ANNOUNCE: KirbyBase 1.7

[Jamey Cribbs]

KirbyBase is a simple, plain-text, database management system written in 
Python.  It can be used either embedded in a python script or in a 
client/server, multi-user mode.  You use python code to express your 
queries instead of having to use another language such as SQL.  
KirbyBase is disk-based, not memory-based.  Database changes are 
immediately written to disk.

You can find more information on KirbyBase at:
http://www.netpromi.com/kirbybase.html

You can download KirbyBase for Python at:
http://www.netpromi.com/files/KirbyBase_Python_1.7.zip

Wow!  It's been almost two years since the initial release of 
KirbyBase.  Time sure does fly!  Version 1.7 includes most of the bug 
fixes that have accumulated over the months and a few enhancements that 
I hope you will enjoy.

I would like to thank everyone who has emailed me with comments, bug 
reports, and enhancement requests/ideas.  Hearing from people who 
actually use KirbyBase is what makes working on it worthwhile.  Please 
keep the emails coming!

I would particularly like to thank Pierre Quentel, the author of 
Karrigell (http://karrigell.sourceforge.net), for his contribution of 
ideas and code for many of the enhancements in version 1.7 of KirbyBase.

For those of you who requested better documentation, the manual has been 
completely re-written.  I'm not saying it's any better, but at least 
it's different.  :)

Changes in Version 1.7:

       ***IMPORTANT - IF YOU ARE UPGRADING THIS COULD BITE YOU!!!***
*  Changed the default value for the keyword argument 'useRegExp' to
    be false instead of true.  This means that, when doing a update,
    delete, or select, records being selected on string fields will
    be matched using exact matching instead of regular expression
    matching.  If you want to do regular expression matching, pass
    'useRegExp = True' to the method.
       ***IMPORTANT***

* Added a keyword argument to select() called returnType.  If set to
   'object', the result list returned will contain Record objects
   where each field name is an attribute of the object, so you could
   refer to a record's field as plane.speed instead of plane[4].  If
   set to 'dict', the result list returned will contain dictionaries
   where each key is a field name and each value is a field value. 
   If set to 'list', the default, the result is a list of lists.

* Added a new method, insertBatch.  It allows you to insert multiple
   records at one time into a table.  This greatly improves the speed
   of batch inserts.

* Added a new public method called validate.  Calling this method
   with a table name will check each record of that table and
   validate that all of the fields have values of the correct type.
   This can be used to validate data you have put into the table by
    means other than through KirbyBase, perhaps by opening the table
    in a text editor and typing in information.

* Fixed a bug in _closeTable where if an exception occurred it was
   blowing up because the variable 'name' did not exist.

* Fixed a bug in _writeRecord where if an exception occured it was
   blowing up because the variable 'name' did not exist.

* Fixed a bug in _getMatches where I was referencing
   self.field_names as a method instead of as a dictionary.

* Added a new private method, _strToBool, that converts string
   values like 'True' to boolean values.

* Added a new private method, _convertInput, and moved to it the
   code that ensures that the data on an insert is in proper list
   format.  I did this so that I did not have duplicate code in both
   the insert and insertBatch methods.

* To accomodate the fact that users can now send a large batch of
   records to be inserted, I changed _sendSocket so that it first
   sends the length of the database command to the server, then it
   actually sends the command itself, which can now be any length.

* Changed the code in _getMatches to precompile the regular
   expression pattern instead of dynamically compiling every time the
   pattern is compared to a table record.  This should speed up
   queries a little bit.

* Changed the code in select that converts table fields back to
   their native types to be more efficient.

* Changed _sendSocket to use StringIO (actually cStringIO) to hold
   the result set of a client/server-based query instead of just
   capturing the result by concatenating records to one big string.
   In informal testing on large result sets, it shaves a few tenths
   of a second off the query time.


Jamey Cribbs
jcribbs at twmi.rr.com