Fix to_pandas() row-doubling when names is set and mcols has no matching row names

src/genomicranges/GenomicRanges.py

@@ -1006,13 +1006,13 @@ def to_pandas(self):
         _rdf["seqnames"] = self.get_seqnames()
         _rdf["strand"] = self.get_strand(as_type="list")
 
-        if self._names is not None:
-            _rdf.index = self._names
-
         if self._mcols is not None:
             if self._mcols.shape[1] > 0:
                 _rdf = pd.concat([_rdf, self._mcols.to_pandas()], axis=1)
 
+        if self._names is not None:
+            _rdf.index = self._names
+
         return _rdf
 
     @classmethod

tests/test_gr_initialize_pandas.py

@@ -1,6 +1,7 @@
 import pytest
 from genomicranges import GenomicRanges
 from biocframe import BiocFrame
+from iranges import IRanges
 from random import random
 import pandas as pd
 
@@ -41,3 +42,29 @@ def test_from_pandas_should_fail():
     )
     with pytest.raises(Exception):
         GenomicRanges.from_pandas(df_gr)
+
+
+def test_to_pandas_with_names_and_mcols_is_positional():
+    # Regression test for a bug where to_pandas() set the index to `names`
+    # before concatenating `mcols`, whose index was still the default
+    # positional RangeIndex. pandas.concat(axis=1) then outer-joined on the
+    # two disjoint indices instead of binding columns positionally, doubling
+    # the row count and NaN-splitting every row.
+    ranges = IRanges(start=[0, 10, 20], width=[5, 5, 5])
+    mcols = BiocFrame({"gene_id": ["g1", "g2", "g3"], "gene_name": ["A", "B", "C"]})
+    gr = GenomicRanges(
+        seqnames=["1", "1", "1"],
+        ranges=ranges,
+        strand=["+", "+", "-"],
+        names=["g1", "g2", "g3"],
+        mcols=mcols,
+    )
+
+    df = gr.to_pandas()
+
+    assert df.shape == (3, 7)
+    assert list(df.index) == ["g1", "g2", "g3"]
+    assert df["gene_id"].notna().all()
+    assert df["seqnames"].notna().all()
+    assert list(df["gene_id"]) == ["g1", "g2", "g3"]
+    assert list(df["gene_name"]) == ["A", "B", "C"]