Re: [Qexo-general] XQParser

[Per Bothner]

Chen Ya Bing wrote:
> > You can create an instance of XQuery, and then call
> > the parseFile method.  That returns a Compilation state.
> > Calling getModule on the latter gives you an abstract
> > syntax tree of the parser.
>
> but I found the parseFile method returns ModuleExp object, right? and the
> object do not have a method called getModule.

Sorry, I should have explained:  I've recently changed how Compilation
is used, and this changed a number of other things.  It used to be
that a Compilation was allocated late in the process, *after* parsing
and conversion to Expression.  I've recently changed the logic so that
the Translation class (used for Scheme/Lisp but not XQuery) inherits
from Compilation, and a Compilation is now created as part of the
parsing process.

So I was referring to the CVS version of Kawa, where parseFile does
return a Compilation.  However, in the released 1.6.99 version
parseFile returns a ModuleExp, which is the actual syntax tree.

> there is print method only.
> or can I use print method to print the syntax expression?

Yes.

> another question
> is, can I separate different parts of a xquery, such as For/Let clause,
> Where clause and Return clause using some methods?

You can, but there aren't any convenience methods.  The framework is
designed for synthesis, not analysis.

To analyzie and/or modify the syntax tree, Kawa uses the ExpWalker
framework.  This recurses through the tree, making various changes,
tests, optimizations, and collecting information.  There are 5
sub-classes of ExpWalker, one for each (intermediate) pass through the
tree.  (Some of them could possibly be combined.)

> yes, I printed it out, but I am sorry to say that it is hard for me
> understand the syntax tree. can I get a picture of how the tree is like in
> terms of query operator tree?

The structure of the tree is language-independent:  There are no
XQuery-specific Expression sub-classes.  Instead, XQuery-specific
operators are represented as function calls using ApplyExp.
The various clauses of a FLWR-expression are represented using
LambdaExp (lambda expressions), which are expression that are
anonymous functions.

> in the case where the xquery is in FLWR expression, I wonder whether the
> tree is constructed clause by clause. I mean, is the tree composed of
> different parts cooresponding to different clauses?

Yes, mostly, but the tree is designed for optimizatio and code
generation, as well as language-independence, so things may
represented in a way that may seem strange.
--
        --Per Bothner
address@hidden   http://www.bothner.com/per/