From 495e1c7366d30f79ef3332d05c94dbf14abd909d Mon Sep 17 00:00:00 2001
From: Matthew Honnibal <honnibal@gmail.com>
Date: Tue, 9 Dec 2014 16:50:01 +1100
Subject: [PATCH] * Use fused type in Tokens.push_back, simplifying the use of
 the cache

---
 spacy/lang.pxd   | 11 +++++++++++
 spacy/lang.pyx   | 39 +++++++++++++++++++++------------------
 spacy/tokens.pxd | 10 +++++++++-
 spacy/tokens.pyx | 12 ++++++------
 4 files changed, 47 insertions(+), 25 deletions(-)

diff --git a/spacy/lang.pxd b/spacy/lang.pxd
index 0307e12fe..8a6aa5f97 100644
--- a/spacy/lang.pxd
+++ b/spacy/lang.pxd
@@ -13,6 +13,17 @@ from .tagger cimport univ_tag_t
 from .utf8string cimport StringStore, UniStr
 
 
+cdef union LexemesOrTokens:
+    const Lexeme* const* lexemes
+    TokenC* tokens
+
+
+cdef struct Cached:
+    LexemesOrTokens data
+    bint is_lex
+    int length
+
+
 cdef class Lexicon:
     cpdef public get_lex_props
     cdef Pool mem
diff --git a/spacy/lang.pyx b/spacy/lang.pyx
index cdae8644a..044bfb7bc 100644
--- a/spacy/lang.pyx
+++ b/spacy/lang.pyx
@@ -137,21 +137,19 @@ cdef class Language:
 
     cdef int _try_cache(self, int idx, hash_t key, Tokens tokens) except -1:
         cdef int i
-        specials = <TokenC*>self._specials.get(key)
-        if specials != NULL:
-            i = 0
-            while specials[i].lex != NULL:
-                tokens.push_back(idx, specials[i].lex)
-                tokens.data[tokens.length - 1].pos = specials[i].pos
-                tokens.data[tokens.length - 1].morph = specials[i].morph
-                tokens.data[tokens.length - 1].lemma = specials[i].lemma
-                tokens.data[tokens.length - 1].sense = specials[i].sense
-                i += 1
+        cdef TokenC* token
+        cached = <Cached*>self._specials.get(key)
+        if cached != NULL:
+            assert not cached.is_lex
+            for i in range(cached.length):
+                token = &cached.data.tokens[i]
+                idx = tokens.push_back(idx, token)
             return True
         else:
-            cached = <const Lexeme* const*>self._cache.get(key)
+            cached = <Cached*>self._cache.get(key)
             if cached != NULL:
-                tokens.extend(i, cached, 0)
+                assert cached.is_lex == True
+                tokens.extend(i, cached.data.lexemes, cached.length)
                 return True
             else:
                 return False
@@ -244,11 +242,14 @@ cdef class Language:
         for i in range(n):
             if tokens[i].lex.id == 1:
                 return 0
-        lexemes = <const Lexeme**>self.mem.alloc(n + 1, sizeof(Lexeme**))
+        cached = <Cached*>self.mem.alloc(1, sizeof(Cached))
+        cached.length = n
+        cached.is_lex = True
+        lexemes = <const Lexeme**>self.mem.alloc(n, sizeof(Lexeme**))
         for i in range(n):
             lexemes[i] = tokens[i].lex
-        lexemes[i + 1] = NULL
-        self._cache.set(key, lexemes)
+        cached.data.lexemes = <const Lexeme* const*>lexemes
+        self._cache.set(key, cached)
 
     cdef int _find_infix(self, Py_UNICODE* chars, size_t length) except -1:
         cdef unicode string = chars[:length]
@@ -287,10 +288,12 @@ cdef class Language:
                 if lemma:
                     tokens[i].lemma = self.lexicon.strings[lemma]
                 set_morph_from_dict(&tokens[i].morph, props)
-            # Null-terminated array
-            tokens[i+1].lex = NULL
+            cached = <Cached*>self.mem.alloc(1, sizeof(Cached))
+            cached.length = len(substrings)
+            cached.is_lex = False
+            cached.data.tokens = tokens
             slice_unicode(&string, chunk, 0, len(chunk))
-            self._specials.set(string.key, tokens)
+            self._specials.set(string.key, cached)
 
 
 cdef int set_morph_from_dict(Morphology* morph, dict props) except -1:
diff --git a/spacy/tokens.pxd b/spacy/tokens.pxd
index f3d6011ec..01bec6815 100644
--- a/spacy/tokens.pxd
+++ b/spacy/tokens.pxd
@@ -30,6 +30,14 @@ cdef struct TokenC:
     int sense
 
 
+ctypedef const Lexeme* const_Lexeme_ptr
+ctypedef TokenC* TokenC_ptr
+
+ctypedef fused LexemeOrToken:
+    const_Lexeme_ptr
+    TokenC_ptr
+
+
 cdef class Tokens:
     cdef Pool mem
     cdef StringStore _string_store
@@ -40,7 +48,7 @@ cdef class Tokens:
     cdef int max_length
 
     cdef int extend(self, int i, const Lexeme* const* lexemes, int n) except -1
-    cdef int push_back(self, int i, const Lexeme* lexeme) except -1
+    cdef int push_back(self, int i, LexemeOrToken lex_or_tok) except -1
     cpdef int set_tag(self, int i, int tag_type, int tag) except -1
 
     cpdef np.ndarray[long, ndim=2] get_array(self, list features)
diff --git a/spacy/tokens.pyx b/spacy/tokens.pyx
index 004d0578c..4075e64d7 100644
--- a/spacy/tokens.pyx
+++ b/spacy/tokens.pyx
@@ -60,16 +60,16 @@ cdef class Tokens:
     def __len__(self):
         return self.length
 
-    cdef int push_back(self, int idx, const Lexeme* lexeme) except -1:
+    cdef int push_back(self, int idx, LexemeOrToken lex_or_tok) except -1:
         if self.length == self.max_length:
             self._realloc(self.length * 2)
         cdef TokenC* t = &self.data[self.length]
-        t.lex = lexeme
-        t.idx = idx
-        t.pos = 0
-        t.sense = 0
+        if LexemeOrToken is TokenC_ptr:
+            t[0] = lex_or_tok[0]
+        else:
+            t.lex = lex_or_tok
         self.length += 1
-        return idx + lexeme.length
+        return idx + t.lex.length
 
     cdef int extend(self, int idx, const Lexeme* const* lexemes, int n) except -1:
         cdef int i