Sporadic test failure with fast io.ascii reader by astrofrog · Pull Request #2964 · astropy/astropy

astrofrog · 2014-11-02T22:49:22Z

I'm seeing this test failure from time to time on the Mac Jenkins instance - it's not systematic, so maybe there is a race condition, or something to do with random numbers (haven't checked in detail):

def test_fill_include_exclude_names():
        """
        fill_include_names and fill_exclude_names should filter missing/empty value handling
        in the same way that include_names and exclude_names filter output columns.
        """
        text = """
    A, B, C
    , 1, 2
    3, , 4
    5, 5,
    """
        table = read_csv(text, fill_include_names=['A', 'B'])
        assert table['A'][0] is ma.masked
        assert table['B'][1] is ma.masked
        assert table['C'][2] is not ma.masked # C not in fill_include_names

>       table = read_csv(text, fill_exclude_names=['A', 'B'])

astropy/io/ascii/tests/test_c_reader.py:435: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = '\nA, B, C\n, 1, 2\n3, , 4\n5, 5,\n'
kwargs = {'fill_exclude_names': ['A', 'B']}

    def read_csv(table, **kwargs):
>       return _read(table, FastCsv, 'csv', **kwargs)

astropy/io/ascii/tests/test_c_reader.py:81: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = '\nA, B, C\n, 1, 2\n3, , 4\n5, 5,\n\n'
Reader = <class 'astropy.io.ascii.fastbasic.FastCsv'>, format = 'csv'
fail_parallel = False, kwargs = {'fill_exclude_names': ['A', 'B']}
reader = <astropy.io.ascii.fastbasic.FastCsv object at 0x10f9cbd90>
t1 = <Table rows=3 names=('A','B','C')>
masked_array(data = [('', '1', 2) ('3', '',...', 'N', 999999),
            dtype = [('A', 'S1'), ('B', 'S1'), ('C', '<i8')])

t2 = <Table rows=3 names=('A','B','C')>
masked_array(data = [('', '1', 2) ('3', '',...', 'N', 999999),
            dtype = [('A', 'S1'), ('B', 'S1'), ('C', '<i8')])

t3 = <Table rows=3 names=('A','B','C')>
masked_array(data = [('', '1', 2) ('3', '',...', 'N', 999999),
            dtype = [('A', 'S1'), ('B', 'S1'), ('C', '<i8')])

t4 = <Table rows=3 names=('A','B','C')>
masked_array(data = [(8070450532247928832, ..., 'N', 999999),
            dtype = [('A', '<i8'), ('B', 'S1'), ('C', '<i8')])

t5 = <Table rows=3 names=('A','B','C')>
masked_array(data = [('', '1', 2) ('3', '',...', 'N', 999999),
            dtype = [('A', 'S1'), ('B', 'S1'), ('C', '<i8')])


    def _read(table, Reader, format, fail_parallel=False, **kwargs):
        # make sure we have a newline so table can't be misinterpreted as a filename
        table += '\n'
        reader = Reader(**kwargs)
        t1 = reader.read(table)
        t2 = reader.read(StringIO(table))
        t3 = reader.read(table.splitlines())
        t4 = ascii.read(table, format=format, guess=False, **kwargs)
        t5 = ascii.read(table, format=format, guess=False, fast_reader=False, **kwargs)
        assert_table_equal(t1, t2)
        assert_table_equal(t2, t3)
>       assert_table_equal(t3, t4)

astropy/io/ascii/tests/test_c_reader.py:55: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

t1 = <Table rows=3 names=('A','B','C')>
masked_array(data = [('', '1', 2) ('3', '',...', 'N', 999999),
            dtype = [('A', 'S1'), ('B', 'S1'), ('C', '<i8')])

t2 = <Table rows=3 names=('A','B','C')>
masked_array(data = [(8070450532247928832, ..., 'N', 999999),
            dtype = [('A', '<i8'), ('B', 'S1'), ('C', '<i8')])


