Skip to content

Read csv category fix #18402

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Nov 22, 2017
Merged

Read csv category fix #18402

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bf61eaf
Fix bug in read_csv for high cardinality category types (#18186)
sam-cohan Nov 21, 2017
ff1945d
Refactor read_csv bug fix with PR comments (#18186)
sam-cohan Nov 21, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions doc/source/whatsnew/v0.21.1.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,6 +82,7 @@ I/O
- Bug in class:`~pandas.io.stata.StataReader` not converting date/time columns with display formatting addressed (:issue:`17990`). Previously columns with display formatting were normally left as ordinal numbers and not converted to datetime objects.
- Bug in :func:`read_csv` when reading a compressed UTF-16 encoded file (:issue:`18071`)
- Bug in :func:`read_csv` for handling null values in index columns when specifying ``na_filter=False`` (:issue:`5239`)
- Bug in :func:`read_csv` when reading numeric category fields with high cardinality (:issue:`18186`)
- Bug in :meth:`DataFrame.to_csv` when the table had ``MultiIndex`` columns, and a list of strings was passed in for ``header`` (:issue:`5539`)
- :func:`read_parquet` now allows to specify the columns to read from a parquet file (:issue:`18154`)
- :func:`read_parquet` now allows to specify kwargs which are passed to the respective engine (:issue:`18216`)
Expand Down
7 changes: 4 additions & 3 deletions pandas/_libs/parsers.pyx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2227,9 +2227,10 @@ def _concatenate_chunks(list chunks):
for name in names:
arrs = [chunk.pop(name) for chunk in chunks]
# Check each arr for consistent types.
dtypes = set(a.dtype for a in arrs)
if len(dtypes) > 1:
common_type = np.find_common_type(dtypes, [])
dtypes = {a.dtype for a in arrs}
numpy_dtypes = {x for x in dtypes if not is_categorical_dtype(x)}
if len(numpy_dtypes) > 1:
common_type = np.find_common_type(numpy_dtypes, [])
if common_type == np.object:
warning_columns.append(str(name))

Expand Down
11 changes: 11 additions & 0 deletions pandas/tests/io/parser/dtypes.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,6 +114,17 @@ def test_categorical_dtype(self):
actual = self.read_csv(StringIO(data), dtype='category')
tm.assert_frame_equal(actual, expected)

@pytest.mark.slow
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

how slow is this test?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

On my machine it is about 4.5 seconds for the high memory parser, and 6.5 seconds for low memory and python parsers.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The minimal range necessary to reproduce the error is range(524289), at least locally for me. Might be beneficial to lower the range in the test to offset some of the slowness? If so, not sure if we should push it down to the limit, or just go down to something like 600k to leave a little buffer room.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Checked and the limit is the same on both ubuntu and OSX so decided to do it as just the limit. This cut down the times to 2.25 seconds and 3 seconds so I removed the slow mark.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added slow mark back. rebased. waiting for tests.

def test_categorical_dtype_high_cardinality_numeric(self):
# GH 18186
data = np.sort([str(i) for i in range(524289)])
expected = DataFrame({'a': Categorical(data, ordered=True)})
actual = self.read_csv(StringIO('a\n' + '\n'.join(data)),
dtype='category')
actual["a"] = actual["a"].cat.reorder_categories(
np.sort(actual.a.cat.categories), ordered=True)
tm.assert_frame_equal(actual, expected)

def test_categorical_dtype_encoding(self):
# GH 10153
pth = tm.get_data_path('unicode_series.csv')
Expand Down