Add a cython test to give perspective on relative ease/performance

mcfletch · mcfletch · commit 3c0c8782c945 · 2017-11-18T02:43:23.000-05:00
The cython implementation is about 2x slower than the Rust
implementation, but it's not clear if the slice in the cython
call is actually doing a copy of the entire memory-area or just
doing a simple reference of the memory.

Note that the cython implementation doesn't build a proper module
with setup.py, it just uses the command-line cythonize operation
to produce a quick-and-dirty extension.
diff --git a/.gitignore b/.gitignore
@@ -5,8 +5,12 @@
 /pyext-myclib/myclib.py
 /pyext-myclib/myclib_wrap.c
 /pyext-myclib/_myclib.cpython*
+/pyext-mycythonlib/*.so
+/pyext-mycythonlib/*.c
+/pyext-mycythonlib/*.html
 .cache
 .benchmarks
 __pycache__
 myrustlib.so
+mycythonlib*.so
 /pyext-myrustlib/.gitignore
diff --git a/Makefile b/Makefile
@@ -1,11 +1,12 @@
-.PHONY: clean clean-test clean-pyc clean-build docs help
+.PHONY: clean clean-test clean-pyc clean-build docs help test-python test-rust test-c test-cython compile-cython compile-rust compile-c
 
 clean: clean-build clean-pyc clean-test ## remove all build, test, coverage and Python artifacts
 
 clean-build: ## remove build artifacts
 	rm -fr build/
 	rm -fr dist/
 	rm -fr .eggs/
+	rm -fr .compiled
 	find . -name '*.egg-info' -exec rm -fr {} +
 	find . -name '*.egg' -exec rm -f {} +
 
@@ -32,6 +33,9 @@ test-rust: ## run tests quickly with the default Python
 test-c: ## run tests quickly with the default Python
 	py.test -v -s doubles_with_c_swig.py
 
+test-cython: # run tests quickly with the default Python
+	py.test -v -s doubles_with_cython.py
+
 test-all: ## run tests quickly with the default Python
 	py.test -v -s doubles_all.py
 
@@ -41,3 +45,14 @@ compile-rust: ## compile new rust lib
 
 compile-c: ## compile new c lib
 	@cd pyext-myclib;python3 setup.py build_ext -i
+
+compile-cython: ## compile new cython lib
+	@cd pyext-mycythonlib;cythonize -a -i mycythonlib.pyx
+	@cp pyext-mycythonlib/mycythonlib*.so ./
+
+compile-all: compile-rust compile-c compile-cython
+
+.compiled: compile-all
+	touch .compiled
+
+test: compile-all test-all
diff --git a/doubles_all.py b/doubles_all.py
@@ -4,6 +4,7 @@
 import itertools
 import numpy as np
 import myrustlib  # <-- Importing Rust Implemented Library
+import mycythonlib # <-- Importing Cython Implemented Library
 
 import sys
 sys.path.append('./pyext-myclib')
@@ -101,3 +102,6 @@ def test_c_swig_bytes_once(benchmark):
 
 # def test_rust_regex(benchmark):
 #     print(benchmark(myrustlib.count_doubles_regex, val))
+
+def test_cython(benchmark):
+    print(benchmark(mycythonlib.count_doubles, val))
diff --git a/pyext-mycythonlib/mycythonlib.pyx b/pyext-mycythonlib/mycythonlib.pyx
@@ -0,0 +1,18 @@
+def count_doubles( unicode source ):
+    """Count number of doubles in a (unicode) string
+    
+    A double is counted for every character where the
+    character at the previous index in the string is
+    the same character as the current character. Thus
+    the string 'aaa' has two doubles.
+    """
+    cdef Py_ssize_t count
+    count = 0
+    if not source:
+        return count
+    char = source[0]
+    for next_char in source[1:]:
+        if next_char == char:
+            count += 1
+        char = next_char
+    return count
diff --git a/requirements.txt b/requirements.txt
@@ -1,3 +1,4 @@
 pytest
 pytest-benchmark
 numpy
+cython

-Original file line number
+Diff line change
@@ @@ -1,3 +1,4 @@ @@
 pytest
 pytest-benchmark
 numpy
 +cython