Use consistent transform names everywhere

Author: ColCarroll

pymc3/model.py

@@ -23,6 +23,55 @@
 FlatView = collections.namedtuple('FlatView', 'input, replacements, view')
 
 
+def get_transformed_name(name, transform):
+    """
+    Consistent way of transforming names
+
+    Parameters
+    ----------
+    name : str
+        Name to transform
+    transform : object
+        Should be a subclass of `transforms.Transform`
+
+    Returns:
+    A string to use for the transformed variable
+    """
+    return "{}_{}__".format(name, transform.name)
+
+
+def is_transformed_name(name):
+    """
+    Quickly check if a name was transformed with `get_transormed_name`
+
+    Parameters
+    ----------
+    name : str
+        Name to check
+
+    Returns:
+    Boolean, whether the string could have been produced by `get_transormed_name`
+    """
+    return name.endswith('__') and name.count('_') >= 3
+
+
+def get_untransformed_name(name):
+    """
+    Undo transformation in `get_transformed_name`. Throws ValueError if name wasn't transformed
+
+    Parameters
+    ----------
+    name : str
+        Name to untransform
+
+    Returns:
+    String with untransformed version of the name.
+    """
+    if not is_transformed_name(name):
+        raise ValueError(u'{} does not appear to be a transformed name'.format(name))
+    return '_'.join(name.split('_')[:-3])
+
+
 class InstanceMethod(object):
     """Class for hiding references to instance methods so they can be pickled.
 
@@ -516,7 +565,7 @@ def Var(self, name, dist, data=None, total_size=None):
                               ' and added transformed {orig_name} to model.'.format(
                                 transform=dist.transform.name,
                                 name=name,
-                                orig_name='{}_{}_'.format(name, dist.transform.name)))
+                                orig_name=get_transformed_name(name, dist.transform)))
                 self.deterministics.append(var)
                 return var
         elif isinstance(data, dict):
@@ -989,13 +1038,13 @@ def __init__(self, type=None, owner=None, index=None, name=None,
         if type is None:
             type = distribution.type
         super(TransformedRV, self).__init__(type, owner, index, name)
-        
+
         self.transformation = transform
 
         if distribution is not None:
             self.model = model
 
-            transformed_name = "{}_{}_".format(name, transform.name)
+            transformed_name = get_transformed_name(name, transform)
 
             self.transformed = model.Var(
                 transformed_name, transform.apply(distribution), total_size=total_size)

pymc3/sampling.py

@@ -7,7 +7,7 @@
 import pymc3 as pm
 from .backends.base import merge_traces, BaseTrace, MultiTrace
 from .backends.ndarray import NDArray
-from .model import modelcontext, Point
+from .model import modelcontext, Point, is_transformed_name, get_untransformed_name
 from .step_methods import (NUTS, HamiltonianMC, Metropolis, BinaryMetropolis,
                            BinaryGibbsMetropolis, CategoricalGibbsMetropolis,
                            Slice, CompoundStep)
@@ -462,14 +462,13 @@ def _update_start_vals(a, b, model):
     """Update a with b, without overwriting existing keys. Values specified for
     transformed variables on the original scale are also transformed and inserted.
     """
-    
     for name in a:
         for tname in b:
-            if tname.startswith(name) and tname!=name:
-                transform_func = [d.transformation for d in model.deterministics if d.name==name]
+            if is_transformed_name(tname) and get_untransformed_name(tname) == name:
+                transform_func = [d.transformation for d in model.deterministics if d.name == name]
                 if transform_func:
                     b[tname] = transform_func[0].forward(a[name]).eval()
-    
+
     a.update({k: v for k, v in b.items() if k not in a})
 
 def sample_ppc(trace, samples=None, model=None, vars=None, size=None,

pymc3/tests/test_advi.py

@@ -264,4 +264,4 @@ def test_sample_vp(self):
             trace = sample_vp(v_params, draws=1, hide_transformed=True)
             assert trace.varnames == ['p']
             trace = sample_vp(v_params, draws=1, hide_transformed=False)
-            assert sorted(trace.varnames) == ['p', 'p_logodds_']
+            assert sorted(trace.varnames) == ['p', 'p_logodds__']

pymc3/tests/test_diagnostics.py

@@ -19,10 +19,10 @@ def get_ptrace(self, n_samples):
         model = build_disaster_model()
         with model:
             # Run sampler
-            step1 = Slice([model.early_mean_log_, model.late_mean_log_])
+            step1 = Slice([model.early_mean_log__, model.late_mean_log__])
             step2 = Metropolis([model.switchpoint])
             start = {'early_mean': 7., 'late_mean': 5., 'switchpoint': 10}
-            ptrace = sample(n_samples, step=[step1, step2], start=start, njobs=2, 
+            ptrace = sample(n_samples, step=[step1, step2], start=start, njobs=2,
                     progressbar=False, random_seed=[20090425, 19700903])
         return ptrace
 
