v0.1.0

anyplotlib 0.1.0 — first public release

anyplotlib is a plotting library built on Python, JavaScript, and
anywidget for fast, interactive, in-browser plotting
— in Jupyter, in the docs, and (new in this release) embedded in standalone apps via an electron bridge.

This first release ships the full core: Figure, Axes, GridSpec /
subplots, the Plot1D / Plot2D / PlotMesh / Plot3D / PlotBar /
PlotXY plot types, a marker system, interactive overlay widgets, and a
two-tier callback registry.

Install

pip install anyplotlib

Requires Python ≥ 3.10. Optional extras: anyplotlib[jupyter], anyplotlib[docs].

Highlights

New plotting surfaces

• PlotXY / Axes.axes2d — a blank data-coordinate 2-D axis (matplotlib
  transData + PathCollection model). Draw scatter / plot / fill /
  text in data coords with per-point face/edge colours and aspect="equal"
  — the surface for stereographic / IPF / pole-figure plots (e.g. an orix
  backend).
• Fast density heatmaps — PlotXY.pcolormesh rasterizes regular meshes into
  a single blit whose cost is independent of cell count (a 256×256 heatmap draws
  as fast as 32×32). The underlying add_raster primitive draws an RGBA image
  between data-coord corners, with an optional clip_path polygon and a
  smooth=True bilinear option.
• InsetAxes — floating overlay sub-plots above the main grid (add_inset),
  supporting every plot type plus interactive minimise / maximise / restore.
• True-colour imshow — (H, W, 3|4) arrays now render as RGB(A) instead of
  silently dropping channels.

3-D and GPU

• Axes.voxels() — volumes as shaded translucent cubes, with a draggable
  PlaneWidget slice selector that fires pointer_move / pointer_up
  callbacks; voxels on the slice plane glow.
• WebGPU acceleration — scatter3d and voxels render on the GPU via WebGPU
  when available (gpu="auto"), with a transparent, identical-looking Canvas2D
  fallback when no GPU is present. Query the active path with plot.gpu_active.
• Richer 3-D scatter — per-point colors=, a bounds= override, single-point
  set_highlight(), a shaded reference set_sphere(), and a proper turntable
  camera (matplotlib azim / elev semantics).

Interactivity

• Touch support — one-finger pan / orbit / drag, two-finger pinch zoom, and
  double-tap, on iPad / iPhone / trackpads, with no API change.
• Data-coordinate picking — double_click on 1-D / PlotXY panels now
  reports ydata alongside xdata.
• Typography — fontsize on labels/titles/colorbars, set_tick_label_size(),
  and a mini-TeX subset inside $...$ (super/subscripts, Greek, \times, \AA,
  \degree) rendered natively on the canvas; text is no longer clipped.

Performance

• Geometry sync channel — large, slow-changing buffers are deduped by content
  hash, so view-only updates (highlight, view, zoom, widget drags) no longer
  re-transmit geometry.
• Figure.batch() — coalesces panel pushes so linked-view handlers send one
  update per changed panel instead of one per mutation.
• Voxels ~2–3× faster via sprite caching and cached projection/depth-sort.
• ~50× faster Pyodide dispatch — interactions call a pre-compiled dispatcher
  instead of recompiling Python source every frame, so the interactive docs keep
  up with gestures.

Embedding outside Jupyter

• fig.save_html() / fig.to_html() export a self-contained interactive page.
• figure_esm.js exports a mount(el, state, opts) entry point for direct JS
  embedding (Electron, MDI windows, plain web pages) with onEvent callbacks,
  live setPanelState, resize, and dispose.
• The new anyplotlib.embed module provides figure_state(), esm_path(), and a
  transport-agnostic FigureBridge for two-way Python ↔ JS sync over any pipe.

Documentation tooling

• The anyplotlib.sphinx_anywidget extension renders live, Pyodide-powered
  figures in the docs (.. anywidget-figure:: directive, automatic wheel
  building, Sphinx Gallery integration).

Bug fixes

• 3-D plane-widget drags no longer snap back (view-only pushes stop clobbering an
  in-progress drag).
• GPU 3-D panels self-heal when a WebGPU device is lost mid-draw (no more vanishing
  voxels/axes until a window resize) — seen on Safari's experimental WebGPU.
• Large voxel volumes no longer render "empty" in WebGPU browsers (e.g. PyCharm's
  JCEF) — the plot canvas background is now transparent while the GPU path is active.
• The 3-D voxel highlight no longer floats onto random voxels in large volumes.
• Interactive (⚡) docs figures are much smoother under Pyodide (the ~50× dispatch
  speedup above).

