This is an automated email from the ASF dual-hosted git repository.
JingsongLi pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/paimon.git
The following commit(s) were added to refs/heads/master by this push:
new 676b40c91f [python] Match newlines with the LIKE single-character
wildcard (#10847)
676b40c91f is described below
commit 676b40c91f17b621f7bb08e4c70c779e765749cd
Author: chenjian2664 <[email protected]>
AuthorDate: Wed Oct 28 20:00:35 2026 +0800
[python] Match newlines with the LIKE single-character wildcard (#10847)
---
paimon-python/pypaimon/common/predicate.py | 2 +-
paimon-python/pypaimon/tests/predicates_test.py | 40 +++++++++++++++++++++++++
2 files changed, 41 insertions(+), 1 deletion(-)
diff --git a/paimon-python/pypaimon/common/predicate.py
b/paimon-python/pypaimon/common/predicate.py
index 6bd6801e6f..80f1d0b91e 100644
--- a/paimon-python/pypaimon/common/predicate.py
+++ b/paimon-python/pypaimon/common/predicate.py
@@ -601,7 +601,7 @@ class Like(Tester):
raise ValueError(
f"Invalid escape sequence '{pattern}' at position
{index}")
elif char == '_':
- regex_parts.append('.')
+ regex_parts.append('(?s:.)')
elif char == '%':
regex_parts.append('(?s:.*)')
else:
diff --git a/paimon-python/pypaimon/tests/predicates_test.py
b/paimon-python/pypaimon/tests/predicates_test.py
index 3e07631c76..684ea038d5 100644
--- a/paimon-python/pypaimon/tests/predicates_test.py
+++ b/paimon-python/pypaimon/tests/predicates_test.py
@@ -442,6 +442,33 @@ class PredicateTest(unittest.TestCase):
predicate = predicate_builder.like('f1', '%bc')
_check_filtered_result(table.new_read_builder().with_filter(predicate),
self.df.loc[0:2])
+ @pytest.mark.python_plan
+ def test_like_newline_pk_merge(self):
+ pa_schema = pa.schema([('f0', pa.int64()), ('f1', pa.string())])
+ self.catalog.create_table('default.test_like_newline',
Schema.from_pyarrow_schema(
+ pa_schema, primary_keys=['f0'], options={
+ 'bucket': '1', 'file.format': 'parquet',
+ 'scan.native-plan.enabled': 'false', 'read.native.enabled':
'false',
+ }), False)
+ table = self.catalog.get_table('default.test_like_newline')
+ df = pd.DataFrame({
+ 'f0': [1, 2, 3, 4, 5],
+ 'f1': ['a\nc', 'abc', 'a\r\nc', 'ac', None],
+ })
+ # A second commit overlaps the first file and requires row-by-row
filtering.
+ for data in (df, df.loc[[0]]):
+ builder = table.new_batch_write_builder()
+ write, commit = builder.new_write(), builder.new_commit()
+ try:
+ write.write_pandas(data)
+ commit.commit(write.prepare_commit())
+ finally:
+ write.close()
+ commit.close()
+ read_builder = table.new_read_builder()
+ predicate = read_builder.new_predicate_builder().like('f1', 'a_c')
+ _check_filtered_result(read_builder.with_filter(predicate),
df.loc[[0, 1]])
+
def test_null_literal_matches_nothing(self):
pa_schema = pa.schema([('f0', pa.int64()), ('f1', pa.string())])
# without stats no file is pruned, so the Arrow filter decides
@@ -647,6 +674,19 @@ class PredicateTest(unittest.TestCase):
self.assertTrue(underscore_pred.test(OffsetRow(['abc'], 0, 1)))
self.assertFalse(underscore_pred.test(OffsetRow(['aXYc'], 0, 1)))
+ def test_like_single_character_wildcard_matches_line_terminators(self):
+ predicate = Predicate(method='like', index=0, field='name',
literals=['a_c'])
+ for terminator in ('\n', '\r', '\u0085', '\u2028', '\u2029'):
+ with self.subTest(terminator=terminator):
+ self.assertTrue(predicate.test(OffsetRow(['a' + terminator +
'c'], 0, 1)))
+
+ self.assertFalse(predicate.test(OffsetRow(['a\r\nc'], 0, 1)))
+ two_characters = predicate.new_literals(['a__c'])
+ self.assertTrue(two_characters.test(OffsetRow(['a\r\nc'], 0, 1)))
+ escaped = predicate.new_literals([r'a\_c'])
+ self.assertTrue(escaped.test(OffsetRow(['a_c'], 0, 1)))
+ self.assertFalse(escaped.test(OffsetRow(['a\nc'], 0, 1)))
+
def test_not_between_value(self):
predicate = Predicate(method='notBetween', index=0, field='val',
literals=[3, 7])
self.assertTrue(predicate.test(OffsetRow([1], 0, 1)))
