Use integers for under/over-flow attributes

otherwise NetCDF will complain

Author: Descanonge

changelog.md

@@ -1,4 +1,8 @@
 
+### 0.2.1
+
+- Fix: under/over flow attributes are int instead of bool to conform with NetCDF
+
 ## 0.2.0
 
 - Use common function for normalization in histogramdd and accessor

docs/source/accessor.rst

@@ -53,8 +53,13 @@ Each bins coordinate may contain attributes:
   to infer the rightmost edge.
 * ``right_edge``: the rightmost edge position, only necessary for Regular and
   Variable bins.
-* ``underflow`` and ``overflow``: booleans that indicate if the corresponding
-  flow bins are present. If not present, will assume no flow bins.
+* ``underflow`` and ``overflow``: integers that indicate if the corresponding
+  flow bins are present (0: not present, 1: present). If not specified, will
+  assume no flow bins.
+
+  .. note::
+
+     We use integers instead of booleans to conform with NetCDF attributes.
 
 Those conventions are coherent with the output of
 ``xarray_histogram.histogram*``, so if you use this package functions you

src/xarray_histogram/accessor.py

@@ -36,7 +36,7 @@ class HistDataArrayAccessor:
       rightmost edge.
     * ``right_edge``: the rightmost edge position, only necessary for Regular and
        Variable bins.
-    * ``underflow`` and ``overflow``: booleans that indicate if the corresponding flow
+    * ``underflow`` and ``overflow``: integers that indicate if the corresponding flow
       bins are present. If not present, will assume no flow bins.
 
     .. rubric:: Backend for computations
@@ -578,6 +578,6 @@ def remove_flow(coord: xr.DataArray) -> xr.DataArray:
     overflow = coord.attrs.get("overflow", False)
     underflow = coord.attrs.get("underflow", False)
     out = coord[slice(1 if underflow else 0, -1 if overflow else None)]
-    out.attrs["underflow"] = False
-    out.attrs["overflow"] = True
+    out.attrs["underflow"] = 0
+    out.attrs["overflow"] = 0
     return out

src/xarray_histogram/core.py

@@ -528,7 +528,9 @@ def get_coord(name: str, ax: bh.axis.Axis, dtype: np.dtype, flow: bool) -> xr.Da
     """
     underflow = flow and ax.traits.underflow
     overflow = flow and ax.traits.overflow
-    attrs = dict(bin_type=type(ax).__name__, underflow=underflow, overflow=overflow)
+    attrs = dict(
+        bin_type=type(ax).__name__, underflow=int(underflow), overflow=int(overflow)
+    )
 
     if isinstance(ax, bh.axis.Integer):
         if dtype.kind not in "uib":

tests/test_histogram.py

@@ -87,8 +87,8 @@ def test_regular(self, underflow: bool, overflow: bool):
         assert coord.attrs["bin_type"] == "Regular"
         assert coord.size == ax.extent
         assert coord.attrs["right_edge"] == 1.0
-        assert coord.attrs["underflow"] is underflow
-        assert coord.attrs["overflow"] is overflow
+        assert coord.attrs["underflow"] is int(underflow)
+        assert coord.attrs["overflow"] is int(overflow)
 
         slc = slice(1 if underflow else 0, -1 if overflow else None)
         assert_allclose(coord[slc], ax.edges[:-1])
@@ -110,8 +110,8 @@ def test_regular_int(self, underflow: bool, overflow: bool):
         assert coord.attrs["bin_type"] == "Regular"
         assert coord.size == ax.extent
         assert coord.attrs["right_edge"] == 10.0
-        assert coord.attrs["underflow"] is underflow
-        assert coord.attrs["overflow"] is overflow
+        assert coord.attrs["underflow"] is int(underflow)
+        assert coord.attrs["overflow"] is int(overflow)
 
         slc = slice(1 if underflow else 0, -1 if overflow else None)
         assert_allclose(coord[slc], ax.edges[:-1])
@@ -133,8 +133,8 @@ def test_int(self, underflow: bool, overflow: bool):
         assert coord.attrs["bin_type"] == "Integer"
         assert coord.size == ax.extent
         assert "right_edge" not in coord.attrs
-        assert coord.attrs["underflow"] is underflow
-        assert coord.attrs["overflow"] is overflow
+        assert coord.attrs["underflow"] is int(underflow)
+        assert coord.attrs["overflow"] is int(overflow)
 
         slc = slice(1 if underflow else 0, -1 if overflow else None)
         assert_allclose(coord[slc], ax.edges[:-1])
@@ -153,8 +153,8 @@ def test_intcat(self, overflow: bool):
         assert coord.size == 5 if overflow else 4
         assert np.isdtype(coord.dtype, kind="signed integer")
         assert coord.attrs["bin_type"] == "IntCategory"
-        assert coord.attrs["underflow"] is False
-        assert coord.attrs["overflow"] is overflow
+        assert coord.attrs["underflow"] == 0
+        assert coord.attrs["overflow"] is int(overflow)
         if overflow:
             assert_allclose(coord, [2, 5, 8, 7, np.iinfo(coord.dtype).max])
         else:
@@ -172,8 +172,8 @@ def test_int_bool(self, underflow: bool, overflow: bool):
         assert coord.attrs["bin_type"] == "Integer"
         assert coord.size == ax.extent
         assert "right_edge" not in coord.attrs
-        assert coord.attrs["underflow"] is underflow
-        assert coord.attrs["overflow"] is overflow
+        assert coord.attrs["underflow"] is int(underflow)
+        assert coord.attrs["overflow"] is int(overflow)
         if underflow or overflow:
             assert np.isdtype(coord.dtype, kind="signed integer")
             slc = slice(1 if underflow else 0, -1 if overflow else None)
@@ -197,8 +197,8 @@ def test_intcat_bool(self, overflow: bool):
         coord = hist["var1_bins"]
         assert coord.size == 3 if overflow else 2
         assert coord.attrs["bin_type"] == "IntCategory"
-        assert coord.attrs["underflow"] is False
-        assert coord.attrs["overflow"] is overflow
+        assert coord.attrs["underflow"] == 0
+        assert coord.attrs["overflow"] is int(overflow)
         if overflow:
             assert np.isdtype(coord.dtype, kind="signed integer")
             assert_allclose(coord[:-1], [0, 1])
@@ -515,7 +515,3 @@ def test_partial_dask(self):
         ref = get_ref_hist(x, y, axes=axes)
 
         assert_allclose(hist.to_numpy(), ref, atol=1, rtol=1e-6)
-
-
-class TestFlowBins:
-    pass