mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-16 04:32:33 +00:00
306 lines
11 KiB
Python
306 lines
11 KiB
Python
# pylint: disable=unused-argument
|
|
# pylint: disable=line-too-long
|
|
# pylint: disable=call-var-from-loop
|
|
# pylint: disable=redefined-outer-name
|
|
|
|
import logging
|
|
|
|
import pytest
|
|
|
|
from helpers.client import QueryRuntimeException
|
|
from helpers.cluster import CLICKHOUSE_CI_MIN_TESTED_VERSION, ClickHouseCluster
|
|
|
|
cluster = ClickHouseCluster(__file__)
|
|
upstream = cluster.add_instance("upstream", use_old_analyzer=True)
|
|
backward = cluster.add_instance(
|
|
"backward",
|
|
image="clickhouse/clickhouse-server",
|
|
tag=CLICKHOUSE_CI_MIN_TESTED_VERSION,
|
|
with_installed_binary=True,
|
|
)
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def start_cluster():
|
|
try:
|
|
cluster.start()
|
|
yield cluster
|
|
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
|
|
def test_aggregate_states(start_cluster):
|
|
"""
|
|
This test goes though all aggregate functions that:
|
|
- has only one argument
|
|
- support string as an argument
|
|
|
|
And do a simple check by creating the aggregate state with one string.
|
|
|
|
Yes this is not covers everything (does not cover functions with
|
|
different number of arguments, types, different states in case of
|
|
multiple values - uniqCombined, but as for uniqCombined it will be
|
|
checked via uniqHLL12), but at least something.
|
|
|
|
And about type, String had been selected, since it more likely that
|
|
there will be used some hash function internally.
|
|
"""
|
|
|
|
aggregate_functions = backward.query(
|
|
"""
|
|
SELECT if(NOT empty(alias_to), alias_to, name)
|
|
FROM system.functions
|
|
WHERE is_aggregate = 1
|
|
"""
|
|
)
|
|
aggregate_functions = aggregate_functions.strip().split("\n")
|
|
aggregate_functions = map(lambda x: x.strip(), aggregate_functions)
|
|
|
|
aggregate_functions = list(aggregate_functions)
|
|
logging.info("Got %s aggregate functions", len(aggregate_functions))
|
|
|
|
skipped = 0
|
|
failed = 0
|
|
passed = 0
|
|
|
|
def get_aggregate_state_hex(node, function_name):
|
|
return node.query(
|
|
f"select hex(initializeAggregation('{function_name}State', 'foo'))"
|
|
).strip()
|
|
|
|
def get_final_value_unhex(node, function_name, value):
|
|
return node.query(
|
|
f"select finalizeAggregation(unhex('{value}')::AggregateFunction({function_name}, String))"
|
|
).strip()
|
|
|
|
for aggregate_function in aggregate_functions:
|
|
logging.info("Checking %s", aggregate_function)
|
|
|
|
try:
|
|
backward_state = get_aggregate_state_hex(backward, aggregate_function)
|
|
except QueryRuntimeException as e:
|
|
error_message = str(e)
|
|
allowed_errors = [
|
|
"ILLEGAL_TYPE_OF_ARGUMENT",
|
|
# sequenceNextNode() and friends
|
|
"UNKNOWN_AGGREGATE_FUNCTION",
|
|
# Function X takes exactly one parameter:
|
|
"NUMBER_OF_ARGUMENTS_DOESNT_MATCH",
|
|
# Function X takes at least one argument
|
|
"TOO_FEW_ARGUMENTS_FOR_FUNCTION",
|
|
# Function X accepts at most 3 arguments, Y given
|
|
"TOO_MANY_ARGUMENTS_FOR_FUNCTION",
|
|
# The function 'X' can only be used as a window function
|
|
"BAD_ARGUMENTS",
|
|
# aggThrow
|
|
"AGGREGATE_FUNCTION_THROW",
|
|
]
|
|
if any(map(lambda x: x in error_message, allowed_errors)):
|
|
logging.info("Skipping %s", aggregate_function)
