Moved weave

1 files changed, 89 insertions, 0 deletions

diff --git a/scipy/weave/examples/md5_speed.py b/scipy/weave/examples/md5_speed.py
new file mode 100644
index 000000000..58edfbc81
--- /dev/null
+++ b/scipy/weave/examples/md5_speed.py
@@ -0,0 +1,89 @@
+"""
+Storing actual strings instead of their md5 value appears to 
+be about 10 times faster.
+
+>>> md5_speed.run(200,50000)
+md5 build(len,sec): 50000 0.870999932289
+md5 retrv(len,sec): 50000 0.680999994278
+std build(len,sec): 50000 0.259999990463
+std retrv(len,sec): 50000 0.0599999427795
+
+This test actually takes several minutes to generate the random
+keys used to populate the dictionaries.  Here is a smaller run,
+but with longer keys.
+
+>>> md5_speed.run(1000,4000)
+md5 build(len,sec,per): 4000 0.129999995232 3.24999988079e-005
+md5 retrv(len,sec,per): 4000 0.129999995232 3.24999988079e-005
+std build(len,sec,per): 4000 0.0500000715256 1.25000178814e-005
+std retrv(len,sec,per): 4000 0.00999999046326 2.49999761581e-006
+
+Results are similar, though not statistically to good because of
+the short times used and the available clock resolution.
+
+Still, I think it is safe to say that, for speed, it is better 
+to store entire strings instead of using md5 versions of 
+their strings.  Yeah, the expected result, but it never hurts
+to check...
+
+"""
+import random, md5, time, cStringIO
+
+def speed(n,m):
+    s = 'a'*n
+    t1 = time.time()            
+    for i in range(m):
+        q= md5.new(s).digest()
+    t2 = time.time()
+    print (t2 - t1) / m
+
+#speed(50,1e6)
+
+def generate_random(avg_length,count):
+    all_str = []
+    alphabet = 'abcdefghijklmnopqrstuvwxyz'
+    lo,hi = [30,avg_length*2+30]
+    for i in range(count):
+        new_str = cStringIO.StringIO()
+        l = random.randrange(lo,hi)
+        for i in range(l):
+            new_str.write(random.choice(alphabet))
+        all_str.append(new_str.getvalue())
+    return all_str    
+    
+def md5_dict(lst):
+    catalog = {}
+    t1 = time.time()
+    for s in lst:
+        key= md5.new(s).digest()
+        catalog[key] = None
+    t2 = time.time()    
+    print 'md5 build(len,sec,per):', len(lst), t2 - t1, (t2-t1)/len(lst)
+    
+    t1 = time.time()
+    for s in lst:
+        key= md5.new(s).digest()
+        val = catalog[key]
+    t2 = time.time()    
+    print 'md5 retrv(len,sec,per):', len(lst), t2 - t1, (t2-t1)/len(lst)
+
+def std_dict(lst):
+    catalog = {}
+    t1 = time.time()
+    for s in lst:
+        catalog[s] = None
+    t2 = time.time()    
+    print 'std build(len,sec,per):', len(lst), t2 - t1, (t2-t1)/len(lst)
+    
+    t1 = time.time()
+    for s in lst:
+        val = catalog[s]
+    t2 = time.time()    
+    print 'std retrv(len,sec,per):', len(lst), t2 - t1, (t2-t1)/len(lst)
+
+def run(m=200,n=10):
+    lst = generate_random(m,n)
+    md5_dict(lst)
+    std_dict(lst)
+
+run(2000,100)    
\ No newline at end of file