[cdec.sa] Explicit feature names in grammar extractor output

+ sparse features in extractor
+ hg.intersect(string)
+ basestring = str|unicode

1 files changed, 34 insertions, 0 deletions

diff --git a/python/src/sa/features.pxi b/python/src/sa/features.pxi
new file mode 100644
index 00000000..fcb93f26
--- /dev/null
+++ b/python/src/sa/features.pxi
@@ -0,0 +1,34 @@
+cdef StringMap FD = StringMap()
+
+INITIAL_CAPACITY = 7 # default number of features
+INCREMENT = INITIAL_CAPACITY # double size
+
+cdef class FeatureVector:
+    def __cinit__(self):
+        self.names = IntList(INITIAL_CAPACITY, INCREMENT)
+        self.values = FloatList(INITIAL_CAPACITY, INCREMENT)
+
+    def set(self, unsigned name, float value):
+        self.names.append(name)
+        self.values.append(value)
+
+    def __iter__(self):
+        cdef unsigned i
+        for i in range(self.names.len):
+            yield (FD.word(self.names[i]), self.values[i])
+
+    def __str__(self):
+        return ' '.join('%s=%s' % feat for feat in self)
+
+cdef class Scorer:
+    cdef models
+    def __init__(self, *models):
+        names = [FD.index(<char *>model.__name__) for model in models]
+        self.models = zip(names, models)
+
+    cdef FeatureVector score(self, Phrase fphrase, Phrase ephrase,
+            unsigned paircount, unsigned fcount, unsigned fsample_count):
+        cdef FeatureVector scores = FeatureVector()
+        for name, model in self.models:
+            scores.set(name, model(fphrase, ephrase, paircount, fcount, fsample_count))
+        return scores