|
|
skipped += 1
|
|
continue
|
|
logging.exception("Failed %s", aggregate_function)
|
|
failed += 1
|
|
continue
|
|
|
|
upstream_state = get_aggregate_state_hex(upstream, aggregate_function)
|
|
if upstream_state != backward_state:
|
|
allowed_changes_if_result_is_the_same = ["anyHeavy"]
|
|
|
|
if aggregate_function in allowed_changes_if_result_is_the_same:
|
|
backward_final_from_upstream = get_final_value_unhex(
|
|
backward, aggregate_function, upstream_state
|
|
)
|
|
upstream_final_from_backward = get_final_value_unhex(
|
|
upstream, aggregate_function, backward_state
|
|
)
|
|
|
|
if backward_final_from_upstream == upstream_final_from_backward:
|
|
logging.info(
|
|
"OK %s (but different intermediate states)", aggregate_function
|
|
)
|
|
passed += 1
|
|
else:
|
|
logging.error(
|
|
"Failed %s, Intermediate: %s (backward) != %s (upstream). Final from intermediate: %s (backward from upstream state) != %s (upstream from backward state)",
|
|
aggregate_function,
|
|
backward_state,
|
|
upstream_state,
|
|
backward_final_from_upstream,
|
|
upstream_final_from_backward,
|
|
)
|
|
failed += 1
|
|
else:
|
|
logging.error(
|
|
"Failed %s, %s (backward) != %s (upstream)",
|
|
aggregate_function,
|
|
backward_state,
|
|
upstream_state,
|
|
)
|
|
failed += 1
|
|
else:
|
|
logging.info("OK %s", aggregate_function)
|
|
passed += 1
|
|
|
|
logging.info(
|
|
"Aggregate functions: %s, Failed: %s, skipped: %s, passed: %s",
|
|
len(aggregate_functions),
|
|
failed,
|
|
skipped,
|
|
passed,
|
|
)
|
|
assert failed == 0
|
|
assert passed > 0
|
|
assert failed + passed + skipped == len(aggregate_functions)
|
|
|
|
|
|
def test_string_functions(start_cluster):
|
|
if (
|
|
upstream.is_built_with_thread_sanitizer()
|
|
or upstream.is_built_with_memory_sanitizer()
|
|
or upstream.is_built_with_address_sanitizer()
|
|
):
|
|
pytest.skip("The test is slow in builds with sanitizer")
|
|
|
|
functions = backward.query(
|
|
"""
|
|
SELECT if(NOT empty(alias_to), alias_to, name)
|
|
FROM system.functions
|
|
WHERE is_aggregate = 0
|
|
"""
|
|
)
|
|
functions = functions.strip().split("\n")
|
|
functions = map(lambda x: x.strip(), functions)
|
|
|
|
excludes = [
|
|
# The argument of this function is not a seed, but an arbitrary expression needed for bypassing common subexpression elimination.
|
|
"rand",
|
|
"rand64",
|
|
"randConstant",
|
|
"randCanonical",
|
|
"generateUUIDv4",
|
|
"generateULID",
|
|
# Syntax error otherwise
|
|
"position",
|
|
"substring",
|
|
"CAST",
|
|
"getTypeSerializationStreams",
|
|
# NOTE: no need to ignore now()/now64() since they will fail because they don't accept any argument
|
|
# 22.8 Backward Incompatible Change: Extended range of Date32
|
|
"toDate32OrZero",
|
|
"toDate32OrDefault",
|
|
# 23.9 changed the base64-handling library from Turbo base64 to aklomp-base64. They differ in the way they deal with base64 values
|
|
# that are not properly padded by '=', for example below test value v='foo'. (Depending on the specification/context, padding is
|
|
# mandatory or optional). The former lib produces a value based on implicit padding, the latter lib throws an error.
|
|
"FROM_BASE64",
|
|
"base64Decode",
|
|
# PR #56913 (in v23.11) corrected the way tryBase64Decode() behaved with invalid inputs. Old versions return garbage, new versions
