[Predator] Move ``SingleFeatureScore`` to LLM.

appengine/findit/crash/loglinear/feature.py

Index: appengine/findit/crash/loglinear/feature.py
diff --git a/appengine/findit/crash/loglinear/feature.py b/appengine/findit/crash/loglinear/feature.py
index 20cda371f4b9ab7c92d40f972f748ebddb8a1f76..097e67ae7355c428172f47ebe6e1245bc8fdf65d 100644
--- a/appengine/findit/crash/loglinear/feature.py
+++ b/appengine/findit/crash/loglinear/feature.py
@@ -138,3 +138,27 @@ class Feature(object):
     information and decide the cuplrit based on all the evidence together.
     """
     raise NotImplementedError()
+
+
+class FeatureFunction(object):
+  """Given a dict of scalar-valued functions, return an dict-valued function.
+
+  Properties:
+    fs (iterable of functions): A collection of curried functions
+    ``X -> Y -> dict(FeatureValue)``. That is, given a particular ``x`` they
+    return a function ``Y -> dict(FeatureValue)``. N.B. each function should
+    have a name property.

[wrengr, 2017/01/12 19:09:09]

The input functions don't return a dict(FV), they just return one FV. With that
change, this entire description is exactly the description of what the
``Feature`` class is capturing. For the current interface, the collection of
inputs is either:

(a) a dict of X->Y->FV functions. The key gives the name, and we will
automatically convert the raw function into a Feature under the covers.

(b) an iterable of Feature objects. Here we use an iterable so we can't get the
name form there; thus we require a Feature object since that specifies both the
X->Y->FV and the name.

Pick one or the other and run with it.

[Sharu Jiang, 2017/01/13 01:08:34]

Oops, I will stick to the option (b).

+  """
+  def __init__(self, fs):
+    self._fs = fs
+
+  def __call__(self, x):
+    """Fuction mapping ``X -> Y -> dict(A.name to A).

[wrengr, 2017/01/12 19:09:09]

The ``A`` should be ``FeatureValue`` here as well

[Sharu Jiang, 2017/01/13 01:08:34]

Done.

+
+    Returns:
+      A function ``X -> Y -> dict(A.name to A)`` where for all ``x``, ``y``, and

[wrengr, 2017/01/12 19:09:09]

ditto

[Sharu Jiang, 2017/01/13 01:08:34]

Done.

+     for a feature f in fs, we have
+      ``FeatureFunction(fs)(x)(y)[f.name] == f(x)(y)``.
+    """
+    name_to_fx = {f.name: f(x) for f in self._fs}
+    return lambda y: {name: fx(y) for name, fx in name_to_fx.iteritems()}