    def assert_table_equal(t1, t2):
        assert_equal(len(t1), len(t2))
        assert_equal(t1.colnames, t2.colnames)
        for name in t1.colnames:
            if len(t1) != 0:
>               assert_equal(t1[name].dtype.kind, t2[name].dtype.kind)

astropy/io/ascii/tests/test_c_reader.py:31: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

a = 'S', b = 'i'

    def assert_equal(a, b):
>       assert a == b
E       assert 'S' == 'i'
E         - S
E         + i

astropy/io/ascii/tests/common.py:41: AssertionError

cc @AmrAS1 @taldcroft

mdmueller · 2014-09-19T22:19:00Z

Hm...at first glance I thought it strange that ascii.read messes up while reader.read doesn't, but then I realized that the latter doesn't receive kwargs. That's not a big deal but I might as well open a PR to fix it for the sake of completeness. Never mind, I just realized thatkwargs is passed in Reader.__init__.

There aren't any random numbers involved, so my guess is this is some problem with memory. The parser has a separate buffer for empty strings, so it could be something to do with that. I'll look closer at the relevant code and see if anything could be problematic.

jfoster17 · 2014-09-26T17:43:17Z

I've also run into this test failing (sporadically) when running tests manually.

mdmueller · 2014-09-27T03:53:06Z

Unfortunately I have no way of testing this myself, but I may possibly have fixed the issue here. Can anyone see if the problem still exist with these changes?

astrofrog · 2014-09-27T08:56:01Z

@AmrAS1 - I set up a script to test this many times and it seems like the issue happened 10% of the time:

Good: 043, Bad: 005, Failure rate:  10.4%

I tested out your branch and the result is similar:

Good: 026, Bad: 003, Failure rate:  10.3%

so I don't think the issue has been fixed.

astrofrog · 2014-09-27T09:08:37Z

@AmrAS1 - in case this is useful, I noticed that it is not always the same assert that fails. Here is an example of one of the failures, with the tables printed out (this is with your branch):

______________________________________________________ test_fill_values ______________________________________________________

    def test_fill_values():
        """
        Make sure that the parameter fill_values works as intended. If fill_values
        is not specified, the default behavior should be to convert '' to 0.
        """
        text = """
    A, B, C
    , 2, nan
    a, -999, -3.4
    nan, 5, -9999
    8, nan, 7.6e12
    """
>       table = read_basic(text, delimiter=',')

astropy/io/ascii/tests/test_c_reader.py:440: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = '\nA, B, C\n, 2, nan\na, -999, -3.4\nnan, 5, -9999\n8, nan, 7.6e12\n', kwargs = {'delimiter': ','}

    def read_basic(table, **kwargs):
>       return _read(table, FastBasic, 'basic', **kwargs)

astropy/io/ascii/tests/test_c_reader.py:133: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = '\nA, B, C\n, 2, nan\na, -999, -3.4\nnan, 5, -9999\n8, nan, 7.6e12\n\n'
Reader = <class 'astropy.io.ascii.fastbasic.FastBasic'>, format = 'basic', fail_parallel = False, kwargs = {'delimiter': ','}
reader = <astropy.io.ascii.fastbasic.FastBasic object at 0x1042c0310>
t1 = <Table rows=4 names=('A','B','C')>
array([(2305843009213693952, 2.0, nan),
   ...an, 7600000000000.0)], 
      dtype=[('A', '<i8'), ('B', '<f8'), ('C', '<f8')])
t2 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, nan) ('a', -... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])

t3 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, nan) ('a', -... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])

t4 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, nan) ('a', -... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])

t5 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, nan) ('a', -... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])


    def _read(table, Reader, format, fail_parallel=False, **kwargs):
        # make sure we have a newline so table can't be misinterpreted as a filename
        table += '\n'
        reader = Reader(**kwargs)
        t1 = reader.read(table)
        t2 = reader.read(StringIO(table))
        t3 = reader.read(table.splitlines())
        t4 = ascii.read(table, format=format, guess=False, **kwargs)
        t5 = ascii.read(table, format=format, guess=False, fast_reader=False, **kwargs)

