Add an 'oldagg' backend pointing to kiva.agg

[jwiggins]

This is step 3 from #414. Before switching the image kiva backend to celiagg, we need to make kiva.agg accessible via an unambiguous name.

[corranwebster]

Do we want a kiva/oldagg.py as well? For now it could just import everything from kiva.image.

[jwiggins]

Come to think of it, we might...

[rahulporuri]

The code changes in this PR LGTM but i'm a bit confused about what the next step will be. I thought the idea was to use kiva.celiagg instead of kiva.agg in enable.<toolkit>.image. IIUC, that doesn't work with the proposed changes in this PR. Right?

[jwiggins]

I'm not sure we're ready to "throw the switch" yet, so it's not being done in this PR

[aaronayres35]

I believe this is similar to Rahul's question, but why do we pull from .image not .agg in these?

Because if we eventually have image point to celiagg, then oldagg effectively becomes an alias for celiagg not agg?

[jwiggins]

.agg doesn't exist. That's what .image is. It's kiva.agg. When .image points to celiagg, this module will be filled with the old contents of .image.

[aaronayres35]

.agg doesn't exist.

I am still a little confused. how does the code in kiva/image.py work then as it seems to just import from .agg, see:

enable/kiva/image.py

Line 19 in fe462b8

from .agg import GraphicsContextArray as GraphicsContext

My understanding was that image was basically just an alias pointing at agg, and it would move to point to celiagg in the future (ie those imports change to .celiagg).

If the image contents are going to eventually be moved into these oldagg modules when .image points to celiagg that makes sense. But the image contents seem to be pulling from .agg so it would make sense to me if these files effectively just copied that code over now rather than later?

I believe I am still missing something

[jwiggins]

OK. We're talking past each other a little bit here. .image is very ambiguous here. Are you talking about kiva.image, enable.null.image, enable.qt4.image, or enable.wx.image? When you make a comment on the file enable.null.oldagg and use the name .image, then I'm sure you're talking about enable.null.image. I'm also sure that .agg does not exist in this context, because there is no such thing as enable.null.agg.

I guess I thought it would be simpler to just have these new oldagg modules get their imports directly from their image neighbors on the odd chance that an image module gets updated before the celiagg switchover.

[aaronayres35]

ahhh I see, my mistake, that makes sense now
thank you!

[rahulporuri]

LGTM. IIUC, the next step (which we will do at somepoint in the future) is to move the contents of *.image into the *.oldagg modules and import *.celiagg stuff into the *.image modules. Is that correct John?

[jwiggins]

That is correct

[rahulporuri]

Still LGTM

[jwiggins]

Thanks for the reviews