Full changelog

See CHANGELOG.rst for the complete, detailed list.

v0.1.0b1

anyplotlib 0.1.0 — first public release

anyplotlib is a plotting library built on Python, JavaScript, and
anywidget for fast, interactive, in-browser plotting
— in Jupyter, in the docs, and (new in this release) embedded in standalone apps.

This first release ships the full core: Figure, Axes, GridSpec /
subplots, the Plot1D / Plot2D / PlotMesh / Plot3D / PlotBar /
PlotXY plot types, a marker system, interactive overlay widgets, and a
two-tier callback registry.

Install

pip install anyplotlib

Requires Python ≥ 3.10. Optional extras: anyplotlib[jupyter], anyplotlib[docs].

Highlights

New plotting surfaces

• PlotXY / Axes.axes2d — a blank data-coordinate 2-D axis (matplotlib
  transData + PathCollection model). Draw scatter / plot / fill /
  text in data coords with per-point face/edge colours and aspect="equal"
  — the surface for stereographic / IPF / pole-figure plots (e.g. an orix
  backend).
• Fast density heatmaps — PlotXY.pcolormesh rasterizes regular meshes into
  a single blit whose cost is independent of cell count (a 256×256 heatmap draws
  as fast as 32×32). The underlying add_raster primitive draws an RGBA image
  between data-coord corners, with an optional clip_path polygon and a
  smooth=True bilinear option.
• InsetAxes — floating overlay sub-plots above the main grid (add_inset),
  supporting every plot type plus interactive minimise / maximise / restore.
• True-colour imshow — (H, W, 3|4) arrays now render as RGB(A) instead of
  silently dropping channels.

3-D and GPU

• Axes.voxels() — volumes as shaded translucent cubes, with a draggable
  PlaneWidget slice selector that fires pointer_move / pointer_up
  callbacks; voxels on the slice plane glow.
• WebGPU acceleration — scatter3d and voxels render on the GPU via WebGPU
  when available (gpu="auto"), with a transparent, identical-looking Canvas2D
  fallback when no GPU is present. Query the active path with plot.gpu_active.
• Richer 3-D scatter — per-point colors=, a bounds= override, single-point
  set_highlight(), a shaded reference set_sphere(), and a proper turntable
  camera (matplotlib azim / elev semantics).

Interactivity

• Touch support — one-finger pan / orbit / drag, two-finger pinch zoom, and
  double-tap, on iPad / iPhone / trackpads, with no API change.
• Data-coordinate picking — double_click on 1-D / PlotXY panels now
  reports ydata alongside xdata.
• Typography — fontsize on labels/titles/colorbars, set_tick_label_size(),
  and a mini-TeX subset inside $...$ (super/subscripts, Greek, \times, \AA,
  \degree) rendered natively on the canvas; text is no longer clipped.

Performance

• Geometry sync channel — large, slow-changing buffers are deduped by content
  hash, so view-only updates (highlight, view, zoom, widget drags) no longer
  re-transmit geometry.
• Figure.batch() — coalesces panel pushes so linked-view handlers send one
  update per changed panel instead of one per mutation.
• Voxels ~2–3× faster via sprite caching and cached projection/depth-sort.
• ~50× faster Pyodide dispatch — interactions call a pre-compiled dispatcher
  instead of recompiling Python source every frame, so the interactive docs keep
  up with gestures.

Embedding outside Jupyter

• fig.save_html() / fig.to_html() export a self-contained interactive page.
• figure_esm.js exports a mount(el, state, opts) entry point for direct JS
  embedding (Electron, MDI windows, plain web pages) with onEvent callbacks,
  live setPanelState, resize, and dispose.
• The new anyplotlib.embed module provides figure_state(), esm_path(), and a
  transport-agnostic FigureBridge for two-way Python ↔ JS sync over any pipe.

Documentation tooling

• The anyplotlib.sphinx_anywidget extension renders live, Pyodide-powered
  figures in the docs (.. anywidget-figure:: directive, automatic wheel
  building, Sphinx Gallery integration).

Bug fixes

• 3-D plane-widget drags no longer snap back (view-only pushes stop clobbering an
  in-progress drag).
• GPU 3-D panels self-heal when a WebGPU device is lost mid-draw (no more vanishing
  voxels/axes until a window resize) — seen on Safari's experimental WebGPU.
• Large voxel volumes no longer render "empty" in WebGPU browsers (e.g. PyCharm's
  JCEF) — the plot canvas background is now transparent while the GPU path is active.
• The 3-D voxel highlight no longer floats onto random voxels in large volumes.
• Interactive (⚡) docs figures are much smoother under Pyodide (the ~50× dispatch
  speedup above).

Full changelog

See CHANGELOG.rst for the complete, detailed list.