use STL map for unicodeindices

Author: andreasvc

src/includes.pxi

@@ -1,11 +1,9 @@
-
 cimport cpython.unicode
-from cython.operator cimport preincrement as inc, dereference as deref
 from libcpp.map cimport map
 from libcpp.string cimport string as cpp_string
+from cython.operator cimport postincrement, dereference
 from cpython.buffer cimport Py_buffer, PyBUF_SIMPLE, PyObject_CheckBuffer, \
         PyObject_GetBuffer, PyBuffer_Release
-from cpython cimport array
 from cpython.version cimport PY_MAJOR_VERSION
 
 

src/match.pxi

@@ -18,7 +18,6 @@ cdef class Match:
 
     property lastgroup:
         def __get__(self):
-
             if self._lastindex < 1:
                 return None
             for name, n in self.re.groupindex.items():
@@ -105,7 +104,8 @@ cdef class Match:
     def groupdict(self):
         result = self._groupdict()
         if self.encoded:
-            return {a: None if b is None else b.decode('utf8') for a, b in result.items()}
+            return {a: None if b is None else b.decode('utf8')
+                    for a, b in result.items()}
         return result
 
     def expand(self, object template):
@@ -264,12 +264,13 @@ cdef class Match:
 
     cdef list _convert_spans(self, spans,
             char * cstring, int size, int * cpos, int * upos):
-        positions = [x for x, _ in spans] + [y for _, y in spans]
-        positions = array.array(b'l' if PY2 else 'l', sorted(set(positions)))
-        posdict = dict(zip(
-                positions,
-                unicodeindices(positions, cstring, size, cpos, upos)))
-        return [(posdict[x], posdict[y]) for x, y in spans]
+        cdef map[int, int] positions
+        cdef int x, y
+        for x, y in spans:
+            positions[x] = x
+            positions[y] = y
+        unicodeindices(positions, cstring, size, cpos, upos)
+        return [(positions[x], positions[y]) for x, y in spans]
 
     def __dealloc__(self):
         delete_StringPiece_array(self.matches)

src/re2.pyx

@@ -106,7 +106,6 @@ include "includes.pxi"
 
 import re
 import sys
-import array
 import warnings
 
 
@@ -404,23 +403,22 @@ cdef utf8indices(char * cstring, int size, int *pos, int *endpos):
     endpos[0] = newendpos
 
 
-cdef array.array unicodeindices(array.array positions,
+cdef void unicodeindices(map[int, int] &positions,
         char * cstring, int size, int * cpos, int * upos):
-    """Convert an array of UTF-8 byte indices to unicode indices."""
+    """Convert UTF-8 byte indices to unicode indices."""
     cdef unsigned char * s = <unsigned char *>cstring
-    cdef int i = 0
-    cdef array.array result = array.clone(positions, len(positions), False)
-
-    if positions.data.as_longs[i] == -1:
-        result.data.as_longs[i] = -1
-        i += 1
-        if i == len(positions):
-            return result
-    if positions.data.as_longs[i] == cpos[0]:
-        result.data.as_longs[i] = upos[0]
-        i += 1
-        if i == len(positions):
-            return result
+    cdef map[int, int].iterator it = positions.begin()
+
+    if dereference(it).first == -1:
+        dereference(it).second = -1
+        postincrement(it)
+        if it == positions.end():
+            return
+    if dereference(it).first == cpos[0]:
+        dereference(it).second = upos[0]
+        postincrement(it)
+        if it == positions.end():
+            return
 
     while cpos[0] < size:
         if s[cpos[0]] < 0x80:
@@ -442,12 +440,11 @@ cdef array.array unicodeindices(array.array positions,
             upos[0] += 1
             emit_endif()
 
-        if positions.data.as_longs[i] == cpos[0]:
-            result.data.as_longs[i] = upos[0]
-            i += 1
-            if i == len(positions):
+        if dereference(it).first == cpos[0]:
+            dereference(it).second = upos[0]
+            postincrement(it)
+            if it == positions.end():
                 break
-    return result
 
 
 __all__ = [