On 04/02/2012 01:46 PM, Darran Lofthouse wrote:
Different subsystems have taken different approaches to this, some
completely fork the parser for each version others adapt the existing
parser to support the additional versions as they are added.
My one comment on this is writing a single parser that can support
multiple versions is very complex and so far I have seen a lot of
mistakes where this is attempted - personally I believe that the long
term maintainability of the parsers should take higher priority than
minimising the amount of code.
Thanks for the input.
For these changes, I opted to subclass the parser and let the child
class handle the new attributes.
However, if the changes were more substantial (modification of tree
hierarchy, etc.), I agree that a new standalone parser would be simpler
to maintain (at the cost of having code duplication in case of bug fix).
jeff
--
Jeff Mesnil
JBoss, a division of Red Hat
http://jmesnil.net/