In my opinion, TGE is a very good tool, especially for new languages when the grammar (and its tree structure) is not well established.
The need for language experimentation is not execution performance,
but rapid development and clean design.
The functional paradigm of TGE (like XSLT, transforming tree) allows to create clean design.
TGE could be used in 2 different stages : parse -> PAST, and PAST -> POST.

The current syntax allows to specified a target language :
    transform name (pattern) :language('PIR') {
        # action
    }
Currently, only PIR is supported.
I think that the support of NQP (action written in NQP, and NQP generation) will be a great improvement.
By this way, TGE becomes a NQP preprocessor.

Comments most welcome.

François Perrad


Reply via email to