[BUG] test_regexp_choice failed

[jlowe]

Nightly integration test run had a failure in test_regexp_choice.

[2024-03-27T19:43:44.838Z] FAILED ../../src/main/python/regexp_test.py::test_regexp_choice[DATAGEN_SEED=1711561563, TZ=UTC, INJECT_OOM] - AssertionError: GPU and CPU string values are different at [609, 'regexp_extract(a, (abc1a$|ab2ab$), 1)']

Details

[2024-03-27T19:43:44.835Z] =================================== FAILURES ===================================

[2024-03-27T19:43:44.835Z] ______________________________ test_regexp_choice ______________________________

[2024-03-27T19:43:44.835Z] [gw0] linux -- Python 3.9.19 /opt/conda/bin/python

[2024-03-27T19:43:44.835Z] 

[2024-03-27T19:43:44.835Z]     def test_regexp_choice():

[2024-03-27T19:43:44.835Z]         gen = mk_str_gen('[abcd]{1,3}[0-9]{1,3}[abcd]{1,3}[ \n\t\r]{0,2}')

[2024-03-27T19:43:44.835Z] >       assert_gpu_and_cpu_are_equal_collect(

[2024-03-27T19:43:44.835Z]                 lambda spark: unary_op_df(spark, gen).selectExpr(

[2024-03-27T19:43:44.835Z]                     'rlike(a, "[abcd]|[123]")',

[2024-03-27T19:43:44.835Z]                     'rlike(a, "[^\n\r]|abcd")',

[2024-03-27T19:43:44.835Z]                     'rlike(a, "abd1a$|^ab2a")',

[2024-03-27T19:43:44.835Z]                     'rlike(a, "[a-c]*|[\n]")',

[2024-03-27T19:43:44.835Z]                     'rlike(a, "[a-c]+|[\n]")',

[2024-03-27T19:43:44.835Z]                     'regexp_extract(a, "(abc1a$|^ab2ab|a3abc)", 1)',

[2024-03-27T19:43:44.835Z]                     'regexp_extract(a, "(abc1a$|ab2ab$)", 1)',

[2024-03-27T19:43:44.835Z]                     'regexp_extract(a, "(ab+|^ab)", 1)',

[2024-03-27T19:43:44.835Z]                     'regexp_extract(a, "(ab*|^ab)", 1)',

[2024-03-27T19:43:44.835Z]                     'regexp_replace(a, "[abcd]$|^abc", "@")',

[2024-03-27T19:43:44.835Z]                     'regexp_replace(a, "[ab]$|[cd]$", "@")',

[2024-03-27T19:43:44.835Z]                     'regexp_replace(a, "[ab]+|^cd1", "@")'

[2024-03-27T19:43:44.835Z]                 ),

[2024-03-27T19:43:44.835Z]             conf=_regexp_conf)

[2024-03-27T19:43:44.835Z] 

[2024-03-27T19:43:44.835Z] ../../src/main/python/regexp_test.py:568: 

[2024-03-27T19:43:44.835Z] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

[2024-03-27T19:43:44.835Z] ../../src/main/python/asserts.py:595: in assert_gpu_and_cpu_are_equal_collect

[2024-03-27T19:43:44.835Z]     _assert_gpu_and_cpu_are_equal(func, 'COLLECT', conf=conf, is_cpu_first=is_cpu_first, result_canonicalize_func_before_compare=result_canonicalize_func_before_compare)

[2024-03-27T19:43:44.835Z] ../../src/main/python/asserts.py:517: in _assert_gpu_and_cpu_are_equal

[2024-03-27T19:43:44.835Z]     assert_equal(from_cpu, from_gpu)

[2024-03-27T19:43:44.835Z] ../../src/main/python/asserts.py:107: in assert_equal

[2024-03-27T19:43:44.835Z]     _assert_equal(cpu, gpu, float_check=get_float_check(), path=[])

[2024-03-27T19:43:44.835Z] ../../src/main/python/asserts.py:43: in _assert_equal

[2024-03-27T19:43:44.835Z]     _assert_equal(cpu[index], gpu[index], float_check, path + [index])

[2024-03-27T19:43:44.835Z] ../../src/main/python/asserts.py:36: in _assert_equal

[2024-03-27T19:43:44.835Z]     _assert_equal(cpu[field], gpu[field], float_check, path + [field])

[2024-03-27T19:43:44.835Z] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

[2024-03-27T19:43:44.835Z] 

[2024-03-27T19:43:44.835Z] cpu = 'ab2ab', gpu = 'ab2ab\r'

[2024-03-27T19:43:44.835Z] float_check = <function get_float_check.<locals>.<lambda> at 0x7f86b47c1040>

[2024-03-27T19:43:44.835Z] path = [609, 'regexp_extract(a, (abc1a$|ab2ab$), 1)']

[2024-03-27T19:43:44.835Z] 

[2024-03-27T19:43:44.835Z]     def _assert_equal(cpu, gpu, float_check, path):

[2024-03-27T19:43:44.835Z]         t = type(cpu)

[2024-03-27T19:43:44.835Z]         if (t is Row):

[2024-03-27T19:43:44.835Z]             assert len(cpu) == len(gpu), "CPU and GPU row have different lengths at {} CPU: {} GPU: {}".format(path, len(cpu), len(gpu))

[2024-03-27T19:43:44.835Z]             if hasattr(cpu, "__fields__") and hasattr(gpu, "__fields__"):

