feat(rust, python): add support for am/pm notation in parse_dates rea…

…d_csv (#5373)

polars/polars-time/src/chunkedarray/utf8/patterns.rs

@@ -24,7 +24,7 @@ pub(super) static DATETIME_D_M_Y: &[&str] = &[
     // --
     // supported by polars' parser
     // ---
-    // 31/12/21 12:54:98
+    // 31/12/21 12:54:48
     "%d/%m/%y %H:%M:%S",
     // 31-12-2021 24:58:01
     "%d-%m-%Y %H:%M:%S",
@@ -44,12 +44,18 @@ pub(super) static DATETIME_D_M_Y: &[&str] = &[
     // no times
     "%d-%m-%Y",
     "%d-%m-%y",
+    // 31/12/2021 11:54:48 PM
+    "%d/%m/%Y %I:%M:%S %p",
+    "%d-%m-%Y %I:%M:%S %p",
+    // 31/12/2021 11:54 PM
+    "%d/%m/%Y %I:%M %p",
+    "%d-%m-%Y %I:%M %p",
 ];
 
 /// NOTE: don't use single letter dates like %F
 /// polars parsers does not support them, so it will be slower
 pub(super) static DATETIME_Y_M_D: &[&str] = &[
-    // 21/12/31 12:54:98
+    // 21/12/31 12:54:48
     "%y/%m/%d %H:%M:%S",
     // 2021-12-31 24:58:01
     "%Y-%m-%d %H:%M:%S",
@@ -82,6 +88,12 @@ pub(super) static DATETIME_Y_M_D: &[&str] = &[
     // no times
     "%Y-%m-%d",
     "%Y/%m/%d",
+    // 2021/12/31 11:54:48 PM
+    "%Y/%m/%d %I:%M:%S %p",
+    "%Y-%m-%d %I:%M:%S %p",
+    // 2021/12/31 11:54 PM
+    "%Y/%m/%d %I:%M %p",
+    "%Y-%m-%d %I:%M %p",
     // --
     // not supported by polars' parser
     // ---

polars/tests/it/io/csv.rs

@@ -602,6 +602,31 @@ fn test_automatic_datetime_parsing() -> PolarsResult<()> {
     Ok(())
 }
 
+#[test]
+#[cfg(feature = "temporal")]
+fn test_automatic_datetime_parsing_default_formats() -> PolarsResult<()> {
+    let csv = r"ts_dmy,ts_dmy_f,ts_dmy_p
+01/01/21 00:00:00,31-01-2021T00:00:00.123,31-01-2021 11:00 AM
+01/01/21 00:15:00,31-01-2021T00:15:00.123,31-01-2021 01:00 PM
+01/01/21 00:30:00,31-01-2021T00:30:00.123,31-01-2021 01:15 PM
+01/01/21 00:45:00,31-01-2021T00:45:00.123,31-01-2021 01:30 PM
+";
+
+    let file = Cursor::new(csv);
+    let df = CsvReader::new(file).with_parse_dates(true).finish()?;
+
+    for col in df.get_column_names() {
+        let ts = df.column(col)?;
+        assert_eq!(
+            ts.dtype(),
+            &DataType::Datetime(TimeUnit::Microseconds, None)
+        );
+        assert_eq!(ts.null_count(), 0);
+    }
+
+    Ok(())
+}
+
 #[test]
 fn test_no_quotes() -> PolarsResult<()> {
     let rolling_stones = r#"linenum,last_name,first_name

py-polars/tests/unit/io/test_csv.py

@@ -126,6 +126,22 @@ def test_datetime_parsing() -> None:
     assert df.dtypes == [pl.Datetime, pl.Float64, pl.Float64]
 
 
+def test_datetime_parsing_default_formats() -> None:
+    csv = textwrap.dedent(
+        """\
+        ts_dmy,ts_dmy_f,ts_dmy_p
+        01/01/21 00:00:00,31-01-2021T00:00:00.123,31-01-2021 11:00 AM
+        01/01/21 00:15:00,31-01-2021T00:15:00.123,31-01-2021 01:00 PM
+        01/01/21 00:30:00,31-01-2021T00:30:00.123,31-01-2021 01:15 PM
+        01/01/21 00:45:00,31-01-2021T00:45:00.123,31-01-2021 01:30 PM
+        """
+    )
+
+    f = io.StringIO(csv)
+    df = pl.read_csv(f, parse_dates=True)
+    assert df.dtypes == [pl.Datetime, pl.Datetime, pl.Datetime]
+
+
 def test_partial_dtype_overwrite() -> None:
     csv = textwrap.dedent(
         """\