mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-28 10:31:57 +00:00
133 lines
4.8 KiB
Python
133 lines
4.8 KiB
Python
import logging
|
|
import os
|
|
|
|
import time
|
|
import pytest
|
|
from helpers.cluster import ClickHouseCluster
|
|
from helpers.test_tools import TSV
|
|
|
|
logging.getLogger().setLevel(logging.INFO)
|
|
logging.getLogger().addHandler(logging.StreamHandler())
|
|
|
|
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def started_cluster():
|
|
try:
|
|
cluster = ClickHouseCluster(__file__)
|
|
cluster.add_instance('h0_0_0', main_configs=['configs/config.xml'], extra_configs=[ 'configs/hdfs-site.xml'], with_hive=True)
|
|
|
|
logging.info("Starting cluster ...")
|
|
cluster.start()
|
|
yield cluster
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
def test_create_parquet_table(started_cluster):
|
|
logging.info('Start testing creating hive table ...')
|
|
node = started_cluster.instances['h0_0_0']
|
|
node.query("set input_format_parquet_allow_missing_columns = true")
|
|
result = node.query("""
|
|
DROP TABLE IF EXISTS default.demo_parquet;
|
|
CREATE TABLE default.demo_parquet (`id` Nullable(String), `score` Nullable(Int32), `day` Nullable(String)) ENGINE = Hive('thrift://hivetest:9083', 'test', 'demo') PARTITION BY(day)
|
|
""")
|
|
logging.info("create result {}".format(result))
|
|
time.sleep(120)
|
|
assert result.strip() == ''
|
|
|
|
def test_create_orc_table(started_cluster):
|
|
logging.info('Start testing creating hive table ...')
|
|
node = started_cluster.instances['h0_0_0']
|
|
result = node.query("""
|
|
DROP TABLE IF EXISTS default.demo_orc;
|
|
CREATE TABLE default.demo_orc (`id` Nullable(String), `score` Nullable(Int32), `day` Nullable(String)) ENGINE = Hive('thrift://hivetest:9083', 'test', 'demo_orc') PARTITION BY(day)
|
|
""")
|
|
logging.info("create result {}".format(result))
|
|
|
|
assert result.strip() == ''
|
|
|
|
def test_create_text_table(started_cluster):
|
|
logging.info('Start testing creating hive table ...')
|
|
node = started_cluster.instances['h0_0_0']
|
|
result = node.query("""
|
|
DROP TABLE IF EXISTS default.demo_text;
|
|
CREATE TABLE default.demo_text (`id` Nullable(String), `score` Nullable(Int32), `day` Nullable(String)) ENGINE = Hive('thrift://hivetest:9083', 'test', 'demo_text') PARTITION BY (tuple())
|
|
""")
|
|
logging.info("create result {}".format(result))
|
|
|
|
assert result.strip() == ''
|
|
|
|
def test_parquet_groupby(started_cluster):
|
|
logging.info('Start testing groupby ...')
|
|
node = started_cluster.instances['h0_0_0']
|
|
result = node.query("""
|
|
SELECT day, count(*) FROM default.demo_parquet group by day order by day
|
|
""")
|
|
expected_result = """2021-11-01 1
|
|
2021-11-05 2
|
|
2021-11-11 1
|
|
2021-11-16 2
|
|
"""
|
|
assert result == expected_result
|
|
def test_orc_groupby(started_cluster):
|
|
logging.info('Start testing groupby ...')
|
|
node = started_cluster.instances['h0_0_0']
|
|
result = node.query("""
|
|
SELECT day, count(*) FROM default.demo_orc group by day order by day
|
|
""")
|
|
expected_result = """2021-11-01 1
|
|
2021-11-05 2
|
|
2021-11-11 1
|
|
2021-11-16 2
|
|
"""
|
|
assert result == expected_result
|
|
|
|
def test_text_count(started_cluster):
|
|
node = started_cluster.instances['h0_0_0']
|
|
result = node.query("""
|
|
SELECT day, count(*) FROM default.demo_orc group by day order by day SETTINGS format_csv_delimiter = '\x01'
|
|
""")
|
|
expected_result = """2021-11-01 1
|
|
2021-11-05 2
|
|
2021-11-11 1
|
|
2021-11-16 2
|
|
"""
|
|
assert result == expected_result
|
|
|
|
def test_parquet_groupby_with_cache(started_cluster):
|
|
logging.info('Start testing groupby ...')
|
|
node = started_cluster.instances['h0_0_0']
|
|
result = node.query("""
|
|
SELECT day, count(*) FROM default.demo_parquet group by day order by day
|
|
""")
|
|
expected_result = """2021-11-01 1
|
|
2021-11-05 2
|
|
2021-11-11 1
|
|
2021-11-16 2
|
|
"""
|
|
assert result == expected_result
|
|
def test_cache_read_bytes(started_cluster):
|
|
node = started_cluster.instances['h0_0_0']
|
|
node.query("set input_format_parquet_allow_missing_columns = true")
|
|
result = node.query("""
|
|
DROP TABLE IF EXISTS default.demo_parquet;
|
|
CREATE TABLE default.demo_parquet (`id` Nullable(String), `score` Nullable(Int32), `day` Nullable(String)) ENGINE = Hive('thrift://hivetest:9083', 'test', 'demo') PARTITION BY(day)
|
|
""")
|
|
result = node.query("""
|
|
SELECT day, count(*) FROM default.demo_parquet group by day order by day
|
|
""")
|
|
result = node.query("""
|
|
SELECT day, count(*) FROM default.demo_parquet group by day order by day
|
|
""")
|
|
expected_result = """2021-11-01 1
|
|
2021-11-05 2
|
|
2021-11-11 1
|
|
2021-11-16 2
|
|
"""
|
|
time.sleep(120)
|
|
assert result == expected_result
|
|
result = node.query("select sum(ProfileEvent_ExternalDataSourceLocalCacheReadBytes) from system.metric_log where ProfileEvent_ExternalDataSourceLocalCacheReadBytes > 0")
|
|
logging.info("Read bytes from cache:{}".format(result))
|
|
assert result.strip() != '0'
|