Recent changes to feature-requests

Windows odbc driver to connect to linux server

Prabakaran Thirumalai — Thu, 14 Apr 2011 17:30:59 -0000

ODBC and JDBC driver porting to windows such that it can connect to CSQLserver running in linux machine.

odbc driver for windows

Prabakaran Thirumalai — Thu, 17 Mar 2011 01:39:06 -0000

port odbc driver and test whether it works with linux CSQL server

tree index support for varchar

Prabakaran Thirumalai — Sun, 13 Mar 2011 13:56:27 -0000

Tree index support for varchar. As our index stores only pointer to key, it should be easy to implement.

Remove obligatoriness of ODBC requirement.

przemoc — Mon, 27 Apr 2009 15:49:59 -0000

unixODBC libraries are not everywhere available. CSQL can work as a simple stand-alone in-memory database, therefore odbc dependency should be reduced only to odbc-related modules.

Currently successful compilation w/o odbc library is impossible. Solving It will require some code changes.

Join Order optimization based on selectivity

Prabakaran Thirumalai — Sat, 31 Jan 2009 16:04:05 -0000

interchanging of left and right tables handles based on selectivity will relax users to specify table names in the FROM clause in any order. Now the user is forced to give low selectivity table at the left hand side.
As predicates are created on the right hand side in case of equi join, this interchange should be done before we set the predicate on the right hand side. see Table::addPredicate() function for more information on this.

optimization for isExclusiveLocked()

Prabakaran Thirumalai — Sat, 31 Jan 2009 05:56:24 -0000

this function is called for every tuple selected. It takes bucket mutex and releases it for checking whether the tuple is locked exclusively by others.
If we use atomic instructions to populate the bucket list (for adding new node and for removing existing node), then reads are not allowed during this by default in a multi processor system.
We need to reorganize locking code to use atomic instructions for bucket node operation and then remove the code which takes mutex for read operation on the bucket elements.
This will improve performance of queries which return more number of records. Also it improves aggregate and join performance with no indexes.

optimization for equi join

Prabakaran Thirumalai — Fri, 30 Jan 2009 10:36:00 -0000

For equi-join, point lookup shall be pushed to the right hand side table to get more performance benefit. Currently it does not use Hash or Tree Index even if is there on right hand table.

Configuration parameters to be moved into shared memory

Kishor Amballi — Mon, 03 Nov 2008 08:22:30 -0000

At present implementation any change in csql.conf file parameter needs the csqlserver to be restarted otherwise the behavior is undefined.

When server boots up it reads csql.conf file and stores each parameter in its address space. When clients connect they also read from csql.conf and
store those parameters in their respective address spaces. Instead the server should store all the server related parameters in shared memory
(system database) and the clients should read these parameters from the shared memory. Clients should read only client related parameters from
the csql.conf file that is specific to each client.

At present the parameters are static and they are not dynamically changeable. This feature should make sure that the parameters are changeable
dynamically.
Related to the bug. 2117333.

Check constraint support

Nihar Ranjan paital — Thu, 25 Sep 2008 09:37:25 -0000

It is a Domain Level Constraint
It is basically used for Validation purpose, which is most important in Real World

Check constraints validate the data values in a given column meet a specific criteria. For example, you could create a check constraint on a varchar2 column so it only can contain the values T or F.

Partial table caching for read only cache

Prabakaran Thirumalai — Sun, 01 Jun 2008 05:38:23 -0000

We support only write through cache tables currently.
We should also support read only cache tables on which we should allow applications to cache only partial set of rows from the table.