Hi everyone, 

I'm having trouble getting error bars to appear on the y-axis when using a split y-axis, has anyone has any experience with this?
Using the minimal example from the PyX website, I've got:


from pyx import *


g = graph.graphxy(width=8, y=graph.axis.split())

g.plot(graph.data.points([(0.1, (0,0.1), 0.08),

                          (0.5, (0,0.2), 0.08),

                          (0.9, (0,0.3), 0.08),

                          (101, (1,0.7), 0.08),

                          (105, (1,0.8), 0.08),

                          (109, (1,0.9), 0.08)], x=1, y=2, dy=3),

  [graph.style.symbol(),graph.style.errorbar()])

g.writePDFfile("minimal")



which is just adding a third column to the data points, and assigning it to dy. This works if the x-axis is split (with y-errorbars), but doesn't work if the y-axis is split. Also, it doesn't generate any error message - the graph is created, it's just missing the errorbars.


I've tried changing the third (dy) data column to be a tuple like the y-column, ie:


g.plot(graph.data.points([(0.1, (0,0.1), (0,0.08)), ...


however this throws the error 


  File "/Library/Python/2.5/site-packages/pyx/graph/axis/axis.py", line 348, in adjustaxis

    assert len(value) == 2, "tuple of size two expected by bar axis '%s'" % errorname

AssertionError: tuple of size two expected by bar axis 'y'



Sticking in a 'print type(value), value)' statement just before this assert, the data going in is:


<type 'tuple'> (0, 0.10000000000000001)

<type 'tuple'> (0, 0.20000000000000001)

<type 'tuple'> (0, 0.29999999999999999)

<type 'tuple'> (1, 0.69999999999999996)

<type 'tuple'> (1, 0.80000000000000004)

<type 'tuple'> (1, 0.90000000000000002)

<type 'tuple'> (0, 0.10000000000000001, 0, 0.080000000000000002)


so obviously this is cycling over the y-data, but then when it gets to the dy column, it still includes the y-data (or the dy has just been added to it).


Anyone any ideas? 

Do the dy values have to be assigned to the subcanvas individually?


Any help is much appreciated!

Also, many thanks to Andre et al for a great piece of software!


Cheers,

Dave.