fixed up tests

Author: jsalvatier

pymc/distributions/__init__.py

@@ -1,6 +1,7 @@
 from continuous import *
 from discrete import * 
 from transforms import * 
+from distribution import *
 
 import multivariate
 

pymc/distributions/continuous.py

@@ -344,10 +344,10 @@ def logp(value):
 
     return locals()
 
-def Tpos(nu, mu=0, lam=1):
+def Tpos(*args, **kwargs):
     """
     Student-t distribution bounded at 0
     see T
     """
-    return Bound(T(nu, mu, lam), 0)
+    return Bound(T(*args,**kwargs), 0)
 

pymc/distributions/distribution.py

@@ -2,6 +2,7 @@
 import numpy as np 
 from ..quickclass import *
 from ..model import *
+__all__ = ['DensityDist', 'TensorDist', 'tensordist', 'continuous', 'discrete', 'arbitrary']
 
 class Distribution(object):
     def __new__(cls, *args, **kwargs):

pymc/model.py

@@ -87,7 +87,7 @@ def d2logpc(model, vars = None):
 
     @property
     def test_point(self):
-        return Point( (var, var.tag.test_value) for var in self.vars)
+        return Point(self, ((var, var.tag.test_value) for var in self.vars))
 
     @property
     def cont_vars(model):
@@ -175,7 +175,7 @@ def cont_inputs(f):
 
 def gradient1(f, v):
     """flat gradient of f wrt v"""
-    return t.flatten(t.grad(f, v))
+    return t.flatten(t.grad(f, v, disconnected_inputs='warn'))
 
 def gradient(f, vars = None):
     if not vars: 

pymc/sample.py

@@ -38,7 +38,7 @@ def sample(model, draws, step, start = None, trace = None, progress_bar = True):
     draws = int(draws)
     if start is None: 
         start = trace[-1]
-    point = Point(start)
+    point = Point(model, start)
 
     if not hasattr(trace, 'record'):
         if trace is None: 

tests/test_distributions.py

@@ -68,23 +68,41 @@ def test_poisson():
 def test_constantdist():
     checkd(ConstantDist, I, {'c' : I})
 
+def test_zeroinflatedpoisson():
+    checkd(ZeroInflatedPoisson, I, {'theta' : Rplus, 'z' : Bool})
+    
+
+def test_densitydist():
+    def logp(x):
+        return -log(2*.5) - abs(x-.5)/.5
+    
+    checkd(DensityDist,R, {}, extra_args = {'logp' : logp})
+
+
+def test_addpotential():
+    with Model() as model: 
+        x = Normal('x', 1,1)
+        model.AddPotential(-x**2)
+
+        check_dlogp(model, x, [R])
 
-def checkd(distfam, valuedomain, vardomains, check_int = True, check_der = True):
 
-    m = Model()
+def checkd(distfam, valuedomain, vardomains, check_int = True, check_der = True, extra_args = {}):
 
-    with m: 
+    with Model() as m: 
         vars = dict((v , Flat(v, dtype = dom.dtype)) for v,dom in vardomains.iteritems())
+        vars.update(extra_args)
+        print vars 
         value = distfam('value', testval = valuedomain[len(valuedomain)//2], **vars)
 
-    vardomains['value'] = np.array(valuedomain)
+        vardomains['value'] = np.array(valuedomain)
 
-    domains = [np.array(vardomains[str(v)]) for v in m.vars]
+        domains = [np.array(vardomains[str(v)]) for v in m.vars]
 
-    if check_int:
-        check_int_to_1(m, value, domains)
-    if check_der: 
-        check_dlogp(m, value, domains)
+        if check_int:
+            check_int_to_1(m, value, domains)
+        if check_der: 
+            check_dlogp(m, value, domains)
 
 
 def check_int_to_1(model, value, domains):