mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-26 01:22:04 +00:00
202 lines
6.2 KiB
Python
Executable File
202 lines
6.2 KiB
Python
Executable File
import time
|
|
|
|
import pytest
|
|
|
|
from helpers.cluster import ClickHouseCluster
|
|
|
|
cluster = ClickHouseCluster(__file__)
|
|
node = cluster.add_instance(
|
|
"node",
|
|
stay_alive=True,
|
|
main_configs=[
|
|
"configs/config.d/query_log.xml",
|
|
"configs/config.d/schema_cache.xml",
|
|
],
|
|
)
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def start_cluster():
|
|
try:
|
|
cluster.start()
|
|
yield cluster
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
|
|
def check_profile_event_for_query(node, file, profile_event, amount=1):
|
|
node.query("system flush logs")
|
|
query_pattern = f"file('{file}'".replace("'", "\\'")
|
|
assert (
|
|
int(
|
|
node.query(
|
|
f"select ProfileEvents['{profile_event}'] from system.query_log where query like '%{query_pattern}%' and query not like '%ProfileEvents%' and type = 'QueryFinish' order by query_start_time_microseconds desc limit 1"
|
|
)
|
|
)
|
|
== amount
|
|
)
|
|
|
|
|
|
def check_cache_misses(node, file, amount=1):
|
|
check_profile_event_for_query(node, file, "SchemaInferenceCacheMisses", amount)
|
|
|
|
|
|
def check_cache_hits(node, file, amount=1):
|
|
check_profile_event_for_query(node, file, "SchemaInferenceCacheHits", amount)
|
|
|
|
|
|
def check_cache_invalidations(node, file, amount=1):
|
|
check_profile_event_for_query(
|
|
node, file, "SchemaInferenceCacheInvalidations", amount
|
|
)
|
|
|
|
|
|
def check_cache_evictions(node, file, amount=1):
|
|
check_profile_event_for_query(node, file, "SchemaInferenceCacheEvictions", amount)
|
|
|
|
|
|
def check_cache_num_rows_hits(node, file, amount=1):
|
|
check_profile_event_for_query(node, file, "SchemaInferenceCacheNumRowsHits", amount)
|
|
|
|
|
|
def check_cache(node, expected_files):
|
|
sources = node.query("select source from system.schema_inference_cache")
|
|
assert sorted(map(lambda x: x.strip().split("/")[-1], sources.split())) == sorted(
|
|
expected_files
|
|
)
|
|
|
|
|
|
def test(start_cluster):
|
|
node.query("insert into function file('data.jsonl') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
node.query("desc file('data.jsonl')")
|
|
check_cache(node, ["data.jsonl"])
|
|
check_cache_misses(node, "data.jsonl")
|
|
|
|
node.query("desc file('data.jsonl')")
|
|
check_cache_hits(node, "data.jsonl")
|
|
|
|
node.query("insert into function file('data.jsonl') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
node.query("desc file('data.jsonl')")
|
|
check_cache_invalidations(node, "data.jsonl")
|
|
|
|
node.query("insert into function file('data1.jsonl') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
node.query("desc file('data1.jsonl')")
|
|
check_cache(node, ["data.jsonl", "data1.jsonl"])
|
|
check_cache_misses(node, "data1.jsonl")
|
|
|
|
node.query("desc file('data1.jsonl')")
|
|
check_cache_hits(node, "data1.jsonl")
|
|
|
|
node.query("insert into function file('data2.jsonl') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
node.query("desc file('data2.jsonl')")
|
|
check_cache(node, ["data1.jsonl", "data2.jsonl"])
|
|
check_cache_misses(node, "data2.jsonl")
|
|
check_cache_evictions(node, "data2.jsonl")
|
|
|
|
node.query("desc file('data2.jsonl')")
