def test_query_formatter(
query: Union[ProcessableQuery, CompositeQuery[Entity]],
formatted: TExpression,
) -> None:
formatted_query = format_query(query) # type: ignore
assert formatted_query == formatted
# make sure there are no empty lines
assert [line for line in formatted_query if not line] == []
def __repr__(self) -> str:
from snuba.query.formatters.tracing import format_query
# NOTE (Vlad): Why the type is cast:
# If you remove the ignore type comment you will get the following error:
#
# Argument 1 to "format_query" has incompatible type
# "ProcessableQuery[TSimpleDataSource]"; expected
# "Union[ProcessableQuery[SimpleDataSource], CompositeQuery[SimpleDataSource]]"
#
# This happens because self in this case is a generic type
# CompositeQuery[TSimpleDataSource] while the function format_query takes a
# SimpleDataSource (a concrete type). It is known by us (and mypy) that
# TSimpleDataSource is bound to SimpleDataSource, which means that all types
# parametrizing this class must be subtypes of SimpleDataSource, mypy is not smart
# enough to know that though and so in order to have a generic repr function
# I cast the type check in this case.
# Making TSimpleDataSource covariant would almost work except that covariant types
# canot be used as parameters: https://github.com/python/mypy/issues/7049
return "\n".join(format_query(cast(CompositeQuery[SimpleDataSource], self)))
def test_query_formatter(query: Union[LogicalQuery, CompositeQuery[Entity]],
formatted: TExpression) -> None:
assert format_query(query) == formatted
def callback_func(
storage: str,
query: Query,
request_settings: RequestSettings,
referrer: str,
results: List[Result[QueryResult]],
) -> None:
cache_hit = False
is_duplicate = False
# Captures if any of the queries involved was a cache hit or duplicate, as cache
# hits may a cause of inconsistency between results.
# Doesn't attempt to distinguish between all of the specific scenarios (one or both
# queries, or splits of those queries could have hit the cache).
if any([result.result.extra["stats"].get("cache_hit", 0) for result in results]):
cache_hit = True
elif any(
[result.result.extra["stats"].get("is_duplicate", 0) for result in results]
):
is_duplicate = True
consistent = request_settings.get_consistent()
if not results:
metrics.increment(
"query_result",
tags={"storage": storage, "match": "empty", "referrer": referrer},
)
return
primary_result = results.pop(0)
primary_result_data = primary_result.result.result["data"]
for result in results:
result_data = result.result.result["data"]
metrics.timing(
"diff_ms",
round((result.execution_time - primary_result.execution_time) * 1000),
tags={
"referrer": referrer,
"cache_hit": str(cache_hit),
"is_duplicate": str(is_duplicate),
"consistent": str(consistent),
},
)
# Do not bother diffing the actual results of sampled queries
if request_settings.get_turbo() or query.get_sample() not in [None, 1.0]:
return
if result_data == primary_result_data:
metrics.increment(
"query_result",
tags={
"storage": storage,
"match": "true",
"referrer": referrer,
"cache_hit": str(cache_hit),
"is_duplicate": str(is_duplicate),
"consistent": str(consistent),
},
)
else:
# Do not log cache hits to Sentry as it creates too much noise
if cache_hit:
continue
reason = assign_reason_category(result_data, primary_result_data, referrer)
metrics.increment(
"query_result",
tags={
"storage": storage,
"match": "false",
"referrer": referrer,
"reason": reason,
"cache_hit": str(cache_hit),
"is_duplicate": str(is_duplicate),
"consistent": str(consistent),
},
)
if len(result_data) != len(primary_result_data):
sentry_sdk.capture_message(
f"Non matching {storage} result - different length",
level="warning",
tags={
"referrer": referrer,
"storage": storage,
"reason": reason,
"cache_hit": str(cache_hit),
"is_duplicate": str(is_duplicate),
"consistent": str(consistent),
},
extras={
"query": format_query(query),
"primary_result": len(primary_result_data),
"other_result": len(result_data),
},
)
break
# Avoid sending too much data to Sentry - just one row for now
for idx in range(len(result_data)):
if result_data[idx] != primary_result_data[idx]:
sentry_sdk.capture_message(
"Non matching result - different result",
level="warning",
tags={
"referrer": referrer,
"storage": storage,
"reason": reason,
"cache_hit": str(cache_hit),
"is_duplicate": str(is_duplicate),
"consistent": str(consistent),
},
extras={
"query": format_query(query),
"primary_result": primary_result_data[idx],
"other_result": result_data[idx],
},
)
break
def callback_func(storage: str, query: Query, referrer: str,
results: List[Result[QueryResult]]) -> None:
if not results:
metrics.increment(
"query_result",
tags={
"storage": storage,
"match": "empty",
"referrer": referrer
},
)
return
primary_result = results.pop(0)
primary_result_data = primary_result.result.result["data"]
for result in results:
result_data = result.result.result["data"]
metrics.timing(
"diff_ms",
round((result.execution_time - primary_result.execution_time) *
1000),
tags={"referrer": referrer},
)
if result_data == primary_result_data:
metrics.increment(
"query_result",
tags={
"storage": storage,
"match": "true",
"referrer": referrer
},
)
else:
metrics.increment(
"query_result",
tags={
"storage": storage,
"match": "false",
"referrer": referrer
},
)
if len(result_data) != len(primary_result_data):
sentry_sdk.capture_message(
f"Non matching {storage} result - different length",
level="warning",
tags={
"referrer": referrer,
"storage": storage
},
extras={
"query": format_query(query),
"primary_result": len(primary_result_data),
"other_result": len(result_data),
},
)
break
# Avoid sending too much data to Sentry - just one row for now
for idx in range(len(result_data)):
if result_data[idx] != primary_result_data[idx]:
sentry_sdk.capture_message(
"Non matching result - different result",
level="warning",
tags={
"referrer": referrer,
"storage": storage
},
extras={
"query": format_query(query),
"primary_result": primary_result_data[idx],
"other_result": result_data[idx],
},
)
break