        try:
>           assert_table_equal(t1, t2)

astropy/io/ascii/tests/test_c_reader.py:55: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

t1 = <Table rows=4 names=('A','B','C')>
array([(2305843009213693952, 2.0, nan),
   ...an, 7600000000000.0)], 
      dtype=[('A', '<i8'), ('B', '<f8'), ('C', '<f8')])
t2 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, nan) ('a', -... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])


    def assert_table_equal(t1, t2):
        assert_equal(len(t1), len(t2))
        assert_equal(t1.colnames, t2.colnames)
        for name in t1.colnames:
            if len(t1) != 0:
>               assert_equal(t1[name].dtype.kind, t2[name].dtype.kind)

astropy/io/ascii/tests/test_c_reader.py:31: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

a = 'i', b = 'S'

    def assert_equal(a, b):
>       assert a == b
E       assert 'i' == 'S'
E         - i
E         + S

astropy/io/ascii/tests/common.py:41: AssertionError
------------------------------------------------------ Captured stdout -------------------------------------------------------
[('A', '<i8'), ('B', '<f8'), ('C', '<f8')]
         A             B       C   
-------------------- ------ -------
 2305843009213693952    2.0     nan
-2305834224835294703 -999.0    -3.4
          4364959746    5.0 -9999.0
     571734331819593    nan 7.6e+12
[('A', 'S3'), ('B', '<f8'), ('C', '<f8')]
 A    B       C   
--- ------ -------
 --    2.0     nan
  a -999.0    -3.4
nan    5.0 -9999.0
  8    nan 7.6e+12

Here is another failure:

__________________________________________________ test_quoted_empty_values __________________________________________________

    def test_quoted_empty_values():
        """
        Quoted empty values spanning multiple lines should be treated correctly.
        """
        text = 'a b c\n1 2 " \n "'
>       table = read_basic(text, fail_parallel=True)

astropy/io/ascii/tests/test_c_reader.py:684: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = 'a b c\n1 2 " \n "', kwargs = {'fail_parallel': True}

    def read_basic(table, **kwargs):
>       return _read(table, FastBasic, 'basic', **kwargs)

astropy/io/ascii/tests/test_c_reader.py:133: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = 'a b c\n1 2 " \n "\n', Reader = <class 'astropy.io.ascii.fastbasic.FastBasic'>, format = 'basic', fail_parallel = True
kwargs = {}, reader = <astropy.io.ascii.fastbasic.FastBasic object at 0x1063d7a50>
t1 = <Table rows=1 names=('a','b','c')>
masked_array(data = [(1, 2, --)],
         ...9999, 999999),
            dtype = [('a', '<i8'), ('b', '<i8'), ('c', '<i8')])

t2 = <Table rows=1 names=('a','b','c')>
masked_array(data = [(1, 2, --)],
         ...9999, 999999),
            dtype = [('a', '<i8'), ('b', '<i8'), ('c', '<i8')])

t3 = <Table rows=1 names=('a','b','c')>
masked_array(data = [(1, 2, --)],
         ...9999, 999999),
            dtype = [('a', '<i8'), ('b', '<i8'), ('c', '<i8')])

t4 = <Table rows=1 names=('a','b','c')>
array([(1, 2, -2305843009213693952)], 
      dtype=[('a', '<i8'), ('b', '<i8'), ('c', '<i8')])
t5 = <Table rows=1 names=('a','b','c')>
masked_array(data = [(1, 2, --)],
         ...9999, 999999),
            dtype = [('a', '<i8'), ('b', '<i8'), ('c', '<i8')])


    def _read(table, Reader, format, fail_parallel=False, **kwargs):
        # make sure we have a newline so table can't be misinterpreted as a filename
        table += '\n'
        reader = Reader(**kwargs)
        t1 = reader.read(table)
        t2 = reader.read(StringIO(table))
        t3 = reader.read(table.splitlines())
        t4 = ascii.read(table, format=format, guess=False, **kwargs)
        t5 = ascii.read(table, format=format, guess=False, fast_reader=False, **kwargs)

