[SciPy-user] Scipy + wxPython ...
Sat Dec 8 10:02:09 CST 2007
Matthieu Brucher wrote:
> 2007/12/8, stef mientki <firstname.lastname@example.org
> A month ago or so we discussed the possibilities to use Vision as a
> to enhance Scipy to make it more look like some of the commercial
> scientific packages.
> I looked if I could use Vision,
> but as some of you already said: "Vision is not very pretty" ,
> and I have to admit I didn't get it working within a limited
> amount of time.
> Also the other packages, Orange and Elefant,
> were too difficult for me to get them working in a short time.
> So therefor I just started to fork another application I'm working on,
> which just uses wxPython, to see what I could accomplish.
> Although it's just a "proof of concept" ,
> made from a quick and dirty glueing of some (most already existing)
> a large part of the program is already working quit well.
> If you're interested, look here:
> (10MB, 8 minutes)
> Any comment, suggestions, domain expertise are very welcome.
> It's great, but the fact that you do real time job as well splits
> applications in two : the one that need it and teh one that must not
> have it. If I have to iterate on a folder and apply something to each
> element in the folder and then save it, I would use a for loop with a
> sub worflow with the load, process and save boxes inside the for loop
> box, not outside.
Yes, I've no ready solution for that.
Labview uses "time-critical" signals for that.
I was thinking of grouping bricks together, so they form a new brick, in
which closed for loops could exists.
> Now, I think that if an input is connected to something, the
> associated editor should not be displayed.
Good point, also makes the design more consistent.
I now see also an error in the demo, I throw all the controls of the
rotation away, while the interpolation should remain.
> But great job, I didn't create such an application that works that
> well even if I started a long time ago ;)
> Hearing that Enthought is working on something similar,
> I'm still very curious to see their concept.
> Yes, their block_canvas will be a huge competitor with the support of
> the Traits library for modifying and/or displaying variables. For what
> I saw, it will beat Vision hands down (their workflow is a Python
> script directly, so adding new functions is generally straightforward).
I'ld love to see some peek previews ;-)
More information about the SciPy-user