[antlr-interest] Re: ASTFactory initialization inconsistencies in Parser & TreeParser

[micheal_jor]

Hi Ric,

> > Why do ANTLR Parsers initialize their ASTFactory appropriately 
whilst
> > TreeParsers have to operate with an uninitialized factory?.
> 
> Which kind of initialization do you mean? You mean the setting of 
an AST
> factory in the treewalker or the initialization of that factory? 
Both have
> to be done manually.

The initialization of the default factory -- pre-loading it with the 
Token-ID-to-ASTNodeType mappings for all the known tokens.

> With the use of a AST superfactory a chain of parser/treeparsers 
must have
> one superfactory (or exact copies). Which will be initialized in 
several
> steps, in each generated (tree)parser an initializeASTFactory() 
method is
> added which adds the AST type data for that parser to the factory. 
It is up
> to the user to ensure things happen in the right order.

OK. This sounds more flexible. initiializeASTFactory() can be called 
after setASTFactory() is called.

> Since antlr does not impose an order in import/exportvocabs in a 
parser
> hierarchy it makes no sense to have a default. Since there is no 
sane
> default.

The sensible default might be to register the closure of all 
importVocab/exportVocab/in-grammar Token IDs in each generated 
lexer/parser/treeparser.

> > The factory fails on create()-ing/dup()-ping valid token ID 
(obtained from
> > the Parser by reflection) when treeparser rules are invoked by 
reflection.
> 
> I guess it is bombing out on a lack of factory? If you're adopting 
the same
> super factory strategy used in C++, then I'd check out the C++ 
examples as
> well.

It's bombing out on account of no Token IDs being registered in the 
default factory.

Cheers,

Micheal


 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/