        try:
            assert_table_equal(t1, t2)
        except:
            print(t1.dtype)
            print(t1)
            print(t2.dtype)
            print(t2)
            raise

        try:
            assert_table_equal(t2, t3)
        except:
            print(t2.dtype)
            print(t2)
            print(t3.dtype)
            print(t3)
            raise

        try:
            assert_table_equal(t3, t4)
        except:
            print(t3.dtype)
            print(t3)
            print(t4.dtype)
            print(t4)
            raise

        try:
>           assert_table_equal(t4, t5)

astropy/io/ascii/tests/test_c_reader.py:82: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

t1 = <Table rows=1 names=('a','b','c')>
array([(1, 2, -2305843009213693952)], 
      dtype=[('a', '<i8'), ('b', '<i8'), ('c', '<i8')])
t2 = <Table rows=1 names=('a','b','c')>
masked_array(data = [(1, 2, --)],
         ...9999, 999999),
            dtype = [('a', '<i8'), ('b', '<i8'), ('c', '<i8')])


    def assert_table_equal(t1, t2):
        assert_equal(len(t1), len(t2))
        assert_equal(t1.colnames, t2.colnames)
        for name in t1.colnames:
            if len(t1) != 0:
                assert_equal(t1[name].dtype.kind, t2[name].dtype.kind)
            if not isinstance(t1[name], MaskedColumn):
                for i, el in enumerate(t1[name]):
                    try:
                        if not isinstance(el, six.string_types) and np.isnan(el):
                            assert_true(not isinstance(t2[name][i], six.string_types) and np.isnan(t2[name][i]))
                        elif isinstance(el, six.string_types):
                            assert_equal(el, t2[name][i])
                        else:
>                           assert_almost_equal(el, t2[name][i])

astropy/io/ascii/tests/test_c_reader.py:40: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

a = -2305843009213693952, b = masked

    def assert_almost_equal(a, b):
>       assert np.allclose(a, b)
E       assert <function allclose at 0x103b08aa0>(-2305843009213693952, masked)
E        +  where <function allclose at 0x103b08aa0> = <function allclose at 0x103b08aa0>
E        +    where <function allclose at 0x103b08aa0> = np.allclose

astropy/io/ascii/tests/common.py:45: AssertionError
------------------------------------------------------ Captured stdout -------------------------------------------------------
[('a', '<i8'), ('b', '<i8'), ('c', '<i8')]
 a   b           c          
--- --- --------------------
  1   2 -2305843009213693952
[('a', '<i8'), ('b', '<i8'), ('c', '<i8')]
 a   b   c 
--- --- ---
  1   2  --

and another

______________________________________________________ test_fill_values ______________________________________________________

    def test_fill_values():
        """
        Make sure that the parameter fill_values works as intended. If fill_values
        is not specified, the default behavior should be to convert '' to 0.
        """
        text = """
    A, B, C
    , 2, nan
    a, -999, -3.4
    nan, 5, -9999
    8, nan, 7.6e12
    """
        table = read_basic(text, delimiter=',')
        # The empty value in row A should become a masked '0'
        assert isinstance(table['A'], MaskedColumn)
        assert table['A'][0] is ma.masked
        # '0' rather than 0 because there is a string in the column
        assert_equal(table['A'].data.data[0], '0')
        assert table['A'][1] is not ma.masked

        table = read_basic(text, delimiter=',', fill_values=('-999', '0'))
        assert isinstance(table['B'], MaskedColumn)
        assert table['A'][0] is not ma.masked # empty value unaffected
        assert table['C'][2] is not ma.masked # -9999 is not an exact match
        assert table['B'][1] is ma.masked
        # Numeric because the rest of the column contains numeric data
        assert_equal(table['B'].data.data[1], 0.0)
        assert table['B'][0] is not ma.masked

        table = read_basic(text, delimiter=',', fill_values=[])
        # None of the columns should be masked
        for name in 'ABC':
            assert not isinstance(table[name], MaskedColumn)

