Added performance testing for atomic operations

README.txt

@@ -1,3 +1,4 @@
 This is the Python 3.4 runtime for AntLR.
-Visit the AntLR web site for more information:
-http://www.antlr.org
+Visit the AntLR web sites for more information:
+http://www.antlr.org
+http://theantlrguy.atlassian.net/wiki/display/ANTLR4/Python+Target

RELEASE-4.5.txt

@@ -0,0 +1,12 @@
+What's in this release?
+
+ - fixed bug where non-ascii input streams would fail
+ - added support for visitor pattern
+ - added support for wildcards in grammar
+
+ Breaking change:
+
+ In version 4.4, the parser/lexer had a tokenNames member.
+ This has been removed in favor of the following members:
+  - lexicalNames, containing the parsed text
+  - symbolicNames, corresponding to tokenNames

setup.py

@@ -8,6 +8,6 @@ setup(
     url='http://www.antlr.org',
     license='BSD',
     author='Eric Vergnaud, Terence Parr, Sam Harwell',
-    author_email='eric@test',
+    author_email='eric.vergnaud@wanadoo.fr',
     description='ANTLR 4.5 runtime for Python 3.4.0'
 )

src/test/TestAtomicPerformance.py

@@ -0,0 +1,87 @@
+__author__ = 'ericvergnaud'
+
+import unittest
+import time
+import uuid
+import io
+
+class TestAtomicPerformance(unittest.TestCase):
+
+    # test to check that on the current platform, a < b < c is faster than b in range(a,c)
+    # x in xrange is intensively used in Interval
+    def test_in_range(self):
+        xstart = time.time()
+        # check with 20 various range sizes
+        for max in range(0,1000,50):
+            r = range(0,max)
+            for i in r:
+                ok = i in r
+        xend = time.time()
+        print(str((xend-xstart)*1000000))
+        ystart = time.time()
+        # check with 20 various range sizes
+        for max in range(0,1000,50):
+            r = range(0,max)
+            for i in r:
+                ok = max > i >= 0
+        yend = time.time()
+        print(str((yend-ystart)*1000000))
+        self.assertTrue((yend-ystart)<(xend-xstart))
+
+    # test to check that on the current platform, hashing string tuples is faster than hashing strings
+    def test_tuple_hash(self):
+        # create an array of random strings
+        s = []
+        for i in range(0,10000):
+            s.append(str(uuid.uuid4()))
+        # hash then using string concat
+        xstart = time.time()
+        for i in range(0,9999):
+            a = hash(s[i] + s[i+1])
+        for i in range(0,9998):
+            a = hash(s[i] + s[i+1] + s[i+2])
+        for i in range(0,9997):
+            a = hash(s[i] + s[i+1] + s[i+2] + s[i+3])
+        xend = time.time()
+        print(str((xend-xstart)*1000000))
+        ystart = time.time()
+        # hash then using string tuple
+        for i in range(0,9999):
+            a = hash((s[i],s[i+1]))
+        for i in range(0,9998):
+            a = hash((s[i],s[i+1],s[i+2]))
+        for i in range(0,9997):
+            a = hash((s[i],s[i+1],s[i+2],s[i+3]))
+        yend = time.time()
+        print(str((yend-ystart)*1000000))
+        self.assertTrue((yend-ystart)<(xend-xstart))
+        zstart = time.time()
+        # hash then using string tuple
+        for i in range(0,9999):
+            b = io.StringIO()
+            b.write(s[i])
+            b.write(s[i+1])
+            a = hash(b.getvalue())
+            b.close()
+        for i in range(0,9998):
+            b = io.StringIO()
+            b.write(s[i])
+            b.write(s[i+1])
+            b.write(s[i+2])
+            a = hash(b.getvalue())
+            b.close()
+        for i in range(0,9997):
+            b = io.StringIO()
+            b.write(s[i])
+            b.write(s[i+1])
+            b.write(s[i+2])
+            b.write(s[i+3])
+            a = hash(b.getvalue())
+            b.close()
+        zend = time.time()
+        print(str((zend-zstart)*1000000))
+        self.assertTrue((yend-ystart)<(zend-zstart))
+
+
+if __name__ == '__main__':
+    unittest.main()