path.append(_path)
import utils.diff
reload(utils.diff)
from utils.diff import _diff_doc_inner_worker, get_backend
b1 = get_backend(*_b1)
b2 = get_backend(*_b2)
_updates = _diff_doc_inner_worker(b1, b2, ids)
return _updates
def main(self, index, collection, diff_filepath, validate=False, wait=60):
self._index = index
self._esi._index = index
diff = loadobj(diff_filepath)
source_collection = diff['source']
add_list = self.add(source_collection, diff['add'])
delete_list = self.delete(collection, diff['delete'])
update_list = self.update(diff['update'])
t00 = time()
print('Adding new {} docs...'.format(len(diff['add'])))
t0 = time()
bulk(self._es, add_list)
print("Done. [{}]".format(timesofar(t0)))
print('Deleting {} docs'.format(len(diff['delete'])))
t0 = time()
bulk(self._es, delete_list)
print("Done. [{}]".format(timesofar(t0)))
print('Updating {} docs'.format(len(diff['update'])))
t0 = time()
bulk(self._es, update_list)
print("Done. [{}]".format(timesofar(t0)))
print("=" * 20)
print("Finished! [{}]".format(timesofar(t00)))
if validate:
print('Waiting {}s to let ES to finish...'.format(wait), end="")
sleep(wait)
print("Done.")
print("Validating...")
t0 = time()
q = {
"query": {
"constant_score": {
"filter": {
"exists": {
"field": collection
}
}
}
}
}
data = self._esi.doc_feeder(query=q, _source=collection)
temp_collection = collection + '_temp_' + get_random_string()
self._src[temp_collection].drop()
load_source(temp_collection, src_data=data)
c1 = get_backend(source_collection, 'mongodb')
c2 = get_backend(temp_collection, 'mongodb')
diff_result = diff_collections(c1, c2, use_parallel=False)
self._src[temp_collection].drop()
print("Done. [{}]".format(t0))
return diff_result
def main(self, index, collection, diff_filepath, validate=False, wait=60):
self._index = index
self._esi._index = index
diff = loadobj(diff_filepath)
source_collection = diff['source']
add_list = self.add(source_collection, diff['add'])
delete_list = self.delete(collection, diff['delete'])
update_list = self.update(diff['update'])
t00 = time()
print('Adding new {} docs...'.format(len(diff['add'])))
t0 = time()
bulk(self._es, add_list)
print("Done. [{}]".format(timesofar(t0)))
print('Deleting {} docs'.format(len(diff['delete'])))
t0 = time()
bulk(self._es, delete_list)
print("Done. [{}]".format(timesofar(t0)))
print('Updating {} docs'.format(len(diff['update'])))
t0 = time()
bulk(self._es, update_list)
print("Done. [{}]".format(timesofar(t0)))
print("="*20)
print("Finished! [{}]".format(timesofar(t00)))
if validate:
print('Waiting {}s to let ES to finish...'.format(wait), end="")
sleep(wait)
print("Done.")
print("Validating...")
t0 = time()
q = {
"query": {
"constant_score": {
"filter": {
"exists": {
"field": collection
}
}
}
}
}
data = self._esi.doc_feeder(query=q, _source=collection)
temp_collection = collection + '_temp_' + get_random_string()
self._src[temp_collection].drop()
load_source(temp_collection, src_data=data)
c1 = get_backend(source_collection, 'mongodb')
c2 = get_backend(temp_collection, 'mongodb')
diff_result = diff_collections(c1, c2, use_parallel=False)
self._src[temp_collection].drop()
print("Done. [{}]".format(t0))
return diff_result
def sync_from_one_diff(index, collection, diff_filepath, validate=False, wait=60, dryrun=False, returncnt=False, save2file=None):
sync = ESSyncer(index=index)
#sync._index = index
#sync._esi._index = index
diff = loadobj(diff_filepath)
source_collection = diff['source']
add_iter = sync.add(source_collection, diff['add'])
delete_iter = sync.delete(collection, diff['delete'])
update_iter = sync.update2(diff['update'], collection, source_collection)
t00 = time()
if save2file:
from itertools import chain
import json
for op in chain(add_iter, delete_iter, update_iter):
json.dump(op, save2file)
print("="*20)
print("Finished! [{}]".format(timesofar(t00)))
return
print('Adding new {} docs...'.format(len(diff['add'])))
t0 = time()
if not dryrun:
try:
bulk(sync._es, add_iter)
except:
pass
print("Done. [{}]".format(timesofar(t0)))
print('Deleting {} docs'.format(len(diff['delete'])))
t0 = time()
if not dryrun:
bulk(sync._es, delete_iter)
print("Done. [{}]".format(timesofar(t0)))
print('Updating {} docs'.format(len(diff['update'])))
t0 = time()
if not dryrun:
bulk(sync._es, update_iter)
print("Done. [{}]".format(timesofar(t0)))
# add flush and refresh
try:
res = sync._es.indices.flush()
print("Flushing...", res)
res = sync._es.indices.refresh()
print("Refreshing...", res)
except:
pass
print("="*20)
print("Finished! [{}]".format(timesofar(t00)))
if returncnt:
cnt = {
'add': len(diff['add']),
'delete': len(diff['delete']),
'update': len(diff['update'])
}
return cnt
if validate:
print('Waiting {}s to let ES to finish...'.format(wait), end="")
sleep(wait)
print("Done.")
print("Validating...")
t0 = time()
q = {
"query": {
"constant_score": {
"filter": {
"exists": {
"field": 'clinvar'
}
}
}
}
}
data = sync._esi.doc_feeder(query=q, _source=collection)
temp_collection = collection + '_temp_' + get_random_string()
sync._src[temp_collection].drop()
load_source(temp_collection, src_data=data)
c1 = get_backend(source_collection, 'mongodb')
c2 = get_backend(temp_collection, 'mongodb')
diff_result = diff_collections(c1, c2, use_parallel=False)
sync._src[temp_collection].drop()
print("Done. [{}]".format(t0))
return diff_result