        table = read_basic(text, delimiter=',', fill_values=[('', '0', 'A'),
>                                   ('nan', '999', 'A', 'C')])

astropy/io/ascii/tests/test_c_reader.py:463: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = '\nA, B, C\n, 2, nan\na, -999, -3.4\nnan, 5, -9999\n8, nan, 7.6e12\n'
kwargs = {'delimiter': ',', 'fill_values': [('', '0', 'A'), ('nan', '999', 'A', 'C')]}

    def read_basic(table, **kwargs):
>       return _read(table, FastBasic, 'basic', **kwargs)

astropy/io/ascii/tests/test_c_reader.py:133: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

table = '\nA, B, C\n, 2, nan\na, -999, -3.4\nnan, 5, -9999\n8, nan, 7.6e12\n\n'
Reader = <class 'astropy.io.ascii.fastbasic.FastBasic'>, format = 'basic', fail_parallel = False
kwargs = {'delimiter': ',', 'fill_values': [('', '0', 'A'), ('nan', '999', 'A', 'C')]}
reader = <astropy.io.ascii.fastbasic.FastBasic object at 0x1084ffb10>
t1 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, --) ('a', -9... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])

t2 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, --) ('a', -9... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])

t3 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, --) ('a', -9... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])

t4 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(5764607523034234880, ...1e+20, 1e+20),
            dtype = [('A', '<i8'), ('B', '<f8'), ('C', '<f8')])

t5 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, --) ('a', -9... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])


    def _read(table, Reader, format, fail_parallel=False, **kwargs):
        # make sure we have a newline so table can't be misinterpreted as a filename
        table += '\n'
        reader = Reader(**kwargs)
        t1 = reader.read(table)
        t2 = reader.read(StringIO(table))
        t3 = reader.read(table.splitlines())
        t4 = ascii.read(table, format=format, guess=False, **kwargs)
        t5 = ascii.read(table, format=format, guess=False, fast_reader=False, **kwargs)

        try:
            assert_table_equal(t1, t2)
        except:
            print(t1.dtype)
            print(t1)
            print(t2.dtype)
            print(t2)
            raise

        try:
            assert_table_equal(t2, t3)
        except:
            print(t2.dtype)
            print(t2)
            print(t3.dtype)
            print(t3)
            raise

        try:
>           assert_table_equal(t3, t4)

astropy/io/ascii/tests/test_c_reader.py:73: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

t1 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(--, 2.0, --) ('a', -9... 1e+20, 1e+20),
            dtype = [('A', 'S3'), ('B', '<f8'), ('C', '<f8')])

t2 = <Table rows=4 names=('A','B','C')>
masked_array(data = [(5764607523034234880, ...1e+20, 1e+20),
            dtype = [('A', '<i8'), ('B', '<f8'), ('C', '<f8')])


    def assert_table_equal(t1, t2):
        assert_equal(len(t1), len(t2))
        assert_equal(t1.colnames, t2.colnames)
        for name in t1.colnames:
            if len(t1) != 0:
>               assert_equal(t1[name].dtype.kind, t2[name].dtype.kind)

astropy/io/ascii/tests/test_c_reader.py:31: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

a = 'S', b = 'i'

    def assert_equal(a, b):
>       assert a == b
E       assert 'S' == 'i'
E         - S
E         + i

astropy/io/ascii/tests/common.py:41: AssertionError
------------------------------------------------------ Captured stdout -------------------------------------------------------
[('A', 'S3'), ('B', '<f8'), ('C', '<f8')]
 A    B       C   
--- ------ -------
 --    2.0      --
  a -999.0    -3.4
 --    5.0 -9999.0
  8    nan 7.6e+12
[('A', '<i8'), ('B', '<f8'), ('C', '<f8')]
         A            B       C   
------------------- ------ -------
5764607523034234880    2.0      --
1152930289512690672 -999.0    -3.4
         4398514178    5.0 -9999.0
    562954351945112    nan 7.6e+12

mdmueller · 2014-09-27T13:58:06Z

@astrofrog - Thanks for the output -- I had been assuming it had something to do with the include/exclude names scheme, which doesn't seem to be the case. Looks like there's a problem with empty values that somehow screws up the rest of the column...I'll see if I can figure out what's up.

