Re: [Libreoffice] errors in idl files

On Thu, 2011-11-24 at 14:42 +0100, Tomas Hlavaty wrote:
> - I write my application in a dynamic programming language with no
>   dependencies on libreoffice code;

        Ok :-) so you would have a native implementation of the UNO RPC
protocol?

> - My application reads the type library (RDB or better a simple XML
>   alternative) of the actually installed libreoffice on the target
>   machines (possibly different versions).

        Fine.

> - My application communicates with libreoffice via UNO using the correct
>   version of the type library (which I didn't exactly know before
>   installing my application on the target machine).

        OK; so you rely on your dynamic language binding to elide / add extra
fields etc. - make sense.

> At the moment, I parse IDL files and hope (or have to make sure) that
> the installed libreoffice is compatible.

        So - in -theory- (for all interesting cases), the UNO API is frozen (at
the moment). In the future we plan to change that, so we can avoid
piling up hacks).

> To an extend, I could probably build the type library dynamically by
> querying all interfaces instead of reading the RDB file, but I think
> that would be unnecessarily slow.

        You could perhaps introspect them via UNO remotely, and cache the data
per interface; not sure how many interfaces people typically use - but,
prolly not -that- many :-)

> There seems to be some Python bridge to libreoffice. Does it use the
> store code underneath a FFI?

        Sure - it links to the URE run-time; we have a stable C ABI that we
export that has all the basic symbols and type information you need; and
of course that makes running in-proc. much cleaner.

        HTH,

                Michael.

-- 
michael.meeks@suse.com  <><, Pseudo Engineer, itinerant idiot