Fix timestamp conversion factor (#960) (#961)

Co-authored-by: Martin Durant <martin.durant@alumni.utoronto.ca>

Author: alexhudspith

fastparquet/test/test_timestamp_conversion_960.py

@@ -0,0 +1,23 @@
+import pandas as pd
+
+import fastparquet as fp
+
+def test_github_960(request, tmp_path):
+    """
+    https://github.com/dask/fastparquet/issues/960
+    """
+    file_path = tmp_path / f'{request.node.name}.parquet'
+    # Write datetime64[us] to TIMESTAMP_MICROS
+    df0 = pd.DataFrame({'Date': ['2025-01-01', pd.NaT]}, index=[0, 1], dtype='datetime64[us]')
+    fp.write(file_path, df0, append=False)
+
+    # Append datetime64[ns] to TIMESTAMP_MICROS
+    df1 = pd.DataFrame({'Date': ['2025-07-31', pd.NaT]}, index=[2, 3], dtype='datetime64[ns]')
+    fp.write(file_path, df1, append=True)
+
+    # Read back
+    pf = fp.ParquetFile(file_path)
+    df_actual = pf.to_pandas()
+
+    df_expected = pd.concat([df0, df1.astype('datetime64[us]')])
+    assert df_expected.equals(df_actual)

fastparquet/writer.py

@@ -321,10 +321,7 @@ def convert(data, se):
         else:
             unit = [k for k, v in se.logicalType.TIMESTAMP.unit._asdict().items() if v is not None][0]
             factor = time_factors[(unit, part)]
-        try:
-            out = data.values.view("int64") * factor
-        except KeyError:
-            breakpoint()
+        out = data.values.view("int64") // factor
     else:
         raise ValueError("Don't know how to convert data type: %s" % dtype)
     return out