@@ -91,7 +91,7 @@ def get_switchpoint(self, n_samples):
         model = build_disaster_model()
         with model:
             # Run sampler
-            step1 = Slice([model.early_mean_log_, model.late_mean_log_])
+            step1 = Slice([model.early_mean_log__, model.late_mean_log__])
             step2 = Metropolis([model.switchpoint])
             trace = sample(n_samples, step=[step1, step2], progressbar=False, random_seed=1)
         return trace['switchpoint']

pymc3/tests/test_dist_math.py

@@ -111,5 +111,5 @@ def test_multinomial_bound():
         p_b = pm.Dirichlet('p', floatX(np.ones(2)))
         MultinomialB('x', n, p_b, observed=x)
 
-    assert np.isclose(modelA.logp({'p_stickbreaking_': [0]}),
-                      modelB.logp({'p_stickbreaking_': [0]}))
+    assert np.isclose(modelA.logp({'p_stickbreaking__': [0]}),
+                      modelB.logp({'p_stickbreaking__': [0]}))

pymc3/tests/test_examples.py

@@ -79,14 +79,14 @@ def build_model(self):
     def too_slow(self):
         model = self.build_model()
         start = {'groupmean': self.obs_means.mean(),
-                 'groupsd_interval_': 0,
-                 'sd_interval_': 0,
+                 'groupsd_interval__': 0,
+                 'sd_interval__': 0,
                  'means': self.obs_means,
                  'floor_m': 0.,
                  }
         with model:
             start = pm.find_MAP(start=start,
-                                vars=[model['groupmean'], model['sd_interval_'], model['floor_m']])
+                                vars=[model['groupmean'], model['sd_interval__'], model['floor_m']])
             step = pm.NUTS(model.vars, scaling=start)
             pm.sample(50, step=step, start=start)
 
@@ -117,8 +117,8 @@ def too_slow(self):
         with model:
             start = pm.Point({
                 'groupmean': self.obs_means.mean(),
-                'groupsd_interval_': 0,
-                'sd_interval_': 0,
+                'groupsd_interval__': 0,
+                'sd_interval__': 0,
                 'means': np.array(self.obs_means),
                 'u_m': np.array([.72]),
                 'floor_m': 0.,
@@ -168,7 +168,7 @@ def test_disaster_model(self):
             # Initial values for stochastic nodes
             start = {'early_mean': 2., 'late_mean': 3.}
             # Use slice sampler for means (other varibles auto-selected)
-            step = pm.Slice([model.early_mean_log_, model.late_mean_log_])
+            step = pm.Slice([model.early_mean_log__, model.late_mean_log__])
             tr = pm.sample(500, tune=50, start=start, step=step)
             pm.summary(tr)
 
@@ -178,7 +178,7 @@ def test_disaster_model_missing(self):
             # Initial values for stochastic nodes
             start = {'early_mean': 2., 'late_mean': 3.}
             # Use slice sampler for means (other varibles auto-selected)
-            step = pm.Slice([model.early_mean_log_, model.late_mean_log_])
+            step = pm.Slice([model.early_mean_log__, model.late_mean_log__])
             tr = pm.sample(500, tune=50, start=start, step=step)
             pm.summary(tr)
 
@@ -260,7 +260,7 @@ def test_run(self):
         model = self.build_model()
         with model:
             start = {'psi': 0.5, 'z': (self.y > 0).astype(int), 'theta': 5}
-            step_one = pm.Metropolis([model.theta_interval_, model.psi_logodds_])
+            step_one = pm.Metropolis([model.theta_interval__, model.psi_logodds__])
             step_two = pm.BinaryMetropolis([model.z])
             pm.sample(50, step=[step_one, step_two], start=start)
 

