[pclasses-devel] S11nNode experiment

[stephan b. <st...@s1...>]

Yo!

Before i finally get to sleep:

i just added a feature to SIO which uses one of s11n's least-known 
features: the ability to work with arbitrary "data node" types. That 
is, it can de/serialize using an arbitrary container type for 
serialized data. We use a node_traits<> type to achieve the node type 
transparency for the core.

Anyway: i took SimplePropertyStore and extended it: P::SIO::S11nNode. i 
simply had to swap out a typedef in SIO to make that the new 
framework-supported node type. Client code doesn't need to change to 
see such a node-swap: test/s11nTest.cpp, which is a fully-functional 
SIO client, worked as-is. It did of course need a recompile to get the 
new class.

Strangely enough: s11n_node is a direct descendant, code-wise, of 
property_store, which is the class i refactored to make 
SimplePropertyStore... which i then subclassed to make S11nNode... by 
copy/pasting the missing functions from s11n_node. Talk about code 
re-use ;).

This was mainly an experiment, in any case. There is a valid reason 
*not* to implement a custom data node type, and that is efficiency: 
s11n::s11n_node is not polymorphic, whereas S11nNode is. Since an 
arbitrary number of nodes can be created for any given tree of data, a 
monomorphic type is probably notably more efficient. (i've tested with 
up to 57k nodes. Theoretically the RAM and system pointer size 
determine the limits.)


-- 
----- st...@s1...   http://s11n.net
"...pleasure is a grace and is not obedient to the commands
of the will." -- Alan W. Watts