Something up with transforms?

[Stuart Axon]

I ran the following test in shoebot and nodebox1:

size(400, 400)
translate(100, 100)
rect(0, 0, 10, 10)
push()
rotate(45)
translate(50, 0)
rect(0, 0, 10, 10)
pop()


The output is quite different, so something is up with our coordinate system. (Possibly this is already known) - this is probably the same thing that happens when running the planets example.


Graphical output is attached.

A couple of side notes:
- We should probably start the canvas with white, to match nodebox.
- Right click and save-as png resulted in a blank image.

[ricardo lafuente]

Gah, crap.

This seems to be an issue with concatenating transforms. I briefly went
through some of the transforms related examples, and they look okay.
Even the planets example looks like it works.

Another issue i found (mostly as a reminder for myself to fix this
asap), is that background() doesn't scale to the actual dimensions set
by size(). IIRC this was reported already, now on to fix it...

Post by Stuart Axon
size(400, 400)
translate(100, 100)
rect(0, 0, 10, 10)
push()
rotate(45)
translate(50, 0)
rect(0, 0, 10, 10)
pop()
The output is quite different, so something is up with our coordinate system. (Possibly this is already known) - this is probably the same thing that happens when running the planets example.
Graphical output is attached.
- We should probably start the canvas with white, to match nodebox.
- Right click and save-as png resulted in a blank image.
_______________________________________________
Shoebot-devel mailing list
http://lists.tinkerhouse.net/listinfo.cgi/shoebot-devel-tinkerhouse.net

[Stuart Axon]

Sent: Tue, January 26, 2010 2:02:53 PM
Subject: Re: [shoebot-devel] Something up with transforms?
Gah, crap.
This seems to be an issue with concatenating transforms. I briefly went
through some of the transforms related examples, and they look okay.
Even the planets example looks like it works.

I'll try this again... however, I'd recommend installing python 2.6 and the
relevant new versions of the libraries, to make sure something in cairo
hasn't changed too.

Another issue i found (mostly as a reminder for myself to fix this
asap), is that background() doesn't scale to the actual dimensions set
by size(). IIRC this was reported already, now on to fix it...

Thats true - background() should be immune to transforms, at the moment it
will rotate and move too (my earlier test bot tests some of this).


If we do make a proper acid test bot (acidbot?) then we should think about
how it should work.

I guess, at a start -
self contained test functions, with docstrings saying what they test
(my original test bot was a little ad-hock and didn't do lots of this).

Post by Stuart Axon
size(400, 400)
translate(100, 100)
rect(0, 0, 10, 10)
push()
rotate(45)
translate(50, 0)
rect(0, 0, 10, 10)
pop()
The output is quite different, so something is up with our coordinate system.

(Possibly this is already known) - this is probably the same thing that happens
when running the planets example.

Post by Stuart Axon
Graphical output is attached.
- We should probably start the canvas with white, to match nodebox.
- Right click and save-as png resulted in a blank image.
_______________________________________________
Shoebot-devel mailing list
http://lists.tinkerhouse.net/listinfo.cgi/shoebot-devel-tinkerhouse.net

_______________________________________________
Shoebot-devel mailing list
http://lists.tinkerhouse.net/listinfo.cgi/shoebot-devel-tinkerhouse.net

[ricardo lafuente]

Post by Stuart Axon
I'll try this again... however, I'd recommend installing python 2.6 and the
relevant new versions of the libraries, to make sure something in cairo
hasn't changed too.

My setup has Python 2.6, Cairo 1.8.8 and PyCairo 1.8.6
(standard Ubuntu Karmic install)

Post by Stuart Axon
If we do make a proper acid test bot (acidbot?) then we should think about
how it should work.
I guess, at a start -
self contained test functions, with docstrings saying what they test
(my original test bot was a little ad-hock and didn't do lots of this).

Hmm, this is a good question. What should our testing framework look like?

I was thinking along the lines of using standard Unit Tests, which would
then compare the bot results with a reference image, preferably
generated by Nodebox to ensure compatibility.

But i'm curious about your AcidBot idea. Were you thinking about
actually porting the Acid tests? Or making a similar all-feature testing
toolkit?

[Stuart Axon]

Sent: Tue, January 26, 2010 3:30:29 PM
Subject: Re: [shoebot-devel] Testing? (was: Something up with transforms?)

Post by Stuart Axon
I'll try this again... however, I'd recommend installing python 2.6 and the
relevant new versions of the libraries, to make sure something in cairo
hasn't changed too.

My setup has Python 2.6, Cairo 1.8.8 and PyCairo 1.8.6
(standard Ubuntu Karmic install)

Post by Stuart Axon
If we do make a proper acid test bot (acidbot?) then we should think about
how it should work.
I guess, at a start -
self contained test functions, with docstrings saying what they test
(my original test bot was a little ad-hock and didn't do lots of this).

Hmm, this is a good question. What should our testing framework look like?
I was thinking along the lines of using standard Unit Tests, which would then
compare the bot results with a reference image, preferably generated by Nodebox
to ensure compatibility.
But i'm curious about your AcidBot idea. Were you thinking about actually
porting the Acid tests? Or making a similar all-feature testing toolkit?

I guess the 'all features testing framework'... except we don't need to test all features, just ones where there are deviations, to start with.

I like the idea of graphical output and checking against a reference graphic as this is easily understood. Whether it needs to be a single graphic, or many is another question.
(For example testing the transform stuff is fairly easy this way).

If we want to run regression tests with a graphical framework, this should be easy(ish) as the output will match previous runs - testing against nodebox will be more difficult as we will never 100% match (different antialiasing etc).

Using a unit test framework sounds like a great idea if we can work out how to do it.

_______________________________________________
Shoebot-devel mailing list
http://lists.tinkerhouse.net/listinfo.cgi/shoebot-devel-tinkerhouse.net