Re: [Qexo-general] XqlServlet

[Per Bothner]

Ivelin Ivanov wrote:
> I am considering writing and donating back to you a servlet (XqlServlet)
> which works like
> a JspServlet, only for XQL files.

I assume you're aware that a Qexo program can be compiled directly
to a servlet, so I'm a little unclear exactly how what you're
attempting is different from what exists.  The automation aspects?

> It will eventually detect changes in .xql source files, compile them, load
> and execute.

This stuff *is* missing - i.e. the integation/infrastructure hooks.

> I have done some research and this what I found so far:
>
> 1) You can run an arbitrary file with:
> Shell.runFile(initFile.getPath())
>
> 2) If you extend a KawaServlet you implement apply(CallContext), where the
> CallContext is a ServletCallContext.
>
> What I cannot figure out easily yet is how do I pass the CallContext to the
> Shell, so that the XQL files has access to the request and response
> variables.

I think that's the wrong approach.  It is better to just compile the
XQuery program directly to a servlet.

However, you could create a ServletCallContext, and then set it
with CallContext.setInstance.  Then when Shell calls
CallContext.getInstance it shoudl get that ServletCallContext,
bcause CallContexts are associated with threads.

> BTW, I am using 1.6.98. I know it is not the latest.

Yes, please use the CVS version.  (I'm trying to get 1.7
released, but my day job will delay things.)

> from your example. It appears that $request/requestURI, $request/method and
> $request/pathInfo are not resolved to what is expected.

This is kind-of a kludge and I'm not 100% sure I want to keep
supporting this syntax.  request-uri(), request-method(), and 
request-path-info() are cleaner and more efficient.  They are
documented in the Kawa manual.
--
        --Per Bothner
address@hidden   http://www.bothner.com/per/