Tests failing due to new awkward release?
rkansal47 opened this issue · comments
Some weird tests failures now [1, 2].
@lgray @jpivarski do the tests perhaps need to be modified because of awkward going from 2.3.1 -> 2.3.3?
They appear to have passed 3 weeks ago with awkward 2.3.1: #244
[1] tests/test_002-exclusive_jets.py::test_listoffset_indexed_input
def test_listoffset_indexed_input():
inputs = ak.Array(
[
[
{"px": 1.2, "py": 3.2, "pz": 5.4, "E": 2.5, "ex": 0.78},
{"px": 32.2, "py": 64.21, "pz": 543.34, "E": 24.12, "ex": 0.35},
{"px": 32.45, "py": 63.21, "pz": 543.14, "E": 24.56, "ex": 0.0},
],
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5, "ex": 0.78},
{"px": 32.2, "py": 64.21, "pz": 543.34, "E": 24.12, "ex": 0.35},
{"px": 32.45, "py": 63.21, "pz": 543.14, "E": 24.56, "ex": 0.0},
],
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5, "ex": 0.78},
{"px": 32.2, "py": 64.21, "pz": 543.34, "E": 24.12, "ex": 0.35},
{"px": 32.45, "py": 63.21, "pz": 543.14, "E": 24.56, "ex": 0.1},
],
],
with_name="Momentum4D",
)
out = ak.Array(
ak.contents.RecordArray(
[
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).px)),
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).py)),
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).pz)),
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).E)),
],
["px", "py", "pz", "E"],
)
)
out = ak.Array(
ak.contents.IndexedArray(
ak.index.Index64([7, 2, 3, 1, 0, 5, 4, 6, 8]), out.layout
)
)
out = ak.Array(ak.contents.ListOffsetArray(inputs.layout.offsets, out.layout))
jetdef = fastjet.JetDefinition(fastjet.antikt_algorithm, 0.6)
cluster = fastjet._pyjet.AwkwardClusterSequence(out, jetdef)
inclusive_jets = [
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5},
{"px": 64.65, "py": 127.41999999999999, "pz": 1086.48, "E": 48.68},
],
[
{"px": 1.2, "py": 3.2, "pz": 5.4, "E": 2.5},
{"px": 64.65, "py": 127.41999999999999, "pz": 1086.48, "E": 48.68},
],
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5},
{"px": 64.65, "py": 127.41999999999999, "pz": 1086.48, "E": 48.68},
],
]
> assert inclusive_jets == cluster.inclusive_jets().to_list()
E AssertionError: assert [[{'px': 11.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5}, {'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}], [{'px': 1.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5}, {'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}], [{'px': 11.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5}, {'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}]] == [[{'px': 6e-323, 'py': 51.18, 'pz': 6e-323, 'E': 51.18}], [{'px': 3e-323, 'py': 51.18, 'pz': 3e-323, 'E': 51.18}], [{'px': 9e-323, 'py': 51.18, 'pz': 9e-323, 'E': 51.18}]][2] tests/test_007-general.py::test_indexed_subtree_input
def test_indexed_subtree_input():
inputs = ak.Array(
[
[
{"px": 1.2, "py": 3.2, "pz": 5.4, "E": 2.5, "ex": 0.78},
{"px": 32.2, "py": 64.21, "pz": 543.34, "E": 24.12, "ex": 0.35},
{"px": 32.45, "py": 63.21, "pz": 543.14, "E": 24.56, "ex": 0.0},
],
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5, "ex": 0.78},
{"px": 32.2, "py": 64.21, "pz": 543.34, "E": 24.12, "ex": 0.35},
{"px": 32.45, "py": 63.21, "pz": 543.14, "E": 24.56, "ex": 0.0},
],
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5, "ex": 0.78},
{"px": 32.2, "py": 64.21, "pz": 543.34, "E": 24.12, "ex": 0.35},
{"px": 32.45, "py": 63.21, "pz": 543.14, "E": 24.56, "ex": 0.1},
],
],
with_name="Momentum4D",
)
out = ak.Array(
ak.contents.RecordArray(
[
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).px)),
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).py)),
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).pz)),
ak.contents.NumpyArray(np.asarray(ak.Array(inputs.layout.content).E)),
],
