Merge 5e9c85c into d57b41a

.appveyor.yml

@@ -92,10 +92,6 @@ install:
   # https://blog.ionelmc.ro/2014/12/21/compiling-python-extensions-on-windows/ for 64bit C compilation
   - ps: copy .\ci\appveyor\vcvars64.bat "C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin\amd64"
   - "ECHO APPVEYOR_PULL_REQUEST_NUMBER is: %APPVEYOR_PULL_REQUEST_NUMBER%"
-  - "%CMD_IN_ENV% python .\\ci\\make_conda_packages.py"
-
-  # test that we can conda install zipline in a new env
-  - conda create -n installenv --yes -q --use-local python=%PYTHON_VERSION% numpy=%NUMPY_VERSION% zipline -c quantopian -c https://conda.anaconda.org/quantopian/label/ci
 
   - ps: $env:BCOLZ_VERSION=(sls "bcolz==([^ ]*)" .\etc\requirements_locked.txt -ca).matches.groups[1].value
   - ps: $env:NUMEXPR_VERSION=(sls "numexpr==([^ ]*)" .\etc\requirements_locked.txt -ca).matches.groups[1].value
@@ -106,7 +102,7 @@ install:
   # Since conda installs latest certifi by default, we would fail to uninstall that new version when trying to install the pinned version using pip later in the build:
   # "Cannot uninstall 'certifi'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall."
   - ps: $env:CERTIFI_VERSION=(sls "certifi==([^ ]*)" .\etc\requirements_locked.txt -ca).matches.groups[1].value
-  - conda create -n testenv --yes -q --use-local "pip<19" python=%PYTHON_VERSION% numpy=%NUMPY_VERSION% pandas=%PANDAS_VERSION% scipy=%SCIPY_VERSION% ta-lib=%TALIB_VERSION% bcolz=%BCOLZ_VERSION% numexpr=%NUMEXPR_VERSION% pytables=%PYTABLES_VERSION% h5py=%H5PY_VERSION% certifi=%CERTIFI_VERSION% -c quantopian -c https://conda.anaconda.org/quantopian/label/ci
+  - conda create -n testenv --yes -q --use-local "pip<19" python=%PYTHON_VERSION% numpy=%NUMPY_VERSION% pandas=%PANDAS_VERSION% scipy=%SCIPY_VERSION% ta-lib=%TALIB_VERSION% bcolz=%BCOLZ_VERSION% numexpr=%NUMEXPR_VERSION% pytables=%PYTABLES_VERSION% h5py=%H5PY_VERSION% certifi=%CERTIFI_VERSION% -c quantopian -c https://conda.anaconda.org/quantopian/label/ci -c conda-forge
   - activate testenv
   - bash etc/dev-install --cache-dir=%LOCALAPPDATA%\pip\Cache\pip_np%CONDA_NPY%py%CONDA_PY%
   - python -m pip freeze | sort

etc/dev-install

@@ -19,7 +19,7 @@ echo
 
 # New releases of pip have frequently caused strange issues. Make sure
 # we know exactly which version we're working with.
-python -m pip install pip==19.2.2 $@
+python -m pip install pip==19.2.2 'setuptools<46' $@
 
 # Install external requirements first: if they share any of our
 # transitive dependencies, we want our pinned versions to win.

tests/pipeline/test_classifier.py

@@ -609,8 +609,9 @@ class C(Classifier):
 
     @parameter_space(
         dtype_and_missing=[(int64_dtype, -1), (categorical_dtype, None)],
+        use_mask=[True, False],
     )
-    def test_peer_count(self, dtype_and_missing):
+    def test_peer_count(self, dtype_and_missing, use_mask):
         class C(Classifier):
             dtype = dtype_and_missing[0]
             missing_value = dtype_and_missing[1]
@@ -636,12 +637,29 @@ class C(Classifier):
                 missing_value=None,
             )
 
-        expected = np.array(
-            [[3, 3, np.nan, 1, 3, np.nan],
-             [4, 1, 1, 4, 4, 4],
-             [np.nan, 1, 3, 3, 3, np.nan],
-             [6, 6, 6, 6, 6, 6]],
-        )
+        if not use_mask:
+            mask = self.build_mask(self.ones_mask(shape=data.shape))
+            expected = np.array(
+                [[3, 3, np.nan, 1, 3, np.nan],
+                 [4, 1, 1, 4, 4, 4],
+                 [np.nan, 1, 3, 3, 3, np.nan],
+                 [6, 6, 6, 6, 6, 6]],
+            )
+        else:
+            # Punch a couple holes in the mask to check that we handle the mask
+            # correctly.
+            mask = self.build_mask(
+                np.array([[1, 1, 1, 1, 0, 1],
+                          [1, 1, 1, 1, 1, 0],
+                          [1, 1, 1, 1, 1, 1],
+                          [1, 1, 0, 0, 1, 1]], dtype='bool')
+            )
+            expected = np.array(
+                [[2, 2, np.nan, 1, np.nan, np.nan],
+                 [3, 1, 1, 3, 3, np.nan],
+                 [np.nan, 1, 3, 3, 3, np.nan],
+                 [4, 4, np.nan, np.nan, 4, 4]],
+            )
 
         terms = {
             'peer_counts': c.peer_count(),
@@ -654,7 +672,7 @@ class C(Classifier):
             terms=terms,
             expected=expected_results,
             initial_workspace={c: data},
-            mask=self.build_mask(self.ones_mask(shape=data.shape)),
+            mask=mask,
         )
 
 

zipline/pipeline/classifiers/classifier.py

@@ -398,11 +398,17 @@ def _to_integral(self, output_array):
             )
         return group_labels, null_label
 
-    def peer_count(self):
+    def peer_count(self, mask=NotSpecified):
         """
         Construct a factor that gives the number of occurrences of
         each distinct category in a classifier.
 
+        Parameters
+        ----------
+        mask : zipline.pipeline.Filter, optional
+            If passed, only count assets passing the filter.  Default behavior
+            is to count all assets.
+
         Examples
         --------
         Let ``c`` be a Classifier which would produce the following output::
@@ -433,7 +439,7 @@ def peer_count(self):
         """
         # Lazy import due to cyclic dependencies in factor.py, classifier.py
         from ..factors import PeerCount
-        return PeerCount(inputs=[self])
+        return PeerCount(inputs=[self], mask=mask)
 
 
 class Everything(Classifier):