[2024-03-27T19:43:44.835Z]                 assert cpu.__fields__ == gpu.__fields__, "CPU and GPU row have different fields at {} CPU: {} GPU: {}".format(path, cpu.__fields__, gpu.__fields__)

[2024-03-27T19:43:44.836Z]                 for field in cpu.__fields__:

[2024-03-27T19:43:44.836Z]                     _assert_equal(cpu[field], gpu[field], float_check, path + [field])

[2024-03-27T19:43:44.836Z]             else:

[2024-03-27T19:43:44.836Z]                 for index in range(len(cpu)):

[2024-03-27T19:43:44.836Z]                     _assert_equal(cpu[index], gpu[index], float_check, path + [index])

[2024-03-27T19:43:44.836Z]         elif (t is list):

[2024-03-27T19:43:44.836Z]             assert len(cpu) == len(gpu), "CPU and GPU list have different lengths at {} CPU: {} GPU: {}".format(path, len(cpu), len(gpu))

[2024-03-27T19:43:44.836Z]             for index in range(len(cpu)):

[2024-03-27T19:43:44.836Z]                 _assert_equal(cpu[index], gpu[index], float_check, path + [index])

[2024-03-27T19:43:44.836Z]         elif (t is tuple):

[2024-03-27T19:43:44.836Z]             assert len(cpu) == len(gpu), "CPU and GPU list have different lengths at {} CPU: {} GPU: {}".format(path, len(cpu), len(gpu))

[2024-03-27T19:43:44.836Z]             for index in range(len(cpu)):

[2024-03-27T19:43:44.836Z]                 _assert_equal(cpu[index], gpu[index], float_check, path + [index])

[2024-03-27T19:43:44.836Z]         elif (t is pytypes.GeneratorType):

[2024-03-27T19:43:44.836Z]             index = 0

[2024-03-27T19:43:44.836Z]             # generator has no zip :( so we have to do this the hard way

[2024-03-27T19:43:44.836Z]             done = False

[2024-03-27T19:43:44.836Z]             while not done:

[2024-03-27T19:43:44.836Z]                 sub_cpu = None

[2024-03-27T19:43:44.836Z]                 sub_gpu = None

[2024-03-27T19:43:44.836Z]                 try:

[2024-03-27T19:43:44.836Z]                     sub_cpu = next(cpu)

[2024-03-27T19:43:44.836Z]                 except StopIteration:

[2024-03-27T19:43:44.836Z]                     done = True

[2024-03-27T19:43:44.836Z]     

[2024-03-27T19:43:44.836Z]                 try:

[2024-03-27T19:43:44.836Z]                     sub_gpu = next(gpu)

[2024-03-27T19:43:44.836Z]                 except StopIteration:

[2024-03-27T19:43:44.836Z]                     done = True

[2024-03-27T19:43:44.836Z]     

[2024-03-27T19:43:44.836Z]                 if done:

[2024-03-27T19:43:44.836Z]                     assert sub_cpu == sub_gpu and sub_cpu == None, "CPU and GPU generators have different lengths at {}".format(path)

[2024-03-27T19:43:44.836Z]                 else:

[2024-03-27T19:43:44.836Z]                     _assert_equal(sub_cpu, sub_gpu, float_check, path + [index])

[2024-03-27T19:43:44.836Z]     

[2024-03-27T19:43:44.836Z]                 index = index + 1

[2024-03-27T19:43:44.836Z]         elif (t is dict):

[2024-03-27T19:43:44.836Z]             # The order of key/values is not guaranteed in python dicts, nor are they guaranteed by Spark

[2024-03-27T19:43:44.836Z]             # so sort the items to do our best with ignoring the order of dicts

[2024-03-27T19:43:44.836Z]             cpu_items = list(cpu.items()).sort(key=_RowCmp)

[2024-03-27T19:43:44.836Z]             gpu_items = list(gpu.items()).sort(key=_RowCmp)

[2024-03-27T19:43:44.836Z]             _assert_equal(cpu_items, gpu_items, float_check, path + ["map"])

[2024-03-27T19:43:44.836Z]         elif (t is int):

[2024-03-27T19:43:44.836Z]             assert cpu == gpu, "GPU and CPU int values are different at {}".format(path)

[2024-03-27T19:43:44.836Z]         elif (t is float):

[2024-03-27T19:43:44.836Z]             if (math.isnan(cpu)):

[2024-03-27T19:43:44.836Z]                 assert math.isnan(gpu), "GPU and CPU float values are different at {}".format(path)

[2024-03-27T19:43:44.836Z]             else:

[2024-03-27T19:43:44.836Z]                 assert float_check(cpu, gpu), "GPU and CPU float values are different {}".format(path)

[2024-03-27T19:43:44.836Z]         elif isinstance(cpu, str):

[2024-03-27T19:43:44.836Z] >           assert cpu == gpu, "GPU and CPU string values are different at {}".format(path)

[2024-03-27T19:43:44.836Z] E           AssertionError: GPU and CPU string values are different at [609, 'regexp_extract(a, (abc1a$|ab2ab$), 1)']

[2024-03-27T19:43:44.836Z] 

[2024-03-27T19:43:44.836Z] ../../src/main/python/asserts.py:85: AssertionError

[sameerz]

Not sure if we should set a fixed seed for this issue, seems like it needs to be investigated and resolved.

[revans2]

We should resolve all of the failures. Setting a fixed seed is a way for the build to stop failing while we do it.