cgi.py inefficiency (with fix)

Thomas Bellman bellman at lysator.liu.se
Wed Nov 17 13:48:01 CET 1999


I was profiling an embedded web server we have written in Python,
and found that it spent a lot of time in FieldStorage.has_key()
and FieldStorage.__getitem__() from cgi.py.  So I looked at the
code for the FieldStorage class, and found that it basically
emulates a normal dictionary, but is implemented using a linear
list of fields.  Thus, has_key() and __getitem__() must do linear
searches to find things.  Is there any reason why this is done,
instead of having a dictionary of lists?

I did a quick stab at replacing the list with a dictionary, and
in our application, we got a threefold increase in speed per
request...  Of course, in our application we have forms sometimes
containing well over a thousand entries.  It is probably not very
common for people to have such large forms, but it is a pity that
the cgi module does not scale well.

Anyway, here is a diff of the changes I made.  I haven't tested
it very extensively (in our application we have subclassed the
FieldStorage class, and use a modified __init__() method, which
makes it difficult to fully test the real FieldStorage class),
but from what I have tested, it seems to work.

I started with revision 1.45 of cgi.py, which I checked out from
the CVS archives this Monday.


===================================================================
RCS file: /projects/cvsroot/python/dist/src/Lib/cgi.py,v
retrieving revision 1.45
diff -u -5 -r1.45 cgi.py
--- cgi.py	1999/06/11 18:26:09	1.45
+++ cgi.py	1999/11/17 12:30:23
@@ -724,11 +724,11 @@
 
     """Like FieldStorage, for use when no file uploads are possible."""
 
     # Dummy attributes
     filename = None
-    list = None
+    dict = None
     type = None
     file = None
     type_options = {}
     disposition = None
     disposition_options = {}
@@ -890,11 +890,12 @@
                 pass
             if maxlen and clen > maxlen:
                 raise ValueError, 'Maximum content length exceeded'
         self.length = clen
 
-        self.list = self.file = None
+        self.file = None
+	self.dict = None
         self.done = 0
         self.lines = []
         if ctype == 'application/x-www-form-urlencoded':
             self.read_urlencoded()
         elif ctype[:10] == 'multipart/':
@@ -912,74 +913,69 @@
             raise AttributeError, name
         if self.file:
             self.file.seek(0)
             value = self.file.read()
             self.file.seek(0)
-        elif self.list is not None:
-            value = self.list
+        elif self.dict is not None:
+            value = self.dict.values()
         else:
             value = None
         return value
 
     def __getitem__(self, key):
         """Dictionary style indexing."""
-        if self.list is None:
+        if self.dict is None:
             raise TypeError, "not indexable"
-        found = []
-        for item in self.list:
-            if item.name == key: found.append(item)
-        if not found:
-            raise KeyError, key
+        found = self.dict[key]
         if len(found) == 1:
             return found[0]
         else:
             return found
 
     def keys(self):
         """Dictionary style keys() method."""
-        if self.list is None:
+        if self.dict is None:
             raise TypeError, "not indexable"
-        keys = []
-        for item in self.list:
-            if item.name not in keys: keys.append(item.name)
-        return keys
+        return self.dict.keys()
 
     def has_key(self, key):
         """Dictionary style has_key() method."""
-        if self.list is None:
+        if self.dict is None:
             raise TypeError, "not indexable"
-        for item in self.list:
-            if item.name == key: return 1
-        return 0
+        return self.dict.has_key(key)
 
     def __len__(self):
         """Dictionary style len(x) support."""
-        return len(self.keys())
+        return len(self.dict)
 
     def read_urlencoded(self):
         """Internal: read data in query string format."""
         qs = self.fp.read(self.length)
-        self.list = list = []
+        self.dict = dict = {}
         for key, value in parse_qsl(qs, self.keep_blank_values,
                                     self.strict_parsing):
-            list.append(MiniFieldStorage(key, value))
+	    l = dict.get(key, [])
+	    l.append(MiniFieldStorage(key, value))
+            dict[key] = l
         self.skip_lines()
 
     FieldStorageClass = None
 
     def read_multi(self, environ, keep_blank_values, strict_parsing):
         """Internal: read a part that is itself multipart."""
-        self.list = []
+        self.dict = {}
         klass = self.FieldStorageClass or self.__class__
         part = klass(self.fp, {}, self.innerboundary,
                      environ, keep_blank_values, strict_parsing)
         # Throw first part away
         while not part.done:
             headers = rfc822.Message(self.fp)
             part = klass(self.fp, headers, self.innerboundary,
                          environ, keep_blank_values, strict_parsing)
-            self.list.append(part)
+	    l = self.dict.get(part.name, [])
+	    l.append(part)
+            self.dict[part.name] = l
         self.skip_lines()
 
     def read_single(self):
         """Internal: read an atomic part."""
         if self.length >= 0:
===================================================================

-- 
Thomas Bellman,   Lysator Computer Club,   Linköping University,  Sweden
"Beware of bugs in the above code; I have    !  bellman @ lysator.liu.se
 only proved it correct, not tried it."      !  Make Love -- Nicht Wahr!




More information about the Python-list mailing list