ClickHouse/dbms/tests/queries/0_stateless/00921_datetime64_compatibility.python
Vasily Nemkov 5fc89bce7c Explicitly specifying timezone when casting
to avoid different results on different machines due to default timezone being used in result datatype.
2020-03-10 16:10:07 +08:00

227 lines
6.1 KiB
Python
Executable File

#!/usr/bin/env python
# encoding: utf-8
import re
import itertools
import sys
import argparse
# Create SQL statement to verify dateTime64 is accepted as argument to functions taking DateTime.
FUNCTIONS="""
toTimeZone(N, 'UTC')
toYear(N)
toQuarter(N)
toMonth(N)
toDayOfYear(N)
toDayOfMonth(N)
toDayOfWeek(N)
toHour(N)
toMinute(N)
toSecond(N)
toUnixTimestamp(N)
toStartOfYear(N)
toStartOfISOYear(N)
toStartOfQuarter(N)
toStartOfMonth(N)
toMonday(N)
toStartOfWeek(N)
toStartOfDay(N)
toStartOfHour(N)
toStartOfMinute(N)
toStartOfFiveMinute(N)
toStartOfTenMinutes(N)
toStartOfFifteenMinutes(N)
toStartOfInterval(N, INTERVAL 1 year)
toStartOfInterval(N, INTERVAL 1 month)
toStartOfInterval(N, INTERVAL 1 day)
toStartOfInterval(N, INTERVAL 15 minute)
toTime(N)
toRelativeYearNum(N)
toRelativeQuarterNum(N)
toRelativeMonthNum(N)
toRelativeWeekNum(N)
toRelativeDayNum(N)
toRelativeHourNum(N)
toRelativeMinuteNum(N)
toRelativeSecondNum(N)
toISOYear(N)
toISOWeek(N)
toWeek(N)
toYearWeek(N)
timeSlot(N)
toYYYYMM(N)
toYYYYMMDD(N)
toYYYYMMDDhhmmss(N)
addYears(N, 1)
addMonths(N, 1)
addWeeks(N, 1)
addDays(N, 1)
addHours(N, 1)
addMinutes(N, 1)
addSeconds(N, 1)
addQuarters(N, 1)
subtractYears(N, 1)
subtractMonths(N, 1)
subtractWeeks(N, 1)
subtractDays(N, 1)
subtractHours(N, 1)
subtractMinutes(N, 1)
subtractSeconds(N, 1)
subtractQuarters(N, 1)
CAST(N as DateTime('Europe/Minsk'))
CAST(N as Date)
CAST(N as UInt64)
CAST(N as DateTime64(0, 'Europe/Minsk'))
CAST(N as DateTime64(3, 'Europe/Minsk'))
CAST(N as DateTime64(6, 'Europe/Minsk'))
CAST(N as DateTime64(9, 'Europe/Minsk'))
# Casting our test values to DateTime(12) will cause an overflow and hence will fail the test under UB sanitizer.
# CAST(N as DateTime64(12))
# DateTime64(18) will always fail due to zero precision, but it is Ok to test here:
# CAST(N as DateTime64(18))
formatDateTime(N, '%C %d %D %e %F %H %I %j %m %M %p %R %S %T %u %V %w %y %Y %%')
""".splitlines()
# Expanded later to cartesian product of all arguments.
# NOTE: {N} to be turned into N after str.format() for keys (format string), but not for list of values!
extra_ops =\
[
# With same type:
(
['N {op} N'],
{
'op':
[
'- ', # does not work, but should it?
'+ ', # does not work, but should it?
'!=', '==', # equality and inequality supposed to take sub-second part in account
'< ',
'<=',
'> ',
'>='
]
}
),
# With other DateTime types:
(
[
'N {op} {arg}',
'{arg} {op} N'
],
{
'op':
[
'-', # does not work, but should it?
'!=', '==',
# these are naturally expected to work, but they don't:
'< ',
'<=',
'> ',
'>='
],
'arg': ['DT', 'D', 'DT64'],
}
),
# With arithmetic types
(
[
'N {op} {arg}',
'{arg} {op} N'
],
{
'op':
[
'+ ',
'- ',
'==',
'!=',
'< ',
'<=',
'> ',
'>='
],
'arg':
[
'toUInt8(1)',
'toInt8(-1)',
'toUInt16(1)',
'toInt16(-1)',
'toUInt32(1)',
'toInt32(-1)',
'toUInt64(1)',
'toInt64(-1)'
],
},
),
]
# Expand extra_ops here
for funcs, args in extra_ops:
args_keys = args.keys()
for args_vals in itertools.product(*args.values()):
for func in funcs:
result_func = func.format(**dict(zip(args_keys, args_vals)))
FUNCTIONS.append(result_func)
# filter out empty lines and commented out lines
COMMENTED_OUT_LINE_RE = re.compile(r"^\s*#")
FUNCTIONS = list(filter(lambda f: len(f) != 0 and COMMENTED_OUT_LINE_RE.match(f) == None, FUNCTIONS))
TYPES = ['D', 'DT', 'DT64']
if sys.version_info[0] > 2:
escape_string_codec = 'unicode_escape'
else:
escape_string_codec = 'string-escape'
def escape_string(s):
return s.encode(escape_string_codec).decode('utf-8')
def execute_functions_for_types(functions, types):
# TODO: use string.Template here to allow lines that do not contain type, like: SELECT CAST(toDateTime64(1234567890), 'DateTime64')
for func in functions:
print("""SELECT 'SELECT {func}';""".format(func=escape_string(func)))
for dt in types:
prologue = "\
WITH \
toDateTime64('2019-09-16 19:20:11.234', 3, 'Europe/Minsk') as DT64, \
toDateTime('2019-09-16 19:20:11', 'Europe/Minsk') as DT, \
toDate('2019-09-16') as D, {X} as N".format(X=dt)
print("""{prologue} SELECT toTypeName(r), {func} as r FORMAT CSV;""".format(prologue=prologue, func=func))
print("""SELECT '------------------------------------------';""")
def main():
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('--functions_re', type=re.compile, help="RE to enable functions", default=None)
parser.add_argument('--types_re',
type=lambda s: re.compile('^(' + s + ')$'),
help="RE to enable types, supported types: " + ",".join(TYPES), default=None)
parser.add_argument('--list_functions', action='store_true', help="List all functions to be tested and exit")
return parser.parse_args()
args = parse_args()
functions = FUNCTIONS
types = TYPES
if args.functions_re:
functions = list(filter(lambda f : args.functions_re.search(f), functions))
if len(functions) == 0:
print("functions list is empty")
return -1
if args.types_re:
types = list(filter(lambda t : args.types_re.match(t), types))
if len(types) == 0:
print("types list is empty")
return -1
if args.list_functions:
print("\n".join(functions))
return 0
execute_functions_for_types(functions, types)
if __name__ == '__main__':
exit(main())