|
|
check_cache_hits(node, "data2.jsonl")
|
|
|
|
node.query("desc file('data1.jsonl')")
|
|
check_cache_hits(node, "data1.jsonl")
|
|
|
|
node.query("desc file('data.jsonl')")
|
|
check_cache(node, ["data.jsonl", "data1.jsonl"])
|
|
check_cache_misses(node, "data.jsonl")
|
|
check_cache_evictions(node, "data.jsonl")
|
|
|
|
node.query("desc file('data2.jsonl')")
|
|
check_cache(node, ["data.jsonl", "data2.jsonl"])
|
|
check_cache_misses(node, "data2.jsonl")
|
|
check_cache_evictions(node, "data2.jsonl")
|
|
|
|
node.query("desc file('data2.jsonl')")
|
|
check_cache_hits(node, "data2.jsonl")
|
|
|
|
node.query("desc file('data.jsonl')")
|
|
check_cache_hits(node, "data.jsonl")
|
|
|
|
node.query("insert into function file('data3.jsonl') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
node.query("desc file('data*.jsonl')")
|
|
check_cache_hits(node, "data*.jsonl")
|
|
|
|
node.query("system drop schema cache for file")
|
|
check_cache(node, [])
|
|
|
|
node.query("desc file('data*.jsonl')")
|
|
check_cache_misses(node, "data*.jsonl", 4)
|
|
|
|
node.query("system drop schema cache")
|
|
check_cache(node, [])
|
|
|
|
node.query("desc file('data*.jsonl')")
|
|
check_cache_misses(node, "data*.jsonl", 4)
|
|
|
|
node.query("system drop schema cache")
|
|
check_cache(node, [])
|
|
|
|
node.query("insert into function file('data.csv') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
res = node.query("select count() from file('data.csv', auto, 'x UInt64')")
|
|
assert int(res) == 100
|
|
check_cache(node, ["data.csv"])
|
|
check_cache_misses(node, "data.csv")
|
|
|
|
res = node.query("select count() from file('data.csv', auto, 'x UInt64')")
|
|
assert int(res) == 100
|
|
check_cache_hits(node, "data.csv")
|
|
|
|
node.query(
|
|
"insert into function file('data.csv', auto, 'x UInt64') select * from numbers(100)"
|
|
)
|
|
time.sleep(1)
|
|
|
|
res = node.query("select count() from file('data.csv', auto, 'x UInt64')")
|
|
assert int(res) == 200
|
|
check_cache_invalidations(node, "data.csv")
|
|
|
|
node.query("insert into function file('data1.csv') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
res = node.query("select count() from file('data1.csv', auto, 'x UInt64')")
|
|
assert int(res) == 100
|
|
check_cache(node, ["data.csv", "data1.csv"])
|
|
check_cache_misses(node, "data1.csv")
|
|
|
|
res = node.query("select count() from file('data1.csv', auto, 'x UInt64')")
|
|
assert int(res) == 100
|
|
check_cache_hits(node, "data1.csv")
|
|
|
|
res = node.query("select count() from file('data*.csv', auto, 'x UInt64')")
|
|
assert int(res) == 300
|
|
check_cache_hits(node, "data*.csv", 2)
|
|
|
|
node.query("system drop schema cache for file")
|
|
check_cache(node, [])
|
|
|
|
res = node.query("select count() from file('data*.csv', auto, 'x UInt64')")
|
|
assert int(res) == 300
|
|
check_cache_misses(node, "data*.csv", 2)
|
|
|
|
node.query("system drop schema cache for file")
|
|
check_cache(node, [])
|
|
|
|
node.query("insert into function file('data.parquet') select * from numbers(100)")
|
|
time.sleep(1)
|
|
|
|
res = node.query("select count() from file('data.parquet')")
|
|
assert int(res) == 100
|
|
check_cache_misses(node, "data.parquet")
|
|
check_cache_hits(node, "data.parquet")
|
|
check_cache_num_rows_hits(node, "data.parquet")
|