mdmueller · 2014-10-05T22:10:27Z

@astrofrog - Sorry I haven't been able to look at this for a while. I managed to get hold of a Mac to do a bit of testing, and the fix in this branch worked for me. Can you test this out and see if the test failures persist?

astrofrog · 2014-10-14T08:29:17Z

@AmrAS1 - unfortunately it looks like the issue is still there. I wrote this script to repeat the test many times:

https://gist.github.com/astrofrog/d7be9e4b6bfc474d5a18

I seem to get a failure between 5 and 10% of the time. I'm wondering whether it has to do with optimization.

astrofrog · 2014-10-14T08:44:19Z

@AmrAS1 - you can also instead use my repeat-option branch (https://github.com/astrofrog/astropy/tree/repeat-option), make sure you use build_ext --inplace, and then I can reproduce the failures although less frequently somehow:

$ py.test --repeat=1000 -k test_fill_include_exclude_names astropy/io/ascii/tests/test_c_reader.py
======================================================================= test session starts ========================================================================
platform darwin -- Python 3.4.1 -- py-1.4.23 -- pytest-2.6.1

Running tests with Astropy version 1.0.dev10080.
Running tests in astropy/io/ascii/tests/test_c_reader.py.

Platform: Darwin-12.5.0-x86_64-i386-64bit

Executable: /opt/local/Library/Frameworks/Python.framework/Versions/3.4/bin/python3.4

Full Python Version: 
3.4.1 (default, May 21 2014, 21:17:51) 
[GCC 4.2.1 Compatible Apple Clang 4.1 ((tags/Apple/clang-421.11.66))]

encodings: sys: utf-8, locale: UTF-8, filesystem: utf-8
byteorder: little
float info: dig: 15, mant_dig: 15

Numpy: 1.10.0.dev-a32625e
Scipy: 0.14.0
Matplotlib: 1.4.0
h5py: 2.3.1

collected 66000 items 

astropy/io/ascii/tests/test_c_reader.py ..................................................................................................................................................................................................................................................................................................................................................................................................................................................F............................................................................................................................................................................................................................................................................F................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................

============================================================================= FAILURES =============================================================================
____________________________________________________________ test_fill_include_exclude_names[217-True] _____________________________________________________________

parallel = True

    @pytest.mark.parametrize("parallel", [True, False])
    def test_fill_include_exclude_names(parallel):
        """
        fill_include_names and fill_exclude_names should filter missing/empty value handling
        in the same way that include_names and exclude_names filter output columns.
        """
        text = """
    A, B, C
    , 1, 2
    3, , 4
    5, 5,
    """
        table = read_csv(text, fill_include_names=['A', 'B'], parallel=parallel)
        assert table['A'][0] is ma.masked
        assert table['B'][1] is ma.masked
        assert table['C'][2] is not ma.masked # C not in fill_include_names

        table = read_csv(text, fill_exclude_names=['A', 'B'], parallel=parallel)
        assert table['C'][2] is ma.masked
        assert table['A'][0] is not ma.masked
        assert table['B'][1] is not ma.masked # A and B excluded from fill handling

>       table = read_csv(text, fill_include_names=['A', 'B'], fill_exclude_names=['B'], parallel=parallel)

astropy/io/ascii/tests/test_c_reader.py:463: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
astropy/io/ascii/tests/test_c_reader.py:84: in read_csv
    return _read(table, FastCsv, 'csv', **kwargs)
astropy/io/ascii/tests/test_c_reader.py:77: in _read
    assert_table_equal(t1, t8)
astropy/io/ascii/tests/test_c_reader.py:31: in assert_table_equal
    assert_equal(t1[name].dtype.kind, t2[name].dtype.kind)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

a = 'i', b = 'U'

    def assert_equal(a, b):
>       assert a == b
E       assert 'i' == 'U'
E         - i
E         + U

