[#187887894]: Median measure

src/cr/cube/cube.py

@@ -427,6 +427,15 @@ def means(self) -> Optional[np.ndarray]:
             return None
         return self._measures.means.raw_cube_array[self._valid_idxs].astype(np.float64)
 
+    @lazyproperty
+    def medians(self) -> Optional[np.ndarray]:
+        """Optional float64 ndarray of the cube_medians if the measure exists."""
+        if self._measures.medians is None:
+            return None
+        return self._measures.medians.raw_cube_array[self._valid_idxs].astype(
+            np.float64
+        )
+
     @lazyproperty
     def missing(self) -> int:
         """Get missing count of a cube."""
@@ -781,6 +790,14 @@ def means(self) -> 'Optional["_MeanMeasure"]':
         mean = _MeanMeasure(self._cube_dict, self._all_dimensions, self._cube_idx_arg)
         return None if mean.raw_cube_array is None else mean
 
+    @lazyproperty
+    def medians(self) -> 'Optional["_MediansMeasure"]':
+        """Optional _MedianMeasure object providing access to means values."""
+        medians = _MediansMeasure(
+            self._cube_dict, self._all_dimensions, self._cube_idx_arg
+        )
+        return None if medians.raw_cube_array is None else medians
+
     @lazyproperty
     def missing_count(self) -> int:
         """numeric representing count of missing rows in cube response."""
@@ -790,6 +807,10 @@ def missing_count(self) -> int:
         # fixtures that don't have valid_counts.
         if self.means is not None:
             return self.means.missing_count
+        # The check on the median measure is needed for retro-compatibility with the old
+        # fixtures that don't have valid_counts.
+        if self.medians is not None:
+            return self.medians.missing_count
         return self._cube_dict["result"].get("missing", 0)
 
     @lazyproperty
@@ -1047,6 +1068,33 @@ def _flat_values(self) -> Optional[np.ndarray]:
         ).flatten()
 
 
+class _MediansMeasure(_BaseMeasure):
+    """Statistical medians values from a cube-response."""
+
+    @lazyproperty
+    def missing_count(self) -> int:
+        """Numeric value representing count of missing rows in response."""
+        return self._cube_dict["result"]["measures"]["median"].get("n_missing", 0)
+
+    @lazyproperty
+    def _flat_values(self) -> Optional[np.ndarray]:
+        """Optional 1D np.ndarray of np.float64 median values as found in cube response.
+
+        Medians data may include missing items represented by a dict like
+        {'?': -1} in the cube response. These are replaced by np.nan in the
+        returned value.
+        """
+        measure_payload = self._cube_dict["result"].get("measures", {}).get("median")
+        if measure_payload is None:
+            return None
+        return np.array(
+            tuple(
+                np.nan if isinstance(x, dict) else x for x in measure_payload["data"]
+            ),
+            dtype=np.float64,
+        ).flatten()
+
+
 class _OverlapMeasure(_BaseMeasure):
     """Overlap values from a cube-response."""
 

src/cr/cube/cubepart.py

@@ -711,6 +711,23 @@ def means(self):
                 "`.means` is undefined for a cube-result without a mean measure"
             )
 
+    @lazyproperty
+    def medians(self):
+        """2D optional np.float64 ndarray of median values for each table cell.
+
+        Cell value is `np.nan` for each cell corresponding to an inserted subtotal
+        (medians of addend cells cannot simply be added to get the mean of the
+        subtotal).
+
+        Raises `ValueError` if the cube-result does not include a median cube-measure.
+        """
+        try:
+            return self._assemble_matrix(self._measures.medians.blocks)
+        except ValueError:
+            raise ValueError(
+                "`.medians` is undefined for a cube-result without a median measure"
+            )
+
     @lazyproperty
     def min_base_size_mask(self):
         return MinBaseSizeMask(self, self._mask_size)
@@ -2005,6 +2022,20 @@ def means(self):
                 "`.means` is undefined for a cube-result without a mean measure"
             )
 
+    @lazyproperty
+    def medians(self):
+        """1D np.float64 ndarray of medians for each row of strand.
+
+        Raises ValueError when accessed on a cube-result that does not contain a median
+        cube-measure.
+        """
+        try:
+            return self._assemble_vector(self._measures.medians.blocks)
+        except ValueError:
+            raise ValueError(
+                "`.medians` is undefined for a cube-result without a median measure"
+            )
+
     @lazyproperty
     def min_base_size_mask(self):
         """1D bool ndarray of True for each row that fails to meet min-base spec.

