From 3f7c1904b3b6a243684f899d9bf322fd5f2b09f2 Mon Sep 17 00:00:00 2001
From: Julian <juliansteden@gmail.com>
Date: Sat, 28 Oct 2023 12:30:38 +0200
Subject: [PATCH] fix(python,rust): str.concat on empty list (#12066)

---
 .../src/chunked_array/strings/concat.rs       |  6 ++++-
 .../unit/namespaces/string/test_string.py     | 22 +++++++++++++++++++
 2 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/crates/polars-ops/src/chunked_array/strings/concat.rs b/crates/polars-ops/src/chunked_array/strings/concat.rs
index 781e1da2175c..16269f06d5c1 100644
--- a/crates/polars-ops/src/chunked_array/strings/concat.rs
+++ b/crates/polars-ops/src/chunked_array/strings/concat.rs
@@ -4,7 +4,11 @@ use polars_core::prelude::*;
 
 // Vertically concatenate all strings in a Utf8Chunked.
 pub fn str_concat(ca: &Utf8Chunked, delimiter: &str) -> Utf8Chunked {
-    if ca.len() <= 1 {
+    if ca.is_empty() {
+        return Utf8Chunked::new(ca.name(), &[""]);
+    }
+
+    if ca.len() == 1 {
         return ca.clone();
     }
 
diff --git a/py-polars/tests/unit/namespaces/string/test_string.py b/py-polars/tests/unit/namespaces/string/test_string.py
index fb2039ab33c3..db2526857470 100644
--- a/py-polars/tests/unit/namespaces/string/test_string.py
+++ b/py-polars/tests/unit/namespaces/string/test_string.py
@@ -28,6 +28,28 @@ def test_str_concat2() -> None:
     assert cast(str, df.item()) == "1-null-2"
 
 
+def test_str_concat_empty_list() -> None:
+    s = pl.Series([], dtype=pl.Utf8)
+    result = s.str.concat()
+    expected = pl.Series([""])
+    assert_series_equal(result, expected)
+
+
+def test_str_concat_empty_list2() -> None:
+    s = pl.Series([], dtype=pl.Utf8)
+    df = pl.DataFrame({"foo": s})
+    result = df.select(pl.col("foo").str.concat()).item()
+    expected = ""
+    assert result == expected
+
+
+def test_str_concat_empty_list_agg_context() -> None:
+    df = pl.DataFrame(data={"i": [1], "v": [None]}, schema_overrides={"v": pl.Utf8})
+    result = df.group_by("i").agg(pl.col("v").drop_nulls().str.concat())["v"].item()
+    expected = ""
+    assert result == expected
+
+
 def test_str_concat_datetime() -> None:
     df = pl.DataFrame({"d": [datetime(2020, 1, 1), None, datetime(2022, 1, 1)]})
     df = df.select(pl.col("d").str.concat("|"))