Make dependeny return optional => predictable.

Readme.md

@@ -28,7 +28,7 @@ result = x(a=(1, 20), b=(2,  30), c=2)
 # used if present. This may be useful to calculate statistical and systemic
 # errors in one go.
 print(result)
-# >> array([ 1.41421356, 35.35533906])
+# >> (1.41421356, 35.35533906)
 
 # As a goodie, you can print out the gaussian error distribution in
 # symbolic form. (Works best in Jupyter Notebooks)
@@ -82,6 +82,10 @@ x(a=([1,2,3], [1,2,3]), b=([1,2,3], 1))
 x(a=([1,2,3], [1,2,3]), b=([1,2], 1))
 ```
 
+If all the returned arrays in the tuple have the same shape, you can
+easily convert that tuple to a numpy array:
+`np.array(x(a=([1,2,3], [1,2,3]), b=([1,2,3], 1)))`
+
 ### Dependencies
 To make the calculation of complex values easier, one can define
 dependencies for a `SecondaryValue`:
@@ -90,14 +94,22 @@ dependencies for a `SecondaryValue`:
 from SecondaryValue import SecondaryValue
 
 dep = SecondaryValue('u')
-x = SecondaryValue("a + b", dependencies={'a': dep})
+x = SecondaryValue("a + b", dependencies=dict(a=dep))
 
 # x will now accept u as an additional kwarg and calculate d==dep on the fly
-# and return a dictionary containing it as a second return value.
+# and return a dictionary containing it as a second return value if you
+# specify `retdeps=True`.
+print(x(b=1, u=(1, 2), retdeps=True))
+# >> ((2.0, 2.0), {'a': ((1.0, 2.0), {})})
+
+# To make the output predictable, the dependencies aren't returned by deafult.
 print(x(b=1, u=(1, 2)))
-# >> (array([2., 2.]), {'a': array([1., 2.])})
+# >> (2.0, 2.0)
 
 # you can overwrite the dependency calculation
 print(x(b=1/2, a=1/2))
 # >> 1.0
 ```
+
+If there are no dependency values, an empty dict will be returned when
+`retdeps=True` is specified.

SecondaryValue/SecondaryValue.py

@@ -66,9 +66,9 @@ class SecondaryValue:
             if name in kwargs:
                 continue
 
-            tmp = sec_val(**kwargs)
+            # we always calculate the depndencies
-            kwargs[name] = tmp[0] if (len(tmp) > 1 and isinstance(tmp, tuple)) \
+            tmp = sec_val(retdeps=True, **kwargs)
-                else tmp
+            kwargs[name] = tmp[0]
             calc_deps[name] = tmp
 
         return kwargs, calc_deps
@@ -111,6 +111,7 @@ class SecondaryValue:
                   dep_values (the values and errors of the dependencies)
 
         :rtype: Tuple
+
         """
 
         kwargs, dep_values = self._calc_deps(**kwargs)
@@ -125,7 +126,6 @@ class SecondaryValue:
         kwargs = {var: val for var, val in kwargs.items() \
                   if var in self._symbols}
 
-
         max_uncertainties = max([len(val) for _, val in kwargs.items() \
                                  if isinstance(val, Iterable)] or [0])
 
@@ -158,7 +158,6 @@ class SecondaryValue:
             central_value = np.empty(value_length)
             for i in range(0, value_length):
                 current_values = join_row(scalar_values, vector_values, i)
-
                 central_value[i] = self._parsed_lambda(**current_values)
         else:
             central_value = self._dtype(self._parsed_lambda(**scalar_values))
@@ -204,22 +203,24 @@ class SecondaryValue:
 
         return terms
 
-    def __call__(self, *args, **kwargs):
+    def __call__(self, *args, retdeps=False, **kwargs):
         """Calculates a value from the expression by substituting
-        variables by the values of the given keyword arguments.  If an
-        argument is specified as a tuplpe of (value, error) the
-        gausssian error propagation will be computed.
 
-        The values and errors can be iterable, but must compatible shapes.
+        The values and errors can be iterable, but must compatible
+        shapes.
 
-        :returns: value or [value, error] or [value, error], dependencies
+        :param retdeps: wether to return a dictionary with the
+            calculated dependencies
+
+        :returns: value or [value, error] or [value, error, ...],
+                  dependencies
 
         :rtype: numpy data type or np array of [value, errors, ...] or
                 a tuple the beforementioned as first element and a
-                dictionary with the calculated dependencies as a second value
+                dictionary with the calculated dependencies as a
+                second value
         """
-
+         # process the keyword arguments
-        # process the keyword arguments
         values, errors, dep_values = self._process_args(*args, **kwargs)
 
         # calulate the central value
@@ -237,7 +238,7 @@ class SecondaryValue:
         result.insert(0, central_value)
         result = tuple(result)
 
-        if dep_values:
+        if retdeps:
             return result, dep_values
 
         return result
@@ -293,5 +294,5 @@ def filter_out_vecotrized(dictionary):
     return scalar, vector
 
 def join_row(scalar, vector, index):
-     return {**scalar, **{key: val[index] \
+    return {**scalar, **{key: val[index] \
-                          for key, val in vector.items()}}
+                         for key, val in vector.items()}}

setup.py

@@ -6,7 +6,7 @@ def readme():
         return f.read()
 
 setup(name='SecondaryValue',
-      version='0.1.3',
+      version='0.1.4',
       description='A helper to calculate the gaussian error propagation.',
       long_description=readme(),
       long_description_content_type='text/markdown',