src/cr/cube/enums.py

@@ -157,6 +157,7 @@ class CUBE_MEASURE(enum.Enum):
     COVARIANCE = "covariance"
     COUNT = "count"
     MEAN = "mean"
+    MEDIAN = "median"
     OVERLAP = "overlap"
     STDDEV = "stddev"
     SUM = "sum"

src/cr/cube/matrix/cubemeasure.py

@@ -25,6 +25,11 @@ def cube_means(self):
         """_BaseCubeMeans subclass object for this cube-result."""
         return _BaseCubeMeans.factory(self._cube, self._dimensions, self._slice_idx)
 
+    @lazyproperty
+    def cube_medians(self):
+        """_BaseCubeMedian subclass object for this cube-result."""
+        return _BaseCubeMedians.factory(self._cube, self._dimensions, self._slice_idx)
+
     @lazyproperty
     def cube_overlaps(self):
         """_BaseCubeOverlaps subclass object for this cube-result."""
@@ -877,6 +882,92 @@ def means(self):
         return self._means[:, 0, :, 0]
 
 
+# === MEDIANs ===
+
+
+class _BaseCubeMedians(_BaseCubeMeasure):
+    """Base class for medians cube-measure variants."""
+
+    def __init__(self, dimensions, medians):
+        super(_BaseCubeMedians, self).__init__(dimensions)
+        self._medians = medians
+
+    @classmethod
+    def factory(cls, cube, dimensions, slice_idx):
+        """Return _BaseCubeMedian subclass instance appropriate to `cube`.
+
+        Raises `ValueError` if the cube-result does not include a cube-median measure.
+        """
+        if cube.medians is None:
+            raise ValueError("cube-result does not contain cube-median measure")
+        dimension_types = cube.dimension_types[-2:]
+        CubeMedianCls = (
+            _MrXMrCubeMedians
+            if dimension_types == (DT.MR, DT.MR)
+            else (
+                _MrXCatCubeMedians
+                if dimension_types[0] == DT.MR
+                else (
+                    _CatXMrCubeMedians
+                    if dimension_types[1] == DT.MR
+                    else _CatXCatCubeMedians
+                )
+            )
+        )
+        return CubeMedianCls(
+            dimensions, cube.medians[cls._slice_idx_expr(cube, slice_idx)]
+        )
+
+    @lazyproperty
+    def medians(self):
+        """2D np.float64 ndarray of cube medians."""
+        raise NotImplementedError(  # pragma: no cover
+            f"`{type(self).__name__}` must implement `.medians`"
+        )
+
+
+class _CatXCatCubeMedians(_BaseCubeMedians):
+    """Medians cube-measure for a slice with no MR dimensions."""
+
+    @lazyproperty
+    def medians(self):
+        """2D np.float64 ndarray of medians for each valid matrix cell."""
+        return self._medians
+
+
+class _CatXMrCubeMedians(_BaseCubeMedians):
+    """Medians cube-measure for a NOT_MR_X_MR slice.
+
+    Note that the rows-dimensions need not actually be CAT.
+    """
+
+    @lazyproperty
+    def medians(self):
+        """2D np.float64 ndarray of medians for each valid matrix cell."""
+        return self._medians[:, :, 0]
+
+
+class _MrXCatCubeMedians(_BaseCubeMedians):
+    """Medians cube-measure for an MR_X_NOT_MR slice.
+
+    Note that the columns-dimension need not actually be CAT.
+    """
+
+    @lazyproperty
+    def medians(self):
+        """2D np.float64 ndarray of medians for each valid matrix cell."""
+        return self._medians[:, 0, :]
+
+
+class _MrXMrCubeMedians(_BaseCubeMedians):
+    """Medians cube-measure for an MR_X_MR slice."""
+
+    @lazyproperty
+    def medians(self):
+        """2D np.float64 ndarray of medians for each valid matrix cell."""
+        return self._medians[:, 0, :, 0]
+
+
 # === OVERLAPS ===
 
 

src/cr/cube/matrix/measure.py

@@ -206,6 +206,11 @@ def means(self):
         """_Means measure object for this cube-result"""
         return _Means(self._dimensions, self, self._cube_measures)
 
+    @lazyproperty
+    def medians(self):
+        """_Medians measure object for this cube-result"""
+        return _Medians(self._dimensions, self, self._cube_measures)
+
     def pairwise_p_vals_for_subvar(self, subvar_idx):
         """_PairwiseSigPValsForSubvar measure object for this cube-result"""
         return _PairwiseSigPValsForSubvar(
@@ -1177,6 +1182,17 @@ def blocks(self):
         )
 
 
+class _Medians(_BaseSecondOrderMeasure):
+    """Provides the medians measure for a matrix."""
+
+    @lazyproperty
+    def blocks(self):
+        """2D array of the four 2D "blocks" making up this measure."""
+        return NanSubtotals.blocks(
+            self._cube_measures.cube_medians.medians, self._dimensions
+        )
+
+
 class _MeansSmoothed(_Means, _SmoothedMeasure):
     """Provides the smoothed mean measure for a matrix."""
 

src/cr/cube/stripe/cubemeasure.py

@@ -26,6 +26,11 @@ def cube_means(self):
         """_BaseCubeMeans subclass object for this stripe."""
         return _BaseCubeMeans.factory(self._cube, self._rows_dimension)
 
+    @lazyproperty
+    def cube_medians(self):
+        """_BaseCubeMedians subclass object for this stripe."""
+        return _BaseCubeMedians.factory(self._cube, self._rows_dimension)
+
     @lazyproperty
     def cube_stddev(self):
         """_BaseCubeStdDev subclass object for this stripe."""
@@ -253,6 +258,57 @@ def means(self):
         return self._means[:, 0]
 
 
+# === MEDIANs ===
+
+
+class _BaseCubeMedians(_BaseCubeMeasure):
+    """Base class for medians cube-measure variants."""
+
+    def __init__(self, rows_dimension, medians):
+        super(_BaseCubeMedians, self).__init__(rows_dimension)
+        self._medians = medians
+
+    @classmethod
+    def factory(cls, cube, rows_dimension):
+        """Return _BaseCubeMedian subclass instance appropriate to `cube`."""
+        if cube.medians is None:
+            raise ValueError("cube-result does not contain cube-median measure")
+        MedianCls = (
+            _MrCubeMedians
+            if rows_dimension.dimension_type == DT.MR
+            else _CatCubeMedians
+        )
+        return MedianCls(rows_dimension, cube.medians)
+
+    @lazyproperty
+    def medians(self):
+        """1D np.float64 ndarray of medians for each stripe row."""
+        raise NotImplementedError(
+            f"`{type(self).__name__}` must implement `.medians`"
+        )  # pragma: no cover
+
+
+class _CatCubeMedians(_BaseCubeMedians):
+    """Medians cube-measure for a non-MR stripe."""
+
+    @lazyproperty
+    def medians(self):
+        """1D np.float64 ndarray of medians for each stripe row."""
+        return self._medians
+
+
+class _MrCubeMedians(_BaseCubeMedians):
+    """Medians cube-measure for an MR stripe.
+
+    Its `.medians` is a 2D ndarray with axes (rows, sel/not).
+    """
+
+    @lazyproperty
+    def medians(self):
+        """1D np.float64 ndarray of medians for each stripe row."""
+        return self._medians[:, 0]
+
+
 # === STD DEV ===
 
 

src/cr/cube/stripe/measure.py

@@ -38,6 +38,11 @@ def means(self):
         """_Means measure object for this stripe."""
         return _Means(self._rows_dimension, self, self._cube_measures)
 
+    @lazyproperty
+    def medians(self):
+        """_Medians measure object for this stripe."""
+        return _Medians(self._rows_dimension, self, self._cube_measures)
+
     @lazyproperty
     def population_proportions(self):
         """_PopulationPrortion measure object for this stripe."""
@@ -231,6 +236,27 @@ def subtotal_values(self):
         return NanSubtotals.subtotal_values(self.base_values, self._rows_dimension)
 
 
+class _Medians(_BaseSecondOrderMeasure):
+    """Provides the medians measure for a stripe.
+
+    Relies on the presence of a medians cube-measure in the cube-result.
+    """
+
+    @lazyproperty
+    def base_values(self):
+        """1D np.float64 ndarray of medians for each row."""
+        return self._cube_measures.cube_medians.medians
+
+    @lazyproperty
+    def subtotal_values(self):
+        """1D ndarray of np.nan for each row-subtotal.
+
+        Medians values cannot be subtotaled and each subtotal value is unconditionally
+        np.nan.
+        """
+        return NanSubtotals.subtotal_values(self.base_values, self._rows_dimension)
+
+
 class _MeansSmoothed(_Means, _SmoothedMeasure):
     """Provides the smoothed means measure for a stripe.