["px", "py", "pz", "E"],
)
)
out = ak.Array(
ak.contents.IndexedArray(
ak.index.Index64([7, 2, 3, 1, 0, 5, 4, 6, 8]), out.layout
)
)
out = ak.Array(ak.contents.ListOffsetArray(inputs.layout.offsets, out.layout))
out = ak.Array(
ak.contents.ListOffsetArray(ak.index.Index64([0, 1, 2, 3]), out.layout)
)
out = ak.Array(
ak.contents.ListOffsetArray(ak.index.Index64([0, 1, 2, 3]), out.layout)
)
out = ak.Array(
ak.contents.ListOffsetArray(ak.index.Index64([0, 1, 2, 3]), out.layout)
)
jetdef = fastjet.JetDefinition(fastjet.antikt_algorithm, 0.6)
cluster = fastjet.ClusterSequence(out, jetdef)
inclusive_jets = [
[
[
[
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5},
{
"px": 64.65,
"py": 127.41999999999999,
"pz": 1086.48,
"E": 48.68,
},
]
]
]
],
[
[
[
[
{"px": 1.2, "py": 3.2, "pz": 5.4, "E": 2.5},
{
"px": 64.65,
"py": 127.41999999999999,
"pz": 1086.48,
"E": 48.68,
},
]
]
]
],
[
[
[
[
{"px": 11.2, "py": 3.2, "pz": 5.4, "E": 2.5},
{
"px": 64.65,
"py": 127.41999999999999,
"pz": 1086.48,
"E": 48.68,
},
]
]
]
],
]
> assert inclusive_jets == cluster.inclusive_jets().to_list()
E AssertionError: assert [[[[[{'px': 11.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5}, {'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}]]]], [[[[{'px': 1.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5}, {'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}]]]], [[[[{'px': 11.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5}, {'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}]]]]] == [[[[[{'px': 6e-323, 'py': 51.18, 'pz': 6e-323, 'E': 51.18}]]]], [[[[{'px': 3e-323, 'py': 51.18, 'pz': 3e-323, 'E': 51.18}]]]], [[[[{'px': 9e-323, 'py': 51.18, 'pz': 9e-323, 'E': 51.18}]]]]]
I don't see anything in the last few releases that would touch something so low-level: it's looking at memory and picking up random junk. In the last test, quoted above, the expected result
[[[[[
{'px': 11.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5},
{'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}]
]]], [[[[
{'px': 1.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5},
{'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}
]]]], [[[[
{'px': 11.2, 'py': 3.2, 'pz': 5.4, 'E': 2.5},
{'px': 64.65, 'py': 127.41999999999999, 'pz': 1086.48, 'E': 48.68}
]]]]]came out as
[
[[[[{'px': 6e-323, 'py': 51.18, 'pz': 6e-323, 'E': 51.18}]]]],
[[[[{'px': 3e-323, 'py': 51.18, 'pz': 3e-323, 'E': 51.18}]]]],
[[[[{'px': 9e-323, 'py': 51.18, 'pz': 9e-323, 'E': 51.18}]]]],
]It has the same depth of structure (the nested Python objects), but there are wrong offsets here (list lengths are different) and wrong values (6e-323 is not a plausible value, and 51.18 seems to be duplicated).
It looks like this is probably caused by a bug in the array-interface implementation, which is fixed by scikit-hep/awkward#2697.
We'll get another release out, and see what happens!
Great, looks these tests are now passing! However, there's one failure [1] now I guess because dask-awkward added pyarrow as a dependency for its testing. I'll add dask-awkward[test] as a dependency for our tests to resolve this, unless @lgray (or others) have better suggestions.
[1]
___________________________ test_inclusive_from_file ___________________________
def test_inclusive_from_file():
from pathlib import Path
import uproot
> from dask_awkward.lib.testutils import assert_eq
tests/test_008-dask.py:103:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
from __future__ import annotations
import random
from typing import Any
import awkward as ak
import numpy as np
> import pyarrow as pa
E ModuleNotFoundError: No module named 'pyarrow'
../../../hostedtoolcache/Python/3.8.17/x64/lib/python3.8/site-packages/dask_awkward/lib/testutils.py:8: ModuleNotFoundError