def do():
yield from asyncio.sleep(0.0)
nonlocal mode
pinfo = {
"category": INSPECTOR_CATEGORY,
"source": "%s" % repr(data_provider),
"step": "",
"description": ""
}
# register begin of inspection (differ slightly depending on type)
if data_provider_type == "source":
registerer_obj.register_status("inspecting",
subkey="inspect")
elif data_provider_type == "build":
registerer_obj.register_status("inspecting",
transient=True,
init=True,
job={"step": "inspect"})
self.logger.info(
"Running inspector on %s (type:%s,data_provider:%s)" %
(repr(data_provider), data_provider_type,
backend_provider))
if sample is not None:
self.logger.info(
"Sample set to %s, inspect only a subset of data",
sample)
if limit is None:
self.logger.info("Inspecting all the documents")
else:
nonlocal batch_size
# adjust batch_size so we inspect only "limit" docs if batch is smaller than the limit
if batch_size > limit:
batch_size = limit
self.logger.info("Inspecting only %s documents", limit)
# make it pickleable
if data_provider_type == "source":
# because register_obj is also used to fetch data, it has to be unprepare() for pickling
registerer_obj.unprepare()
else:
# NOTE: do not unprepare() the builder, we'll loose the target name
# (it's be randomly generated again) and we won't be able to register results
pass
cnt = 0
doccnt = 0
jobs = []
# normalize mode param and prepare global results
if type(mode) == str:
mode = [mode]
converters, mode = btinspect.get_converters(mode)
inspected = {}
for m in mode:
inspected.setdefault(m, {})
backend = create_backend(backend_provider).target_collection
for ids in id_feeder(backend, batch_size=batch_size):
if sample is not None:
if random.random() > sample:
continue
cnt += 1
doccnt += batch_size
if limit and doccnt > limit:
break
pinfo["description"] = "batch #%s" % cnt
def batch_inspected(bnum, i, f):
nonlocal inspected
nonlocal got_error
nonlocal mode
try:
res = f.result()
for m in mode:
inspected[m] = btinspect.merge_record(
inspected[m], res[m], m)
except Exception as e:
got_error = e
self.logger.error(
"Error while inspecting data from batch #%s: %s"
% (bnum, e))
raise
pre_mapping = "mapping" in mode # we want to generate intermediate mapping so we can merge
# all maps later and then generate the ES mapping from there
self.logger.info("Creating inspect worker for batch #%s" %
cnt)
job = yield from self.job_manager.defer_to_process(
pinfo,
partial(inspect_data,
backend_provider,
ids,
mode=mode,
pre_mapping=pre_mapping,
**kwargs))
job.add_done_callback(partial(batch_inspected, cnt, ids))
jobs.append(job)
yield from asyncio.gather(*jobs)
# compute metadata (they were skipped before)
for m in mode:
if m == "mapping":
try:
inspected["mapping"] = es.generate_es_mapping(
inspected["mapping"])
# metadata for mapping only once generated
inspected = btinspect.compute_metadata(
inspected, m)
except es.MappingError as e:
inspected["mapping"] = {
"pre-mapping": inspected["mapping"],
"errors": e.args[1]
}
else:
inspected = btinspect.compute_metadata(inspected, m)
# just potential converters
btinspect.run_converters(inspected, converters)
def fully_inspected(res):
nonlocal got_error
try:
res = btinspect.stringify_inspect_doc(res)
_map = {"results": res}
_map["data_provider"] = repr(data_provider)
_map["started_at"] = started_at
_map["duration"] = timesofar(t0)
# when inspecting with "stats" mode, we can get huge number but mongo
# can't store more than 2^64, make sure to get rid of big nums there
def clean_big_nums(k, v):
