python - Fastest pcolor-like method with log scale support -

i've been experimenting different pcolor-like methods find fastest 1 support log scaling (of y axis) , kind of image smoothing (shading or interpolation). i'm asking here make sure i'm not missing something.

the goal

the goal plot 2d array s representing scalogram, t array representing time samples , f representing logarithmically sampled frequencies. shapes aren't pcolor-style material, because represent points, not bounds: s.shape == (f.size, t.size). ideally want method takes point coordinates, not bounds.

the result should fast enough interactive plotting sizes map(log10, s.shape) around (3-4, 2-3).

my research far

• matplotlib.image.nonuniformimage
  • provides interface want, expects point coordinates
  • is fast
  • supports bilinear interpolation
  • does not support log scales. axis ticks rescaled after call yscale('log'), image isn't.
  • uses matplotlib.image.pcolor internally, undocumented
    • the code seems show not support larger arrays
• matplotlib.image.pcolorimage similar above
  • has pcolor-style interface
  • has same deficiencies regarding log scaling.
  • uses matplotlib.image.pcolor2 internally, undocumented
    • judging code not support interpolation
• matplotlib.axes.axes.pcolorfast:
  • for 1d point coordinates uses pcolorimage
  • for 2d point coordinates (created e.g. meshgrid) uses quadmesh supports log scaling.
    • does not support gouraud shading
• matplotlib.axes.axes.pcolormesh seems winner far
  • supports gouraud shading, quite slow
  • also uses quadmesh, not find significant speedups in pcolorfast in quadmesh case (when not using gouraud shading).

alternative approaches

an alternative use contourf, many levels slow down too.

another possibility use nonuniformimage or imshow, scale coordinates , handle axis ticks manually.