Eclipse Community Forums: M2M (model-to-model transformation) » [QVTO] Large script

Home » Archived » M2M (model-to-model transformation) » [QVTO] Large script - editor slow

[QVTO] Large script - editor slow [message #869583]

Thu, 03 May 2012 12:15

Eclipse User

Hello,

I have a large transformation script and the qvto editor becomes
extremely slow. (I'm not sure yet if this is also a problem of using
custom types which the editor doesn't know). Anyway I can edit fragments
while strategically commenting out other areas, but when I enable all
qvto code, the editor chokes on the script (CPU busy for many minutes).

I may have issues in the script, but is there a way to split scripts
into smaller fragments?

TIA
Marius

Re: [QVTO] Large script - editor slow [message #869585 is a reply to message #869583]

Thu, 03 May 2012 12:21

Eclipse User

On 03.05.2012 18:15, Marius Gröger wrote:
> while strategically commenting out other areas, but when I enable all
> qvto code, the editor chokes on the script (CPU busy for many minutes).

What's even worse: when the editor is chewing on the script, I can
hardly use Eclipse anymore. If I open any other(!) editor window,
Eclipse hangs completely and the CPU is idle. I then need to kill
Eclipse and start over. Not good.

Marius

Re: [QVTO] Large script - editor slow [message #869742 is a reply to message #869583]

Fri, 04 May 2012 03:19

Eclipse User

On 03.05.2012 18:15, Marius Gröger wrote:
> I have a large transformation script and the qvto editor becomes
> extremely slow. (I'm not sure yet if this is also a problem of using
> custom types which the editor doesn't know). Anyway I can edit fragments

Turns out this was indeed the problem. In my workspace I have the
referenced Ecore models, but the qvt editor doesn't know about them
(rather, the models aren't registered in the global EMF registry). So I
had written a small helper plugin which loads a list of .ecore models
into the global EMF registry on startup. However, when such a model
references custom types, the qvto chokes badly on each reference trying
to resolve that custom type. I'm now only loading the XSD models with
that helper, and deploy the generated ecore models as plugins into my
dropins/ folder. Now the model is complete and the qvt editor is happy
(and so am I.)

Regards
Marius

Re: [QVTO] Large script - editor slow [message #869820 is a reply to message #869742]

Fri, 04 May 2012 08:02

Eclipse User

Hi

I suspect that you have side stepped the problem. EMF works quite well
with dynamic models; a bit slower but not dramatically so. Perhaps your
problem was that you were repeatedly loading the model again and again.
Using a single Java model may just have avoided the repetition.

Regards

Ed Willink

On 04/05/2012 08:19, Marius Gröger wrote:
> On 03.05.2012 18:15, Marius Gröger wrote:
>> I have a large transformation script and the qvto editor becomes
>> extremely slow. (I'm not sure yet if this is also a problem of using
>> custom types which the editor doesn't know). Anyway I can edit fragments
> Turns out this was indeed the problem. In my workspace I have the
> referenced Ecore models, but the qvt editor doesn't know about them
> (rather, the models aren't registered in the global EMF registry). So I
> had written a small helper plugin which loads a list of .ecore models
> into the global EMF registry on startup. However, when such a model
> references custom types, the qvto chokes badly on each reference trying
> to resolve that custom type. I'm now only loading the XSD models with
> that helper, and deploy the generated ecore models as plugins into my
> dropins/ folder. Now the model is complete and the qvt editor is happy
> (and so am I.)
>
> Regards
> Marius
>

Re: [QVTO] Large script - editor slow [message #869841 is a reply to message #869820]

Fri, 04 May 2012 09:35

Eclipse User

On 04.05.2012 14:02, Ed Willink wrote:
> I suspect that you have side stepped the problem. EMF works quite well
> with dynamic models; a bit slower but not dramatically so. Perhaps your
> problem was that you were repeatedly loading the model again and again.
> Using a single Java model may just have avoided the repetition.

Perhaps... anyway my (hopefully final) approach to this is to simply
start my Ecore model in a runtime workbench, add my XSD model there also
using my special startup-loader plugin and edit the QVT there in a
project that's linked to my original workspace. That way I don't need to
re-deploy and restart Eclipse, I merely need to restart the runtime
workbench which is much more lightweight.

Everything works great now, and I really love QVT. Hopefully development
will continue (I'm afraid I won't be able to give much back... well,
perhaps a bug report here and there).

Regards
Marius

Previous Topic:	Model contains "definitions", Ecore specifies "Definition"
Next Topic:	Connecting RP to OM2M

Goto Forum:

-=] Back to Top [=-

Current Time: Wed Apr 30 11:53:35 EDT 2025

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter