Skip to content
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

Make string methods return a Series with a useful Index #12814

Merged
merged 6 commits into from
Mar 9, 2023
Merged

Make string methods return a Series with a useful Index #12814

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
50f1bf5
Make character_ngrams return the right index
shwina Feb 21, 2023
bea690c
Merge branch 'branch-23.04' of github.com:rapidsai/cudf into HEAD
shwina Mar 7, 2023
b8e7cd5
Fix index for character_tokenize
shwina Mar 7, 2023
55578e1
tokenize
shwina Mar 7, 2023
b0ec6de
docs
shwina Mar 7, 2023
9907096
Merge branch 'branch-23.04' into fix-str-methods-index
shwina Mar 9, 2023
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
109 changes: 59 additions & 50 deletions python/cudf/cudf/core/column/string.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4564,22 +4564,22 @@ def tokenize(self, delimiter: str = " ") -> SeriesOrIndex:
>>> ser = cudf.Series(data)
>>> ser.str.tokenize()
0 hello
0 world
1 goodbye
1 world
2 hello
2 goodbye
3 world
4 hello
5 goodbye
dtype: object
"""
delimiter = _massage_string_arg(delimiter, "delimiter", allow_col=True)

if isinstance(delimiter, Column):
return self._return_or_inplace(
result = self._return_or_inplace(
libstrings._tokenize_column(self._column, delimiter),
retain_index=False,
)
elif isinstance(delimiter, cudf.Scalar):
return self._return_or_inplace(
result = self._return_or_inplace(
libstrings._tokenize_scalar(self._column, delimiter),
retain_index=False,
)
Expand All @@ -4588,6 +4588,11 @@ def tokenize(self, delimiter: str = " ") -> SeriesOrIndex:
f"Expected a Scalar or Column\
for delimiters, but got {type(delimiter)}"
)
if isinstance(self._parent, cudf.Series):
result.index = self._parent.index.repeat( # type: ignore
self.token_count()
)
return result

def detokenize(
self, indices: "cudf.Series", separator: str = " "
Expand Down Expand Up @@ -4641,41 +4646,43 @@ def character_tokenize(self) -> SeriesOrIndex:
>>> data = ["hello world", None, "goodbye, thank you."]
>>> ser = cudf.Series(data)
>>> ser.str.character_tokenize()
0 h
1 e
2 l
3 l
4 o
5
6 w
7 o
8 r
9 l
10 d
11 g
12 o
13 o
14 d
15 b
16 y
17 e
18 ,
19
20 t
21 h
22 a
23 n
24 k
25
26 y
27 o
28 u
29 .
0 h
0 e
0 l
0 l
0 o
0
0 w
0 o
0 r
0 l
0 d
2 g
2 o
2 o
2 d
2 b
2 y
2 e
2 ,
2
2 t
2 h
2 a
2 n
2 k
2
2 y
2 o
2 u
2 .
dtype: object
"""
result_col = libstrings.character_tokenize(self._column)
if isinstance(self._parent, cudf.Series):
return cudf.Series(result_col, name=self._parent.name)
lengths = self.len().fillna(0)
Copy link
Contributor

Choose a reason for hiding this comment

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

Is the fillna(0) matching pandas behavior? It looks like the test is constructing the expected output manually to be the same, so it's not clear if it's going to match pandas or if the test is just constructing the same as what this outputs.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

There is no Pandas behaviour (Pandas doesn't have a tokenize).