pymc3/tests/test_model.py

@@ -3,6 +3,7 @@
 import numpy as np
 import pymc3 as pm
 from pymc3.distributions import HalfCauchy, Normal
+from pymc3.distributions.transforms import Transform
 from pymc3 import Potential, Deterministic
 from pymc3.theanof import generator
 from .helpers import select_by_precision
@@ -193,3 +194,28 @@ def test_gradient_with_scaling(self):
             g1 = grad1(1)
             g2 = grad2(1)
             np.testing.assert_almost_equal(g1, g2)
+
+
+class TestTransformName(object):
+    cases = [
+        ('var', 'var_test__'),
+        ('var_test_', 'var_test__test__')
+    ]
+    transform_name = 'test'
+
+    def test_get_transformed_name(self):
+        test_transform = Transform()
+        test_transform.name = self.transform_name
+        for name, transformed in self.cases:
+            assert pm.model.get_transformed_name(name, test_transform) == transformed
+
+    def test_is_transformed_name(self):
+        for name, transformed in self.cases:
+            assert pm.model.is_transformed_name(transformed)
+            assert not pm.model.is_transformed_name(name)
+
+    def test_get_untransformed_name(self):
+        for name, transformed in self.cases:
+            assert pm.model.get_untransformed_name(transformed) == name
+            with pytest.raises(ValueError):
+                pm.model.get_untransformed_name(name)

pymc3/tests/test_models_linear.py

@@ -30,7 +30,7 @@ def setup_class(cls):
 
     def test_linear_component(self):
         vars_to_create = {
-            'sigma_interval_',
+            'sigma_interval__',
             'y_obs',
             'lm_x0',
             'lm_Intercept'
@@ -41,7 +41,7 @@ def test_linear_component(self):
                 self.data_linear['y'],
                 name='lm'
             )   # yields lm_x0, lm_Intercept
-            sigma = Uniform('sigma', 0, 20)     # yields sigma_interval_
+            sigma = Uniform('sigma', 0, 20)     # yields sigma_interval__
             Normal('y_obs', mu=lm.y_est, sd=sigma, observed=self.y_linear)  # yields y_obs
             start = find_MAP(vars=[sigma])
             step = Slice(model.vars)
@@ -68,7 +68,7 @@ def test_linear_component_from_formula(self):
     def test_glm(self):
         with Model() as model:
             vars_to_create = {
-                'glm_sd_log_',
+                'glm_sd_log__',
                 'glm_y',
                 'glm_x0',
                 'glm_Intercept'

pymc3/tests/test_plots.py

@@ -57,7 +57,7 @@ def test_multichain_plots():
     model = build_disaster_model()
     with model:
         # Run sampler
-        step1 = Slice([model.early_mean_log_, model.late_mean_log_])
+        step1 = Slice([model.early_mean_log__, model.late_mean_log__])
         step2 = Metropolis([model.switchpoint])
         start = {'early_mean': 2., 'late_mean': 3., 'switchpoint': 50}
         ptrace = sample(1000, step=[step1, step2], start=start, njobs=2)

pymc3/tests/test_sampling.py

@@ -119,12 +119,12 @@ def test_soft_update_empty(self):
         test_point = {'a': 3, 'b': 4}
         pm.sampling._soft_update(start, test_point)
         assert start == test_point
-        
+
     def test_soft_update_transformed(self):
         start = {'a': 2}
-        test_point = {'a_log_': 0}
+        test_point = {'a_log__': 0}
         pm.sampling._soft_update(start, test_point)
-        assert assert_almost_equal(start['a_log_'], np.log(start['a']))
+        assert assert_almost_equal(start['a_log__'], np.log(start['a']))
 
 
 class TestNamedSampling(SeededTest):

pymc3/tests/test_stats.py

@@ -378,5 +378,5 @@ def test_row_names(self):
             step = Metropolis()
             trace = pm.sample(100, step=step)
         ds = df_summary(trace, batches=3, include_transformed=True)
-        npt.assert_equal(np.array(['x_interval_', 'x']),
+        npt.assert_equal(np.array(['x_interval__', 'x']),
                          ds.index)

pymc3/tests/test_variational_inference.py

@@ -116,7 +116,7 @@ def test_sample(self):
                 assert trace.varnames == ['p']
                 assert len(trace) == 1
                 trace = app.sample(draws=10, hide_transformed=False)
-                assert sorted(trace.varnames) == ['p', 'p_logodds_']
+                assert sorted(trace.varnames) == ['p', 'p_logodds__']
                 assert len(trace) == 10
 
         def test_sample_node(self):