[antlr-interest] Why use trees rather than building up your own code?

[Colin Yates]

Hi all,

I understand that the ANTLR grammar can produce an intermediate form tree
which is then parsed or you can provide code snippets to build up your own
internal representation.

I have always used the second, for example:

departmentDefinition returns[DepartmentDefinition value]

        @init {$value = new DepartmentDefinition(); }

        : DEPARTMENT singleReferenceDataDefinition {
$value.setDepartmentDefinition($singleReferenceDataDefinition.value);
} departmentStatements[$value]+




This seems a much more natural approach, and I have implemented a handful of
pretty complex grammars this way, but I get the sense that trees are the
preferred approach.

My question is why - what benefit do they get you?

The reason I want to know is I am about to embark on a library for editing
documents defined by an ANTLR grammar with syntax highlighting etc. (see my
next email :)) and I want to make sure I am not sending myself up an alley
:)

Col