[SciPy-user] Presentation of pymachine, a python package for machine learning

[Matthieu Brucher]

>
> You can implement a for filter with a fixed number of outputs ( (el1,
> el2? el3) in you example). I do agree that zip, sequence unpacking, etc
> are a bit more tricky. These might need special approach. However for the
> general case I think you are better of having fixed numbers of outputs.
> This will make designing your UI way simpler. I think a node needs to
> display how many inputs and outputs it has.



I agree with your point. I have to digg this a little more to find a usable
solution.


The compute method is call automatically when the inputs are changed.
> There is no need for some manual update code. Similarily, dialogs can
> update objects live, and can reflect the current value of the attributes,
> even if they change.
>
> This will really make your code simpler and more readable, I think.
>

Exactly what is needed in such an application, I would say ! From the table
of contents of yout tutorial, I had come to the same conclusion, so I will
read it with attention - be cause, there is a part on threads too ;) -

Matthieu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20070514/e6260efa/attachment.html>