Fix crossmatch for empty left partition (#315)

src/lsdb/dask/crossmatch_catalog_data.py

@@ -42,6 +42,7 @@ def perform_crossmatch(
     algorithm,
     suffixes,
     right_columns,
+    meta_df,
     **kwargs,
 ):
     """Performs a crossmatch on data from a HEALPix pixel in each catalog
@@ -52,6 +53,9 @@ def perform_crossmatch(
     if right_pix.order > left_pix.order:
         left_df = filter_by_hipscat_index_to_pixel(left_df, right_pix.order, right_pix.pixel)
 
+    if len(left_df) == 0:
+        return meta_df
+
     right_joined_df = concat_partition_and_margin(right_df, right_margin_df, right_columns)
 
     return algorithm(
@@ -125,22 +129,22 @@ def crossmatch_catalog_data(
     # get lists of HEALPix pixels from alignment to pass to cross-match
     left_pixels, right_pixels = get_healpix_pixels_from_alignment(alignment)
 
-    # perform the crossmatch on each partition pairing using dask delayed for lazy computation
+    # generate meta table structure for dask df
+    meta_df = generate_meta_df_for_joined_tables(
+        [left, right], suffixes, extra_columns=crossmatch_algorithm.extra_columns
+    )
 
+    # perform the crossmatch on each partition pairing using dask delayed for lazy computation
     joined_partitions = align_and_apply(
         [(left, left_pixels), (right, right_pixels), (right.margin, right_pixels)],
         perform_crossmatch,
         crossmatch_algorithm,
         suffixes,
         right.columns,
+        meta_df,
         **kwargs,
     )
 
-    # generate meta table structure for dask df
-    meta_df = generate_meta_df_for_joined_tables(
-        [left, right], suffixes, extra_columns=crossmatch_algorithm.extra_columns
-    )
-
     return construct_catalog_args(joined_partitions, meta_df, alignment)
 
 

tests/lsdb/catalog/test_crossmatch.py

@@ -211,6 +211,22 @@ def test_self_crossmatch(algo, small_sky_catalog, small_sky_dir):
         assert len(xmatched) == len(small_sky_catalog.compute())
         assert all(xmatched["_dist_arcsec"] == 0)
 
+    @staticmethod
+    def test_crossmatch_empty_left_partition(algo, small_sky_order1_catalog, small_sky_xmatch_catalog):
+        ra = 300
+        dec = -60
+        radius_arcsec = 1 * 3600
+        cone = small_sky_order1_catalog.cone_search(ra, dec, radius_arcsec)
+        assert len(cone.get_healpix_pixels()) == 2
+        assert len(cone.get_partition(1, 44)) == 2
+        # There is an empty partition in the left catalog
+        assert len(cone.get_partition(1, 46)) == 0
+        xmatched = cone.crossmatch(
+            small_sky_xmatch_catalog, radius_arcsec=0.01 * 3600, algorithm=algo
+        ).compute()
+        assert len(xmatched) == 2
+        assert all(xmatched["_dist_arcsec"] <= 0.01 * 3600)
+
 
 # pylint: disable=too-few-public-methods
 class MockCrossmatchAlgorithm(AbstractCrossmatchAlgorithm):