launchpad-dev team mailing list archive

[Gary Poster <gary.poster@xxxxxxxxxxxxx>]

On Nov 28, 2010, at 7:46 AM, Manish Sinha wrote:

> On 11/26/2010 11:34 PM, Robert Collins wrote:
>> 
>> On the dynamic introspection vs compilation angle - I'm skeptical that
>> runtime resolution is a sensible way to handle code generation simply
>> because it has no room for manual fixups, you become very dependent on
>> the compiler logic and input file being complete - and those bugs you
>> referenced show that this isn't a simple or trivial thing to get
>> right.
> 
> Actually I tried generating it dynamically only to find it riddled with bugs.
> One big problem is that WADL standard(still in progress) is somewhat insufficient
> for being processed by a tool. Two biggest hindrances are noted down in these
> two bug reports
> * https://bugs.launchpad.net/launchpad-foundations/+bug/650967
> * https://bugs.launchpad.net/launchpad-foundations/+bug/579516
> 
> Due to this the generated code is never perfect. Leonard himself told that
> WADL is not like WSDL (latter is tool friendly)
> 
> Due to this, every time whenever code is generated it needs to be fixed.  Automatic
> code generation, building or using Reflection.Emit to generate CLR
> assemblies wont be sufficient.

Yes.

As noted before, we recommend following our lead in this regard: have the WADL parser generate what it can, and have a wrapper that adds the lazr.restful interpretations (the equivalent of lazr.restfulclient) and the launchpad-specific interpretations (the equivalent of launchpadlib).

Gary