On this page
Bring matplotlib figures¶
i.from_matplotlib() reads what a matplotlib figure plots and redraws it as
Inklet charts. Your plotting code stays as it is; the result gets Inklet's
print-sized type, measured layout, aligned panels and panel letters.
import matplotlib.pyplot as plt
import inklet as i
fig, (left, right) = plt.subplots(1, 2, figsize=(7.2, 2.8))
left.plot(times, signal, label='treated')
left.fill_between(times, lower, upper, alpha=0.25)
right.bar(groups, means, yerr=errors)
...
i.from_matplotlib(fig).save('figure.pdf')
The matplotlib figure, as matplotlib draws it:
The same figure object after i.from_matplotlib(fig):
The bridge reads data, not pixels. Every mark is vector output and the result is an ordinary chart or layout. You can add marks, change the preset or check the report like any other chart.
What carries over¶
| matplotlib | Inklet |
|---|---|
plot (lines, markers, markevery, dashes) |
line and scatter |
scatter (sizes, colours, c= with a colormap) |
scatter, with a colour ramp and bar |
bar / barh / hist |
bars |
errorbar, bar(yerr=) |
errorbars with the series' line and markers |
fill_between |
band |
axhline / axvline |
hline / vline |
imshow |
matrix with a colour bar |
bar(color=[...]) |
bars, one colour per bar |
bar(alpha=), plot(mfc='none'), scatter(facecolors='none') |
Opacity, hollow markers |
errorbar(capsize=), linewidth= |
Error-bar caps, line width |
text, annotate (data coordinates) |
text, annotate |
fig.colorbar(im, label=), cb.set_label() |
The colour bar's title |
set_xticks(ticks, labels), invert_yaxis(), grid(True) |
Ticks as placed, inverted axes, gridlines |
| Axis labels, title, log scales, explicit limits, categorical ticks, legend | Chart options |
| A grid of subplots | A layout with panel letters; a colour bar beside a panel keeps the grid |
matplotlib's automatic colour cycle (C0, C1, ...) maps onto the preset
palette slot by slot, so series that shared a colour still share one. Colours
you set explicitly are kept. Pass keep_colors=True to keep the cycle
colours as well.
The width follows the matplotlib figure, snapped to a single (89 mm) or
double (183 mm) column when it is close to one. Pass width= to override it,
and style= or palette= to choose a preset or
palette.
What does not¶
Nothing is dropped silently. Each property the bridge does not carry is listed
in a MatplotlibWarning and on the chart's skipped list, so a figure that
loses something says what it lost. The cases:
- Hatches on bars and bands are drawn solid.
- Outlines and fills: bar and marker edges in their own colour are drawn without them, and bars with no fill are drawn in the default colour.
- Images: colour images (RGB or RGBA pixels) are not kept. Colour scales
other than linear, such as
LogNorm, are drawn linear, and a colour map Inklet does not know is drawn as viridis. - Markers: shapes other than circle, square, triangle, diamond, cross and plus are drawn as circles. Hollow markers stay hollow.
- Lines:
drawstyle="steps"is drawn as straight segments. Dashed lines fromhlinesandvlinesare drawn solid, and a labelled set of lines in several colours loses its label. - Text: rotated text is drawn level. Text placed in axes or figure coordinates is not converted, and neither is an annotation whose point is in axes coordinates (data coordinates are).
- Axes:
set_aspectis not forced, so panels fill their space. The legend title and legend labels given by hand are not kept, since the key uses the artists' own labels. Inset axes are not drawn. - Twin axes (
twinx,twiny) are not drawn. The second axes is reported and left out, so its lines and bars are missing from the figure. - Figure level:
suptitle,fig.textand figure legends are not drawn. - Other artists: patches other than bars (circles, polygons), contours, meshes, quivers, violins and collections that are not a band between two curves are reported by name.
Colour bars are recognised and replaced by Inklet colour bars, with the label they were given as their title.