# TODO: same with float/double? seems mongo handles more there ?
if isinstance(v, int) and v > 2**64:
return k, math.nan
else:
return k, v
dict_traverse(_map, clean_big_nums)
# register begin of inspection (differ slightly depending on type)
if "mapping" in mode and "errors" in res[
"mapping"] and "pre-mapping" in res["mapping"]:
registerer_obj.register_status("failed",
subkey="inspect",
inspect=_map)
got_error = InspectorError(
res["mapping"]["errors"])
else:
if data_provider_type == "source":
registerer_obj.register_status(
"success", subkey="inspect", inspect=_map)
elif data_provider_type == "build":
registerer_obj.register_status(
"success",
job={"step": "inspect"},
build={"inspect": _map})
except Exception as e:
self.logger.exception(
"Error while inspecting data: %s" % e)
got_error = e
if data_provider_type == "source":
registerer_obj.register_status("failed",
subkey="inspect",
err=repr(e))
elif data_provider_type == "build":
registerer_obj.register_status(
"failed", job={"err": repr(e)})
fully_inspected(inspected)
if data_provider_type is None:
return
if got_error:
raise got_error
def do():
yield from asyncio.sleep(0.0)
nonlocal mode
pinfo = {
"category": INSPECTOR_CATEGORY,
"source": "%s" % repr(data_provider),
"step": "",
"description": ""
}
# register begin of inspection (differ slightly depending on type)
if data_provider_type == "source":
registerer_obj.register_status("inspecting",
subkey="inspect")
elif data_provider_type == "build":
registerer_obj.register_status("inspecting",
transient=True,
init=True,
job={"step": "inspect"})
self.logger.info("Running inspector on %s (type:%s,data_provider:%s)" % \
(repr(data_provider),data_provider_type,backend_provider))
# make it pickleable
if data_provider_type == "source":
# because register_obj is also used to fetch data, it has to be unprepare() for pickling
registerer_obj.unprepare()
else:
# NOTE: do not unprepare() the builder, we'll loose the target name
# (it's be randomly generated again) and we won't be able to register results
pass
cnt = 0
jobs = []
# normalize mode param and prepare global results
if type(mode) == str:
mode = [mode]
inspected = {}
for m in mode:
inspected.setdefault(m, {})
backend = create_backend(backend_provider).target_collection
for ids in id_feeder(backend, batch_size=batch_size):
cnt += 1
pinfo["description"] = "batch #%s" % cnt
def batch_inspected(bnum, i, f):
nonlocal inspected
nonlocal got_error
nonlocal mode
try:
res = f.result()
for m in mode:
inspected[m] = btinspect.merge_record(
inspected[m], res[m], m)
except Exception as e:
got_error = e
self.logger.error(
"Error while inspecting data from batch #%s: %s"
% (bnum, e))
raise
pre_mapping = "mapping" in mode # we want to generate intermediate mapping so we can merge
# all maps later and then generate the ES mapping from there
self.logger.info("Creating inspect worker for batch #%s" %
cnt)
job = yield from self.job_manager.defer_to_process(
pinfo,
partial(inspect_data,
backend_provider,
ids,
mode=mode,
pre_mapping=pre_mapping,
**kwargs))
job.add_done_callback(partial(batch_inspected, cnt, ids))
jobs.append(job)
yield from asyncio.gather(*jobs)
# compute metadata (they were skipped before)
for m in mode:
if m == "mapping":
try:
inspected["mapping"] = es.generate_es_mapping(
inspected["mapping"])
# metadata for mapping only once generated
inspected = btinspect.compute_metadata(
inspected, m)
except es.MappingError as e:
inspected["mapping"] = {
"pre-mapping": inspected["mapping"],
"errors": e.args[1]
}
else:
inspected = btinspect.compute_metadata(inspected, m)
def fully_inspected(res):
nonlocal got_error
try:
res = btinspect.stringify_inspect_doc(res)
_map = {"results": res}
_map["data_provider"] = repr(data_provider)
_map["started_at"] = started_at
_map["duration"] = timesofar(t0)
# register begin of inspection (differ slightly depending on type)
if "mapping" in mode and "errors" in res[
"mapping"] and "pre-mapping" in res["mapping"]:
registerer_obj.register_status("failed",
subkey="inspect",
inspect=_map)
got_error = InspectorError(
res["mapping"]["errors"])
else:
if data_provider_type == "source":
registerer_obj.register_status(
"success", subkey="inspect", inspect=_map)
elif data_provider_type == "build":
registerer_obj.register_status(
"success",
job={"step": "inspect"},
build={"inspect": _map})
except Exception as e:
self.logger.exception(
"Error while inspecting data: %s" % e)
got_error = e
if data_provider_type == "source":
registerer_obj.register_status("failed",
subkey="inspect",
err=repr(e))
elif data_provider_type == "build":
registerer_obj.register_status(
"failed", job={"err": repr(e)})
fully_inspected(inspected)
if data_provider_type is None:
return
if got_error:
raise got_error