astropy/io/ascii/tests/common.py:41: AssertionError
____________________________________________________________ test_fill_include_exclude_names[351-False] ____________________________________________________________

parallel = False

    @pytest.mark.parametrize("parallel", [True, False])
    def test_fill_include_exclude_names(parallel):
        """
        fill_include_names and fill_exclude_names should filter missing/empty value handling
        in the same way that include_names and exclude_names filter output columns.
        """
        text = """
    A, B, C
    , 1, 2
    3, , 4
    5, 5,
    """
>       table = read_csv(text, fill_include_names=['A', 'B'], parallel=parallel)

astropy/io/ascii/tests/test_c_reader.py:453: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
astropy/io/ascii/tests/test_c_reader.py:84: in read_csv
    return _read(table, FastCsv, 'csv', **kwargs)
astropy/io/ascii/tests/test_c_reader.py:53: in _read
    assert_table_equal(t1, t2)
astropy/io/ascii/tests/test_c_reader.py:31: in assert_table_equal
    assert_equal(t1[name].dtype.kind, t2[name].dtype.kind)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

a = 'U', b = 'i'

    def assert_equal(a, b):
>       assert a == b
E       assert 'U' == 'i'
E         - U
E         + i

astropy/io/ascii/tests/common.py:41: AssertionError
================================================== 64000 tests deselected by '-ktest_fill_include_exclude_names' ===================================================
==================================================== 2 failed, 1998 passed, 64000 deselected in 341.92 seconds =====================================================

(I had to rebase your branch on master to avoid the 'too many files' error)

mdmueller · 2014-11-02T22:30:55Z

@astrofrog - I think I might have a solution with the latest commit in this branch; I repeated tests 100 times and didn't get an error. Can you test this out?

Sorry it's taken so long for me to look more into this--I've been busier than I anticipated. By the way, I changed my Github username (previously amras1).

astrofrog · 2014-11-02T22:49:39Z

@mdmueller - works for me! I've attached the branch to this issue. Thanks!

astrofrog · 2014-11-03T17:11:07Z

@mdmueller - just to check, do you consider this ready to merge?

mdmueller · 2014-11-03T18:20:48Z

Yes, this is good from my perspective.

astrofrog · 2014-11-03T18:25:15Z

Ok, thanks! Merging :)

Sporadic test failure with fast io.ascii reader

astrofrog added the io.ascii label Sep 19, 2014

embray added Bug Affects-dev PRs and issues that do not impact an existing Astropy release labels Sep 29, 2014

Fixed error in io.ascii on Mac

b67d1ea

astrofrog mentioned this pull request Oct 6, 2014

Added a --repeat option for testing #3019

Merged

Modified empty field check at the beginning of columns

5cf1c0e

astrofrog added this to the v1.0.0 milestone Nov 2, 2014

astrofrog added a commit that referenced this pull request Nov 3, 2014

Merge pull request #2964 from mdmueller/fix-mac-issue

0610e28

Sporadic test failure with fast io.ascii reader

astrofrog merged commit 0610e28 into astropy:master Nov 3, 2014

mdmueller deleted the fix-mac-issue branch November 3, 2014 19:28

Conversation

astrofrog commented Nov 2, 2014

Uh oh!

mdmueller commented Sep 19, 2014

Uh oh!

jfoster17 commented Sep 26, 2014

Uh oh!

mdmueller commented Sep 27, 2014

Uh oh!

astrofrog commented Sep 27, 2014

Uh oh!

astrofrog commented Sep 27, 2014

Uh oh!

mdmueller commented Sep 27, 2014

Uh oh!

mdmueller commented Oct 5, 2014

Uh oh!

astrofrog commented Oct 14, 2014

Uh oh!

astrofrog commented Oct 14, 2014

Uh oh!

mdmueller commented Nov 2, 2014

Uh oh!

astrofrog commented Nov 2, 2014

Uh oh!

astrofrog commented Nov 3, 2014

Uh oh!

mdmueller commented Nov 3, 2014

Uh oh!

astrofrog commented Nov 3, 2014

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants