Merge pull request #916 from keflavich/issue915

fix for 915: account for different use cases of median

spectral_cube/dask_spectral_cube.py

@@ -287,15 +287,6 @@ def __exit__(self, *args):
     def _compute(self, array):
         return array.compute(**self._scheduler_kwargs)
 
-    def _warn_slow(self, funcname):
-        if self._is_huge and not self.allow_huge_operations:
-            raise ValueError("This function ({0}) requires loading the entire "
-                             "cube into memory, and the cube is large ({1} "
-                             "pixels), so by default we disable this operation. "
-                             "To enable the operation, set "
-                             "`cube.allow_huge_operations=True` and try again."
-                             .format(funcname, self.size))
-
     def _get_filled_data(self, view=(), fill=np.nan, check_endian=None, use_memmap=None):
 
         if check_endian:

spectral_cube/spectral_cube.py

@@ -1191,6 +1191,17 @@ def median(self, axis=None, iterate_rays=False, **kwargs):
         except ImportError:
             bnok = False
 
+        how = kwargs.pop('how', None)
+        if how == 'slice' and (not isinstance(axis, (list, tuple)) or len(axis) != 2):
+            raise ValueError("Cannot compute median slicewise unless you're compressing over two axes.")
+        elif how == 'ray':
+            if axis not in (0, 1, 2):
+                raise ValueError("Cannot compute median raywise unless you're compressing over one axis.")
+            else:
+                if not iterate_rays:
+                    iterate_rays = True
+                    warnings.warn("how='ray' was specified in call to median; this is setting iterate_rays=True")
+
         # slicewise median is nonsense, must force how = 'cube'
         # bottleneck.nanmedian does not allow axis to be a list or tuple
         if bnok and not iterate_rays and not isinstance(axis, (list, tuple)):
@@ -1203,7 +1214,12 @@ def median(self, axis=None, iterate_rays=False, **kwargs):
             log.debug("Using numpy nanmedian")
             result = self.apply_numpy_function(np.nanmedian, axis=axis,
                                                projection=True, unit=self.unit,
-                                               how='cube',**kwargs)
+                                               how='cube', **kwargs)
+        elif iterate_rays:
+            result = self.apply_numpy_function(
+                nanmedian if bnok else np.nanmedian if hasattr(np, 'nanmedian') else np.median,
+                axis=axis, projection=True, unit=self.unit, how='ray',
+                check_endian=True, **kwargs)
         else:
             log.debug("Using numpy median iterating over rays")
             result = self.apply_function(np.median, projection=True, axis=axis,

spectral_cube/utils.py

@@ -1,4 +1,5 @@
 import warnings
+import inspect
 
 from functools import wraps
 
@@ -33,19 +34,26 @@ def warn_slow(function):
     @wraps(function)
     def wrapper(self, *args, **kwargs):
         # if the function accepts a 'how', the 'cube' approach requires the whole cube in memory
-        warn_how = (kwargs.get('how') == 'cube') or 'how' not in kwargs
-        if self._is_huge and not self.allow_huge_operations and warn_how:
-            raise ValueError("This function ({0}) requires loading the entire "
-                             "cube into memory, and the cube is large ({1} "
-                             "pixels), so by default we disable this operation. "
-                             "To enable the operation, set "
-                             "`cube.allow_huge_operations=True` and try again.  "
-                             "Alternatively, you may want to consider using an "
-                             "approach that does not load the whole cube into "
-                             "memory by specifying how='slice' or how='ray'.  "
-                             "See {bigdataurl} for details."
-                             .format(str(function), self.size,
-                                     bigdataurl=bigdataurl))
+        argspec = inspect.getfullargspec(function)
+        accepts_how_keyword = 'how' in argspec.args or argspec.varkw == 'how'
+
+        warn_how = accepts_how_keyword and ((kwargs.get('how') == 'cube') or 'how' not in kwargs)
+
+        if self._is_huge and not self.allow_huge_operations:
+            warn_message = ("This function ({0}) requires loading the entire "
+                            "cube into memory, and the cube is large ({1} "
+                            "pixels), so by default we disable this operation. "
+                            "To enable the operation, set "
+                            "`cube.allow_huge_operations=True` and try again.  ").format(str(function), self.size)
+
+            if warn_how:
+                warn_message += ("Alternatively, you may want to consider using an "
+                                 "approach that does not load the whole cube into "
+                                 "memory by specifying how='slice' or how='ray'.  ")
+
+            warn_message += ("See {bigdataurl} for details.".format(bigdataurl=bigdataurl))
+
+            raise ValueError(warn_message)
         elif warn_how and not self._is_huge:
             # TODO: add check for whether cube has been loaded into memory
             warnings.warn("This function ({0}) requires loading the entire cube into "