No registered users in community xowiki
in last 10 minutes
in last 10 minutes
[Xotcl] Announcing Storm v0.1 A transparent queryable object storage for Tcl/XOTcl
From: Kristoffer Lawson <setok_at_scred.com>
Date: Tue, 18 May 2010 00:53:05 +0300
Continuing on where Spindle leaves off, I created a system which again
reflects how I believe storage should work. After endless problems
with RDBMS and the object-relational mismatch, I decided to hell with
that, and built a transparent persistence layer which can use a file
or Sqlite backend (no reason why others could not be used too). The
idea is that any class created from the basic PersistenceClass
metaclass can be attached to storage. Changes are automatically
written out. Even more interestingly objects which are referred to,
but which do not exist in memory, are automatically loaded as well.
You can even destroy them at will and, when required, they'll happily
be loaded back in. There is also a simple query language for getting
objects from their class.
It's not perfect, and a lot could be done with it still, but I like
the idea. For many things full blown direct RDBMS access is just not
necessary, and can cause major headaches. I also think that a lot
could be done to optimise the Storm system so that performance
differences wouldn't be huge, for many cases. Spindle (the web
framework) does not require Storm in any way, and you can use any
storage backend you want with it. I did, however, create them
simultaneously so perhaps they subconsciously share some fabric.
http://github.com/Setok/Storm
Date: Tue, 18 May 2010 00:53:05 +0300
Continuing on where Spindle leaves off, I created a system which again
reflects how I believe storage should work. After endless problems
with RDBMS and the object-relational mismatch, I decided to hell with
that, and built a transparent persistence layer which can use a file
or Sqlite backend (no reason why others could not be used too). The
idea is that any class created from the basic PersistenceClass
metaclass can be attached to storage. Changes are automatically
written out. Even more interestingly objects which are referred to,
but which do not exist in memory, are automatically loaded as well.
You can even destroy them at will and, when required, they'll happily
be loaded back in. There is also a simple query language for getting
objects from their class.
It's not perfect, and a lot could be done with it still, but I like
the idea. For many things full blown direct RDBMS access is just not
necessary, and can cause major headaches. I also think that a lot
could be done to optimise the Storm system so that performance
differences wouldn't be huge, for many cases. Spindle (the web
framework) does not require Storm in any way, and you can use any
storage backend you want with it. I did, however, create them
simultaneously so perhaps they subconsciously share some fabric.
http://github.com/Setok/Storm
-- Kristoffer Lawson, Co-Founder, Scred // http://www.scred.com/