CHANGELOG

2.01
=======================

*issues*
--------------

FIXD  - runQuery and runInteraction needed ConnectionPool._error, which was
        removed.  Fixed.

2.0
=======================

*issues*
--------------

FIXD  - Errback handling fixes
        
FIXD  - Removed useless explicit conversion of integers to
        NUMBER when no special formatting is done on them

FIXD  - Exposed __int__, __float__ and friends in NUMBER
        just in case a NUMBER instance is tossed to format()

FIXD  - The above two fixes mean you can now do %03d style
        formatting in format()

FIXD  - A trailing semicolon can be included on execute()
        commands now

FIXD  - Added boolean support (thanks: Andrea Arcangeli)
        
*enhancements*
--------------

 - Experimental Unicode Support (thanks Matt Goodall)

 - Unix socket support added (thanks Stephen Early)

 - convertBinary function to get stored binary back out
   of the database



1.15b
=======================

*issues*
--------------

FIXD  - cleaned up a few circular references and added a gc.collect()
        to the pool's prune() routine to remove memory leaking

FIXD  - Made it so the second attempt to errback in runOperation is 
        ignored on exception in cases that the errback has already been
		called


1.14b
=======================

*issues*
--------------

FIXD  - fixed typo in runOperation, improved errback handling

FIXD  - fixed a few concurrency issues, dropped connection handling


1.13b
=======================

*enhancements*
--------------

 - Pool size control implemented


1.12b
=======================

*issues*
--------------

FIXD  - added connect() method to ConnectionPool


1.11b
=======================

*enhancements*
--------------

 - Transaction and ConnectionPool (adbapi compat) added


*issues*
--------------

FIXD  - protocol layer no longer calls back with stale transaction 
        status flag


1.10b
=======================

*issues*
--------------

FIXD  - format() and execute() now take all pyformat argument
        types, not just dicts

FIXD  - closed connections handle failed query execution more
        gracefully


1.9b
=======================

*enhancements*
--------------

 - connection.cursor() is no longer deferred; you get your cursor object
   immediately.  Now, everything is queued, including before a real
   connection is made to the db, and *no* callbacks are necessary unless 
   you care about a query response.

 - pgasync knows the difference between "dirty" and "clean" transactions.
   It will avoid issuing ROLLBACK when only selects have been done, and
   similarly it will resume a select-only transaction across different 
   cursors to avoid a needless BEGIN

 - connection.exFetch was implemented for convenience


1.8b
=======================

*issues*
--------------

FIXD  - str, int, float, True/False, datetime instances returned from 
        fetch() queries instead of pgtypes

FIXD  - date/time/timestamp handling corrected

FIXD  - None passed to execute/exFetch results in NULL insertion

FIXD  - Premature cursor release issues resolved; removed automatic 
        cursor release 

FIXD  - Simplified pgtypes for output-only usage; removed
        stack inspection


1.7b
=======================

*issues*
--------------

FIXD  - fetchone() was corrected to return values consistent
        with the PEP


1.6b
=======================

*issues*
--------------

FIXD  - STRING and BINARY types now inherit from str, not
        types.StringType


1.5b
=======================

*issues*
--------------

FIXD  - execute() now strips query when looking for SELECT

FIXD  - params are optional to execute and exFetch

FIXD  - docstrings reflect the new param style


1.4b
=======================

*issues*
--------------

FIXD  - Changed references to imaginary 'pystyle' format
        to 'pyformat'

FIXD  - Changed execute/exFetch functions to accept a
        dictionary instead of kwargs in compliance with
		the DB API convention


1.3b
=======================

*enhancements*
--------------
 - execute (and variants) can take a datetime.datetime now
   as a format argument

*additions*
-----------
 - examples/wiki

*issues*
--------------

FIXD  - error handling is now corrected.  addErrback/errback works 
        for all library deferred exceptions
       
FIXD  - under high concurrency, release() is necessary to give the
        cursor back to the pool.. so release() was implemented


1.2b
=======================

*issues*
--------------

FIXD  - Old examples distributed instead of new


1.1b
=======================
Note: connection.cursor() returns a Deferred now

*enhancements*
--------------
 - Query queues implemented.  Slight performance increase, and much cleaner 
   API.  

*removals*
----------
 - adbapi compat layer.  

*issues*
--------------

FIXD  - Problem with setup.py *always* requiring Pyrex (credit: Bob Ippolito)



1.0b
=======================

*enhancements*
--------------
 - Change to twisted's deferreds
 - adbapi wrapper

*issues*
--------------

FIXD  - string quote escape sequence was "\'", now it's "''"



3b
=======================

*enhancements*
--------------

 - connect is now an instantaneous operation
 - exFetch added to the fe.py.  



2b
=======================

*enhancements*
--------------

 - Simplified the net.py code b/c we're *only using byte count terminators;
   moved it from my arraybuffer class to (c)StringIO.  Faster now.
 - Renamed cursor.conn to cursor.connection in compliance with optional
   DB API 2.0 spec

*issues*
-------

FIXD  - Hanging in C fixed when the move to cStringIO buffering happened
        strings when things speed up, concurrency issues
FIXD  - __del__ / gc wasn't happening because cursor._callbackInfo held a
        reference
FIXD  - cursor.rowcount reflects modifications on INSERT, UPDATE
FIXD  - cursor 'none' error 
FIXD  - Ensure that the convert.c things work correcty on Linux, other systems
FIXD  - Source documentation
FIXD  - Callbacks need a queue, not just a var