1.2.4
 - added check for unicode problems

1.2.3
 - added compatibility to newer versions of mxdatetime

1.2.2
 - Option added to sort(), enabling case insensitive sorting

1.2.1
 - bugfix release

1.2.0
 - added mysql superclass wrapper with auto-reconnect

1.1.0
 - overloading the time conversion method in MySQLdb with mxdatetime properly

1.0.0
 - bugfix: updating virtual columns always resulted in an exception
 - added refresh() method to result sets. So you can easily stay up
   to date with result sets that you already have.

0.9.9
 - bugfix: another bug with multi-target xrefs, this time in recursive deletion
 - added replicate() method. It allows you to replicate the contents of a
   database result object into a new table or into a different data set in the
   same table while conveniently adding or changing data at the same time.

0.9.8
 - bugfix: recursive deletion crashed with multi-target xrefs
 - bugfix: MySQLdb API change worked around

0.9.7:
 - added index() method for result sets

0.9.6:
 - bugfix in cross referencing code
 - optimization for len() calls

0.9.5:
 - Added "sort" option to select() method, so it is possible to issue
   ORDER BY calls without leaving python ground.
 - some compatibility fixes

0.9.4:
 - Added various new methods (e.g. __iter__()) to enhance the user experience
   when using pSQL with python 2.3 and above
 - The insert() method does now return the index of the freshly inserted 
   column. (At least if that table happens to have an index...)
 - License changed to GPL V2
 - Added functionality to reduce the number of columns to be queried in a
   table, thanks to a suggestion by Bernhard Reiter. You can use the method
   table.use_columns(<column_list>) to define a set of column names or
   table.ignore_column(<column_name>) to just remove a single column from
   the list of used columns. Please note that NO actual data (columns) will
   be removed from your database - they will just be ignored by that table
   object.
 - Added an update() method to the result object. This way you can update
   an entire row with just one call. 
   Use it like this: paul = drinkers.select({"Name": "Paul", "Age":20})[0]
                     paul.update({"Age": 21, "Beers": 999})

0.9.3:
 - fixed another problem with inserting binary data.
 - compiling more accurate SQL queries now. Please note that pSQL now uses
   LIKE syntax automatically if there is a '%' in your expression. If you
   want to match a '%' exactly you will have to use table.sql_select().
 - compiling more failsafe SQL queries now.
 - the speed of most queries should have significantly improved

0.9.2:
 - fixed pSQL to continue functioning with mySQLdb 0.9.2 and up which
   had its internal character escaping routine moved.
 - added pop() method for result sets
 - added new option to x_ref: multi
   When you set an x_ref with multi=2 you will get back a result array
   for your reference and no exception if the reference hits zero or
   more than one data set.
   With multi=0 you will get "None" instead of an exception for a
   dangling cross reference.
   multi=1 is the default. It behaves like always.
 - Added quoting for table names. This bug was here for more than 2
   years and nobody sent a report! Maybe you all just use simple tables.

0.9.1:
 - internal data conversion now several times faster.
 - no more problems writing binary data!
    -> f = open("core")
       mydatabase.mytable.File = f.read() # thats all to get your file in!
 - sub("regsub.gsub", "re.sub") and a number of other small fixes.
   Really 'cool things' will have to wait for summer to happen. Sorry.
   This version has not been tested with 1.5.2 any longer but it might
   still work.

0.9.0:
 - DateTime import statement changed to mx.DateTime
 - preliminary support for python versions 2.0 and above. This is far from
   being complete! Expect cool things this spring!
   This version has not been tested with 1.5.2 any longer but it should
   still work.

0.8.9:
 - bugfix: wrong identifier used in result object setitem call.

   0.8.9 IS THE LAST pSQL version guaranteed to work with python 1.5.2

0.8.8:

 - FIXED PYTHON 1.5.2 BUILTIN str() to leave out the dreaded "L" for long() !!
   The patch is applied automatically when the module is being loaded.
   Through its nature, this fix is global to the scope of the interpreter and
   could break other 1.5.2 software. On the other hand, the fixed str() now
   works like the str() of all other python versions which should be ok.
   If you dont want it, comment out the marked line at the very beginning of
   pSQL.py
 - fixed a bug that could lead to false results for concatenated queries
 - fixed problems with updating of strings that contained "'" or '"'.
 - fixed a bug that occured when iterating through concatenated result objects
 - introduced virtual columns for result objects, suitable to attach random
   data to a result object (e.g. related data sets from othe tables to be used
   for further computation). Virtual columns are attached on a per-object basis
   only, as opposed to cross references which are table-wide.
   The methods to be used are setVirtual(name, data) and delVirtual(name).
   Retrieval an manipulation of virtual columns is the same as with standard
   columns in the result object. They are not shown by repr(). Take them as
   a kind of 'sticky notes' for result objects.
 - added reverse() method for result set objects. To sort some result by some
   number in descending order just use res.sort("colname"); res.reverse()
 - added has_key() method to result objects. (why did I forget that?)
 - keys() and has_keys() now refer to all data items of a result object,
   including cross references and virtual columns
 - column() generates result set objects for columns that contain result objects
   (e.g. cross references or virtuals)
 - sort() can sort cross references and virtuals as well, now. You can even
   sort by columns of the cross referenced data! Cool, eh?
 - because the keys() methods report additional keys, the values() method of the
   result object now returns the respective values.
 - to query just the column names without cross references use the new methods
   in the table and result object: columns() and has_column(name)
 - to query just the values of the database row represented by a result object
   without additional cross refs or virtuals, use the row() method.

0.8.7:

 - fixed updating of float() values which could raise unwanted exceptions
 - added comparison methods for result set-, table- and database objects
 - added append() method to the result set object (aka 'res1.append(res2)')
 - added a valid __setitem__ operation to result set objects
 - added a valid __add__ operation to result set objects (aka 'res1 + res2')
 - added result object slicing (aka 'resobj[2:4]')