]> git.ipfire.org Git - thirdparty/sqlalchemy/sqlalchemy.git/commitdiff
changeset about connection pool rel_0_4beta4
authorMike Bayer <mike_mp@zzzcomputing.com>
Wed, 22 Aug 2007 19:36:54 +0000 (19:36 +0000)
committerMike Bayer <mike_mp@zzzcomputing.com>
Wed, 22 Aug 2007 19:36:54 +0000 (19:36 +0000)
CHANGES

diff --git a/CHANGES b/CHANGES
index 1ced053e7dcfa46b358cab8b60a5afe344f3d059..cba42858953a0b861cc220d7d7b7fed0fa2a40a7 100644 (file)
--- a/CHANGES
+++ b/CHANGES
@@ -19,6 +19,15 @@ CHANGES
   - the 'Smallinteger' compatiblity name (small i!) is no longer imported, but
     remains in schema.py for now.  SmallInteger (big I!) is still imported.
 
+- the connection pool uses a "threadlocal" strategy internally to return
+  the same connection already bound to a thread, for "contextual" connections;
+  these are the connections used when you do a "connectionless" execution
+  like insert().execute().  This is like a "partial" version of the
+  "threadlocal" engine strategy but without the thread-local transaction part 
+  of it.  We're hoping it reduces connection pool overhead as well as 
+  database usage.  However, if it proves to impact stability in a negative way,
+  we'll roll it right back.
+  
 - Fix to bind param processing such that "False" values (like blank strings)
   still get processed/encoded.