I would like to contribute versioning, please give your opinions:

Hi!

I would like to contribute to Xindice and especially I would like to
contribute a two resolvers that implement versioning, like CVS, but
for XML/XIndice.

I am pretty new to Java and I wonder if I can get some pointers on how
to implement this and where to start.

Here is my first thoughts on how I want to implement things. I would
find it great if I can get pointers by some of you guys what you think
about it.

To enable versioning on a database the following configuration will be used:

<xindice>
   <root-collection dbroot="./db/" name="db">
      <queryengine>
          <resolver autoindex="false" versioning-collection="version"
class="org.apache.xindice.core.query.XPathVersionedQueryResolver" />
          <resolver  versioning-collection="version"
class="org.apache.xindice.core.xupdate.XUpdateVersionedQueryResolver"
/>
      </queryengine>
   </root-collection>
</xindice>

Upon initialization the system will create the versioning collection
specified by the "versioning-collection" attribute, this collection
then only stores versioning information.

The _version_-collection then holds all information needed to extract
different revisions and versions from the database, while the db
(root-collection) actually just is unchanged, it's a view of
HEAD/TRUNK.

When a client queries something from the versioned collection
XPathVersionedQueryResolver add attributes to the result including
revision and branch information (something like
version:info="HEAD:1.43").

When a client updates the a versioned collection the
XUpdateVersionedQueryResolver saves the XUpdate snippet in the
versioning-collection.

When a client wants to get an old revision the
XPathVersionedQueryResolver just "combines" all xupdate-snippets from
the versioning-collection and returns a document, I would want to use
a special hierarchy in the versioning collection to enable to just use
XPath to get different versions:

(Of course this is not the final design, I have not even started
coding, but I would like to start a discussion about this)

Consider a collection "db" with an attached versioning collection
called "version", the db-collection initially contains:

<product>
  <name>My Product</name>
  <price currency="USD">32.00</price>
</product>

while the version collection contains (abbreviated for readability):

<branch name="HEAD">
  <revision rev="1.1">
    <info>
       <comment>Initial version</comment>
    </info>
    <data>
      <product>
        <name>My Product</name>
        <price currency="USD">32.00</price>
      </product>
    </data>
  </revision>
</branch>

When the client changes the price from 32.00 to let's say 30.00, the
version collection contains (abbreviated for readability):

<branch name="HEAD">
  <revision rev="1.1">
    <info>
       <comment>Initial version</comment>
    </info>
    <data>
      <product>
        <name>My Product</name>
        <price currency="USD">32.00</price>
      </product>
    </data>
  </revision>
  <revision rev="1.2">
    <info>
       <comment>Change price</comment>
    </info>
    <data>
      <xupdate:update select="/product[1]/price">
        30.00
      </xupdate:update>
    </data>
  </revision>
</branch>

To get a get a special version of a document, one would use the
following XPath expression:

/product[1]/branch[@name='HEAD']/revision[@rev='1.2']/complete

(I have not yet really figured out a XPath format)

As you can see the complete tag does not really exist in the database,
but it should be generated on-the-fly by combining all preceding
revisions.

Likewise the XPath expression:

/product[1]/branch[@name='HEAD']/revision[@rev='1.2']/diff/branch[@name="HEAD"]/revision[@rev='1.1']

Would generate a xml with the differences between HEAD 1.1 and HEAD
1.2 of /product[1].



I wouldn't want this email to become very long, so I will just stop
here, I do really look forward to your opinions though.



Regards,


Mikael Olenfalk