Modify profile viewer so that when in 'Sum' mode, parts of profiles with no valid values are NaN rather than zero

CHANGES.md

@@ -1,6 +1,12 @@
 Full changelog
 ==============
 
+v1.4.0 (unreleased)
+-------------------
+
+* Modify profile viewer so that when in 'Sum' mode, parts of profiles
+  with no valid values are NaN rather than zero.
+
 v1.3.0 (2022-04-22)
 -------------------
 

glue/utils/array.py

@@ -396,6 +396,23 @@ def format_minimal(values):
     return fmt, strings
 
 
+def nansum_with_nan_for_empty(values, axis=None):
+    """
+    Wrapper around nansum which returns NaN instead of zero when no non-NaN
+    values are summed.
+    """
+    result = np.nansum(values, axis=axis)
+    reset = np.sum(~np.isnan(values), axis=axis)
+    if np.isscalar(result):
+        if reset == 0:
+            return np.nan
+        else:
+            return result
+    else:
+        result[reset == 0] = np.nan
+        return result
+
+
 PLAIN_FUNCTIONS = {'minimum': np.min,
                    'maximum': np.max,
                    'mean': np.mean,
@@ -407,7 +424,7 @@ def format_minimal(values):
                  'maximum': np.nanmax,
                  'mean': np.nanmean,
                  'median': np.nanmedian,
-                 'sum': np.nansum,
+                 'sum': nansum_with_nan_for_empty,
                  'percentile': np.nanpercentile}
 
 

glue/utils/tests/test_array.py

@@ -10,7 +10,7 @@
                      shape_to_string, check_sorted, pretty_number, unbroadcast,
                      iterate_chunks, combine_slices, nanmean, nanmedian, nansum,
                      nanmin, nanmax, format_minimal, compute_statistic, categorical_ndarray,
-                     index_lookup, broadcast_arrays_minimal)
+                     index_lookup, broadcast_arrays_minimal, nansum_with_nan_for_empty)
 
 
 @pytest.mark.parametrize(('before', 'ref_after', 'ref_indices'),
@@ -194,6 +194,14 @@ def test_nanfunctions(function, axis):
     assert_allclose(function(ARRAY, axis=axis), np_func(ARRAY, axis=axis))
 
 
+def test_custom_nansum():
+    assert np.isnan(nansum_with_nan_for_empty([]))
+    assert_equal(nansum_with_nan_for_empty([np.nan]), np.nan)
+    assert_equal(nansum_with_nan_for_empty([[np.nan, np.nan], [np.nan, 1]]), 1)
+    assert_equal(nansum_with_nan_for_empty([[np.nan, np.nan], [1, np.nan]], axis=0), [1, np.nan])
+    assert_equal(nansum_with_nan_for_empty([[np.nan, np.nan], [1, np.nan]], axis=1), [np.nan, 1])
+
+
 SLICE_CASES = [
     (slice(None), slice(5), 10),
     (slice(None), slice(1, 5), 10),