Fix transparent markers in PDF backend. Closes #1410

[mdboom]

No description provided.

[dmcdougall]

@mdboom Does this work with alpha transparency?

[mdboom]

alpha works to the extent that it always did -- i.e. not a lot of fine control. This only addresses the case where mfc="none".

[dmcdougall]

Sweet. This gets my +1 when the Travis tests pass. Also, I'd like to express that I am very impressed with your response on this bug! :)

[efiring]

What I have not figured out is why it did not work the way it was; it would seem that check_gc already stores rgbFace in gc._fillcolor, and that gc.filp() checks that for None. So it seems odd to need this extra block of code here. It might be worth putting in a comment to explain it.

[dmcdougall]

This is check_gc:

def check_gc(self, gc, fillcolor=None):
    orig_fill = gc._fillcolor
    gc._fillcolor = fillcolor

    delta = self.gc.delta(gc)
    if delta: self.file.output(*delta)

    # Restore gc to avoid unwanted side effects
    gc._fillcolor = orig_fill

It looks like the last line restores it. Might that be the cause?

[efiring]

On 2012/10/17 11:29 AM, Damon McDougall wrote:

In lib/matplotlib/backends/backend_pdf.py:

@@ -1556,7 +1556,10 @@ def draw_markers(self, gc, marker_path, marker_trans, path, trans, rgbFace=None)
return

     self.check_gc(gc, rgbFace)

•    fillp = gc.fillp()
  

•    if rgbFace:
  

This is |check_gc|:

def check_gc(self, gc, fillcolor=None):
orig_fill = gc._fillcolor
gc._fillcolor = fillcolor

     delta  =  self.gc.delta(gc)
     if  delta:  self.file.output(*delta)

     # Restore gc to avoid unwanted side effects
     gc._fillcolor  =  orig_fill

It looks like the last line restores it. Might that be the cause?

Yes, I missed that.

@mdboom,
I think a better alternative would be to modify the gc.fillp() method to
take an optional rgbFace arg (not a kwarg); draw_markers would then call
it with that arg, and all other uses would call it without. If the arg
is present, it would be used inside fillp() in place of self._fillcolor.