Fix bug whereby extractors that validated their direct dependencies b…

…roke Feature.version

featureflow/__init__.py

@@ -1,4 +1,4 @@
-__version__ = '1.8.9'
+__version__ = '1.8.10'
 
 from model import BaseModel
 

featureflow/feature.py

@@ -4,6 +4,7 @@
 from decoder import JSONDecoder, Decoder, GreedyDecoder, DecoderNode, \
     BZ2Decoder, PickleDecoder
 from datawriter import DataWriter, StringIODataWriter
+from persistence import PersistenceSettings
 
 
 class Feature(object):
@@ -49,7 +50,13 @@ def __str__(self):
 
     @property
     def version(self):
-        return self.extractor(**self.extractor_args).version
+        # KLUDGE: Build a shallow version of the extractor.  Building a deep
+        # version with re-usable code is more difficult, because
+        # self._build_extractor relies on this version property, so there's
+        # a circular dependency.
+        dependencies = [f.extractor(**f.extractor_args) for f in self.needs]
+        e = self.extractor(needs=dependencies, **self.extractor_args)
+        return e.version
 
     def copy(
             self,
@@ -205,7 +212,7 @@ def _depends_on(self, _id, graph, persistence):
             needs.append(e)
         return needs
 
-    def _build_extractor(self, _id, graph, persistence):
+    def _build_extractor(self, _id, graph, persistence, never_store=False):
         try:
             return graph[self.key]
         except KeyError:
@@ -218,7 +225,7 @@ def _build_extractor(self, _id, graph, persistence):
             setattr(e, '_reader', reader)
 
         graph[self.key] = e
-        if not self.store:
+        if never_store or not self.store:
             return e
 
         key = self.key

featureflow/test_integration.py

@@ -70,6 +70,16 @@ def _process(self, data):
         yield self._version
 
 
+class ValidatesDependencies(Node):
+    def __init__(self, needs=None):
+        if not needs:
+            raise ValueError('you must supply at least one dependency')
+        super(ValidatesDependencies, self).__init__(needs=needs)
+
+    def _process(self, data):
+        yield data
+
+
 class Echo(Node):
     def __init__(self, needs=None):
         super(Echo, self).__init__(needs=needs)
@@ -320,6 +330,21 @@ class MultipleRoots(BaseModel):
 
 class BaseTest(object):
 
+    def test_can_use_node_that_validates_its_dependency_list(self):
+        class D1(BaseModel, self.Settings):
+            stream = Feature(TextStream, store=True)
+            words = Feature(Tokenizer, needs=stream, store=False)
+            count = JSONFeature(WordCount, needs=words, store=True)
+            timestamp = JSONFeature(
+                    TimestampEmitter,
+                    version='1',
+                    needs=stream,
+                    store=True)
+            validated = Feature(ValidatesDependencies, needs=stream, store=True)
+
+        _id = D1.process(stream='mary')
+        self.assertTrue(_id)
+
     def test_recomputes_when_necessary(self):
         class D1(BaseModel, self.Settings):
             stream = Feature(TextStream, store=True)