Question

I have to do several subplots, each with an inset at the same custom location relative to the subplot.

It can be done in principle with axes_grid1 as given in this example, however, the location parameter loc= in inset_axes or zoomed_inset_axes is not specific enough for my purpose.

I have also tried with GridSpec‘s get_grid_positions, but I cannot make sense of the parameters it returns.

How can I get insets to look the same in each subplot without being forced to put them at specific locations?

Answer 1

You can do it by hand:

ncols = 2
nrows = 2

inset_hfrac = .3
inset_vfrac = .3

inset_hfrac_offset = .6
inset_vfrac_offset = .6

top_pad = .1
bottom_pad = .1
left_pad = .1
right_pad = .1

hspace = .1
vspace = .1

ax_width = (1 - left_pad - right_pad - (ncols - 1) * hspace) / ncols
ax_height = (1 - top_pad - bottom_pad - (nrows - 1) * vspace) / nrows

fig = figure()

ax_lst = []
for j in range(ncols):
    for k in range(nrows):
        a_bottom = bottom_pad + k * ( ax_height + vspace)
        a_left = left_pad + j * (ax_width + hspace)

        inset_bottom = a_bottom + inset_vfrac_offset * ax_height
        inset_left = a_left + inset_hfrac_offset * ax_width

        ax = fig.add_axes([a_left, a_bottom, ax_width, ax_height])
        ax_in = fig.add_axes([inset_left, inset_bottom, ax_width * inset_hfrac, ax_height *  inset_vfrac])
        ax_lst.append((ax,ax_in))

This costs you the ability to use tight_layout, but gains you a tremendous amount of control.