Alright, this was my shot at it:

https://gist.github.com/goretkin/d34f287062f6b27e8846

It seems like the assumption of an axis-aligned axes is baked in pretty

strongly, and maybe that's for the best...

I couldn't figure out how tick lines work and why they weren't rotated. The

Line2D object has the correct transform. They are lines whose xy_data

contain only one coordinate, though something must specify the other end of

the tick line (what gives it its length otherwise?) and that is assuming

that the axes is axis-aligned.

In [4]: [l.get_xydata() for l in ax.xaxis.get_ticklines()]

Out[4]:

[array([[ 0., 0.]]),

array([[ 0., 1.]]),

array([[ 2., 0.]]),

array([[ 2., 1.]]),

array([[ 4., 0.]]),

array([[ 4., 1.]]),

array([[ 6., 0.]]),

array([[ 6., 1.]]),

array([[ 8., 0.]]),

array([[ 8., 1.]]),

array([[ 10., 0.]]),

array([[ 10., 1.]]),

array([[ 12., 0.]]),

array([[ 12., 1.]])]

## ···

On Fri, Feb 26, 2016 at 3:12 PM, Gustavo Goretkin < gustavo.goretkin at gmail.com> wrote:

Hello,

(sorry if you get this message twice. I originally posted to the old

sourceforge list)

How can I add an axes to a figure, but specify the transform from axes

coordinates to figure coordinates? I want a set of axes (ticks, spines,

data, everything) that are rotated?

Figure.add_axes [1] accepts a box left-bottom-width-height axis aligned

box specification. I can't follow in the code where a transform is made out

of that, but I basically want a generic transform. Do tick labels, ticks,

titles, etc. assume that the axes-to-figure transform is purely a

translation?

[1]

https://github.com/matplotlib/matplotlib/blob/e12d103a11cd08adbed348547bdb8182e414e0bb/lib/matplotlib/figure.py#L840

Thanks,

Gustavo

-------------- next part --------------

An HTML attachment was scrubbed...

URL: <http://mail.python.org/pipermail/matplotlib-users/attachments/20160229/50112bf4/attachment.html>