From e1dd3c9f4f14bce514542f8985eab296c9a4e7fb Mon Sep 17 00:00:00 2001
From: Simon Lin <simonlin.rqmmw@slmail.me>
Date: Wed, 31 Jul 2024 08:36:37 +1000
Subject: [PATCH] fix: Incorrect filter on categorical columns from parquet
 files

---
 crates/polars-io/src/predicates.rs           |  3 ++-
 py-polars/tests/unit/io/test_lazy_parquet.py | 25 ++++++++++++++++++++
 2 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/crates/polars-io/src/predicates.rs b/crates/polars-io/src/predicates.rs
index ad20aa102a9f..be3bf0ebff03 100644
--- a/crates/polars-io/src/predicates.rs
+++ b/crates/polars-io/src/predicates.rs
@@ -190,7 +190,8 @@ impl ColumnStats {
 
 /// Returns whether the [`DataType`] supports minimum/maximum operations.
 fn use_min_max(dtype: &DataType) -> bool {
-    dtype.to_physical().is_numeric()
+    dtype.is_numeric()
+        || dtype.is_temporal()
         || matches!(
             dtype,
             DataType::String | DataType::Binary | DataType::Boolean
diff --git a/py-polars/tests/unit/io/test_lazy_parquet.py b/py-polars/tests/unit/io/test_lazy_parquet.py
index 6c4b2842ee13..7b7350bdf496 100644
--- a/py-polars/tests/unit/io/test_lazy_parquet.py
+++ b/py-polars/tests/unit/io/test_lazy_parquet.py
@@ -450,3 +450,28 @@ def test_parquet_schema_mismatch_panic_17067(tmp_path: Path, streaming: bool) ->
 
     with pytest.raises(pl.exceptions.SchemaError):
         pl.scan_parquet(tmp_path).collect(streaming=streaming)
+
+
+@pytest.mark.write_disk()
+def test_predicate_push_down_categorical_17744(tmp_path: Path) -> None:
+    path = tmp_path / "1"
+
+    df = pl.DataFrame(
+        data={
+            "n": [1, 2, 3],
+            "ccy": ["USD", "JPY", "EUR"],
+        },
+        schema_overrides={"ccy": pl.Categorical("lexical")},
+    )
+    df.write_parquet(path)
+    expect = df.head(1).with_columns(pl.col(pl.Categorical).cast(pl.String))
+
+    lf = pl.scan_parquet(path)
+
+    for predicate in [pl.col("ccy") == "USD", pl.col("ccy").is_in(["USD"])]:
+        assert_frame_equal(
+            lf.filter(predicate)
+            .with_columns(pl.col(pl.Categorical).cast(pl.String))
+            .collect(),
+            expect,
+        )