Mon, 12 May 1997 18:45:20 +0000 (GMT)
> Georg Fuellen wrote:
> > SteveC wrote,
> > > I very much agree. I have a number of examples of how I am using Perl
> > > and Java in precisely this manner. One of the big issues with the BioWidget
> > > folks is to avoid getting caught up in sematics. "Drive sematics out," they
> > > say (after David Searls). They want to create a semantically neutral system
> > > for displaying and manipulating graphically any kind of genomic data. But
> > > at some point you must deal with the semantics of biological objects
> > > (primarily, this is a data model issue, I believe). I think Perl is well
> > > suited to deal with semantic issues. This could be a good angle to present
> > > at the conference. For example, we could cast Perl as an intermediary
> > > between a semantically complex biological object and a semantically neutral
> > > Java bioWidget. What do you think?
> > This looks exciting, but I'm not clear in which way the
> > ``semantically complex biological object'' is represented, so that Perl can
> > act as an intermediary between this object and a Java bioWidget.
> I'm not quite sure myself, but the basic idea would be to use a parallel
> object schema in both the Perl and Java to facilitate mapping between
> Perl and Java objects.
Did you see:
From: firstname.lastname@example.org (Bart Schuller)
Subject: Re: Perl CORBA IIOP module?
Date: 10 May 1997 13:12:31 GMT
Organization: Lunatech Research
[comp.lang.perl is long dead, Newsgroups trimmed]
In article <email@example.com>,
Oracle News 2 <firstname.lastname@example.org> wrote:
>A related question. Is there any CORBA-Perl mapping definition underway
>right now? I understand that Perl does not have the notion of datatypes
>which makes such a mapping difficult. Has anyone thought about how to
>overcome this difficulty?
I've set up a mailinglist to discuss these issues, as well as my own
implementation of CORBA in Perl. The list is at email@example.com
Details on htpp://www.lunatech.com/cope/
For your question: A CORBA mapping is not hard. Perl does have plenty of
datatypes, it's just that they are not strongly typed.
For example, a short and a string will both map to the same perl datatype
(a scalar). But a stub method knows that for example the CORBA object
expects a short and so interprets the parameter as a short. You could
hand this method a string and it will probably be interpreted as 0. That's
a case of "Don't do that, then". After all, the IDL for the method told you
it expects a number.
The same goes for other datatypes. I represent structs as associative-
arrays. What happens if you put more keys in them? That's undefined.
The bottom line is: as long as you remember that you're talking to a
strongly typed system and play by the rules of that system, all will
Bart Schuller firstname.lastname@example.org At Lunalabs, where the
Lunatech Research http://www.lunatech.com/ future is made today..
Partner of The Perl Institute http://www.perl.org/ Linux http://www.li.org/
This seems to touch some of the issues re: strongly-typed Java <-> Perl...
> > Also, am I right that ``semantically neutral'' means that a bioWidget
> > can display any piece of data that conforms to a certain syntax, or ?
> Yes. A bioWidget is expected to conform to an common API that allows
> for various display tasks such as selection, highlighting, and coordinated
> scrolling/zooming. This permits a consistent interface across widgets.
> At some level, though, there has to be some semantic tie-in so that a
> specific bioWidget knows how handle the peculiarities of biological
> information from a specific domain. One could imagine there being
> subtypes of bioWidgets that specify more about the domain data which
> would act as a base class for all widgets of that type (e.g.,
> SequenceWidget, MapWidget).
> Check the bioWidget homepage for more about this:
I hope to take an intensive look at all this in a month or so.