|
|
# return an empty string (as it was always documented).
|
|
"tryBase64Decode",
|
|
# Removed in 23.9
|
|
"meiliMatch",
|
|
# These functions require more than one argument.
|
|
"parseDateTimeInJodaSyntaxOrZero",
|
|
"parseDateTimeInJodaSyntaxOrNull",
|
|
"parseDateTimeOrNull",
|
|
"parseDateTimeOrZero",
|
|
"parseDateTime",
|
|
# The argument is effectively a disk name (and we don't have one with name foo)
|
|
"filesystemUnreserved",
|
|
"filesystemCapacity",
|
|
"filesystemAvailable",
|
|
# Exclude it for now. Looks like the result depends on the build type.
|
|
"farmHash64",
|
|
]
|
|
functions = filter(lambda x: x not in excludes, functions)
|
|
|
|
functions = list(functions)
|
|
logging.info("Got %s functions", len(functions))
|
|
|
|
skipped = 0
|
|
failed = 0
|
|
passed = 0
|
|
|
|
def get_function_value(node, function_name, value):
|
|
return node.query(f"select {function_name}('{value}')").strip()
|
|
|
|
v = "foo"
|
|
for function in functions:
|
|
logging.info("Checking %s('%s')", function, v)
|
|
|
|
try:
|
|
backward_value = get_function_value(backward, function, v)
|
|
except QueryRuntimeException as e:
|
|
error_message = str(e)
|
|
allowed_errors = [
|
|
# Messages
|
|
"Cannot load time zone ",
|
|
"No macro ",
|
|
"Should start with ", # POINT/POLYGON/...
|
|
"Cannot read input: expected a digit but got something else:",
|
|
# ErrorCodes
|
|
"ILLEGAL_TYPE_OF_ARGUMENT",
|
|
"DICTIONARIES_WAS_NOT_LOADED",
|
|
"CANNOT_PARSE_UUID",
|
|
"CANNOT_PARSE_DOMAIN_VALUE_FROM_STRING",
|
|
"ILLEGAL_COLUMN",
|
|
"TYPE_MISMATCH",
|
|
"SUPPORT_IS_DISABLED",
|
|
"CANNOT_PARSE_DATE",
|
|
"UNKNOWN_SETTING",
|
|
"CANNOT_PARSE_BOOL",
|
|
"FILE_DOESNT_EXIST",
|
|
"NOT_IMPLEMENTED",
|
|
"BAD_GET",
|
|
"UNKNOWN_TYPE",
|
|
# addressToSymbol
|
|
"FUNCTION_NOT_ALLOWED",
|
|
# Date functions
|
|
"CANNOT_PARSE_TEXT",
|
|
"CANNOT_PARSE_DATETIME",
|
|
# Function X takes exactly one parameter:
|
|
"NUMBER_OF_ARGUMENTS_DOESNT_MATCH",
|
|
# Function X takes at least one argument
|
|
"TOO_FEW_ARGUMENTS_FOR_FUNCTION",
|
|
# Function X accepts at most 3 arguments, Y given
|
|
"TOO_MANY_ARGUMENTS_FOR_FUNCTION",
|
|
# The function 'X' can only be used as a window function
|
|
"BAD_ARGUMENTS",
|
|
# String foo is obviously not a valid IP address.
|
|
"CANNOT_PARSE_IPV4",
|
|
"CANNOT_PARSE_IPV6",
|
|
]
|
|
if any(map(lambda x: x in error_message, allowed_errors)):
|
|
logging.info("Skipping %s", function)
|
|
skipped += 1
|
|
continue
|
|
logging.exception("Failed %s", function)
|
|
failed += 1
|
|
continue
|
|
|
|
upstream_value = get_function_value(upstream, function, v)
|
|
if upstream_value != backward_value:
|
|
logging.warning(
|
|
"Failed %s('%s') %s (backward) != %s (upstream)",
|
|
function,
|
|
v,
|
|
backward_value,
|
|
upstream_value,
|
|
)
|
|
failed += 1
|
|
else:
|
|
logging.info("OK %s", function)
|
|
passed += 1
|
|
|
|
logging.info(
|
|
"Functions: %s, failed: %s, skipped: %s, passed: %s",
|
|
len(functions),
|
|
failed,
|
|
skipped,
|
|
passed,
|
|
)
|
|
assert failed == 0
|
|
assert passed > 0
|
|
assert failed + passed + skipped == len(functions)
|