The fillna(0) is there because nulls are ignored during tokenization, corresponding to 0 rows in the result (and thus, length = 0).

index = self._parent.index.repeat(lengths)
return cudf.Series(result_col, name=self._parent.name, index=index)
elif isinstance(self._parent, cudf.BaseIndex):
return cudf.core.index.as_index(result_col, name=self._parent.name)
else:
Expand Down Expand Up @@ -4780,20 +4787,20 @@ def character_ngrams(
>>> str_series = cudf.Series(['abcd','efgh','xyz'])
>>> str_series.str.character_ngrams(2)
0 ab
1 bc
2 cd
3 ef
4 fg
5 gh
6 xy
7 yz
0 bc
0 cd
1 ef
1 fg
1 gh
2 xy
2 yz
dtype: object
>>> str_series.str.character_ngrams(3)
0 abc
1 bcd
2 efg
3 fgh
4 xyz
0 bcd
1 efg
1 fgh
2 xyz
dtype: object
>>> str_series.str.character_ngrams(3,True)
0 [abc, bcd]
Expand All @@ -4802,8 +4809,6 @@ def character_ngrams(
dtype: list
"""
ngrams = libstrings.generate_character_ngrams(self._column, n)
if as_list is False:
return self._return_or_inplace(ngrams, retain_index=False)

# convert the output to a list by just generating the
# offsets for the output list column
Expand All @@ -4820,7 +4825,11 @@ def character_ngrams(
null_count=self._column.null_count,
children=(oc, ngrams),
)
return self._return_or_inplace(lc, retain_index=False)
result = self._return_or_inplace(lc, retain_index=True)

if isinstance(result, cudf.Series) and not as_list:
return result.explode()
return result

def ngrams_tokenize(
self, n: int = 2, delimiter: str = " ", separator: str = "_"
Expand Down
40 changes: 32 additions & 8 deletions python/cudf/cudf/tests/test_text.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
# Copyright (c) 2019-2022, NVIDIA CORPORATION.
# Copyright (c) 2019-2023, NVIDIA CORPORATION.

from io import StringIO

Expand All @@ -24,7 +24,7 @@ def test_tokenize():
]
)

expected = cudf.Series(
expected_values = cudf.Series(
[
"the",
"quick",
Expand All @@ -43,6 +43,8 @@ def test_tokenize():
"sofa",
]
)
expected_index = strings.index.repeat(strings.str.token_count())
expected = cudf.Series(expected_values, index=expected_index)

actual = strings.str.tokenize()

Expand Down Expand Up @@ -231,7 +233,7 @@ def test_ngrams(n, separator, expected_values):


@pytest.mark.parametrize(
"n, expected_values, as_list",
"n, expected_values, expected_index, as_list",
[
(
2,
Expand All @@ -247,21 +249,41 @@ def test_ngrams(n, separator, expected_values):
"he",
"er",
"re",
cudf.NA,
],
[1, 1, 1, 2, 3, 4, 4, 4, 5, 5, 5, 6],
False,
),
(
3,
[
"thi",
"his",
cudf.NA,
cudf.NA,
"boo",
"ook",
"her",
"ere",
cudf.NA,
],
[1, 1, 2, 3, 4, 4, 5, 5, 6],
False,
),
(3, ["thi", "his", "boo", "ook", "her", "ere"], False),
(
3,
[["thi", "his"], [], [], ["boo", "ook"], ["her", "ere"], []],
[1, 2, 3, 4, 5, 6],
True,
),
],
)
def test_character_ngrams(n, expected_values, as_list):
strings = cudf.Series(["this", "is", "my", "book", "here", ""])
def test_character_ngrams(n, expected_values, expected_index, as_list):
strings = cudf.Series(
["this", "is", "my", "book", "here", ""], index=[1, 2, 3, 4, 5, 6]
)

expected = cudf.Series(expected_values)
expected = cudf.Series(expected_values, index=expected_index)

actual = strings.str.character_ngrams(n=n, as_list=as_list)

Expand Down Expand Up @@ -314,7 +336,7 @@ def test_character_tokenize_series():
),
]
)
expected = cudf.Series(
expected_values = cudf.Series(
[
"h",
"e",
Expand Down Expand Up @@ -402,6 +424,8 @@ def test_character_tokenize_series():
"DŽ",
]
)
expected_index = sr.index.repeat(sr.str.len().fillna(0))
expected = cudf.Series(expected_values, index=expected_index)

actual = sr.str.character_tokenize()
assert_eq(expected, actual)
Expand Down