Changelog

All notable changes to this project will be documented in this file.

[0.5.1] - 2026-05-19

render/3 opt :responsive — appends dynamic to set terminal svg …, so the captured SVG has no fixed width/height attrs and scales to its container via viewBox alone. Fixes mobile/responsive layouts where the previous baked pixel dims overflowed cards
render/3 opt :css_theme — post-processes captured SVG to swap baked colors for currentColor (text, axes, strokes) and transparent (page background), letting host CSS color: … drive chart chrome. Enables dark-mode and themed-page embedding without downstream String.replace/3 hacks

Both opts default to false (no behavior change) and are no-ops for non-SVG terminals.

Widen :decimal version constraint — {:decimal, "~> 2.0 or ~> 3.0", optional: true}. The previous ~> 2.0 constraint silently downgraded Decimal 3.x to 2.x in apps that pulled in gnuplot_ex alongside Ecto 3.13+, Postgrex 0.22+, or Phoenix LiveView 1.1+ (all on Decimal 3.x). Mirrors the dual constraint used by ecto/postgrex/jason

Candlestick (OHLC) series — first-class GnuplotEx.candlestick/1,2,3 builder for financial charts. Supports bull/bear body coloring via :up_color/:down_color (conditional lc rgb variable) and optional wick caps via :whiskerbars
MultiPlot composition — new %GnuplotEx.MultiPlot{} struct plus GnuplotEx.multiplot/0,1, add_panel/2, layout/2, shared_x/2 builders for multi-panel layouts (e.g. price + volume + indicator stacks). Shared x-axis mode aligns time-series panels
MultiPlot figure-level styling — :background (canvas color appended to set terminal) and :commands (figure-level raw gnuplot commands emitted before set multiplot) for consistent styling across panels
MultiPlot :row_heights — unequal vertical panel heights (e.g. [0.6, 0.2, 0.2] for price-dominant financial stacks). Emits bare set multiplot with per-panel set size/set origin instead of forced-equal set multiplot layout R,C
LiveView accepts MultiPlot — <.live_gnuplot plot={@multiplot} /> renders composed layouts through the existing component, cache, and async paths

Series.using honored for all data sources — previously dropped for inline "-" and datablock $name sources, breaking time-axis line/impulses/steps plots that gnuplot requires to declare their using clause
Multiplot inline-streaming — gnuplot 6 refuses plot '-' inside set multiplot. Panel series data is now hoisted into named datablocks ($mp_pN_sM << EOD … EOD) before emission
Float-tuple ranges — {94.5, 105.5} now renders as [94.5:105.5] instead of 94.5,105.5 (invalid gnuplot syntax). Elixir Range doesn't support floats, so float tuples are emitted as gnuplot range literals directly
Duplicate SVG ids in LiveView — multiple <.live_gnuplot> components on one page no longer collide on gnuplot's hardcoded ids (gnuplot_canvas, gpPt0–gpPt14, textbox, greybox, etc.). The component rewrites every id=/xlink:href=#/href=#/url(#…) to a per-component prefix, handling both single- and double-quoted forms

Real-gnuplot smoke tests (@tag :gnuplot) added for multiplot, candlestick, and the using/range fixes so future regressions fail on render rather than slipping past dry-mode checks

render/2 now returns actual SVG/PNG output - Previously returned the gnuplot script string instead of rendered output, breaking LiveView components
datablock/3 works with command/2 plots - Named datablocks are now defined before custom commands, allowing references like $stats in custom plot commands
Pie/donut charts render correctly - Fixed inner_radius calculation so pie charts display as full circles, not donuts

Pie/donut slice labels - Labels are now rendered at the center of each slice using set label commands

Initial public release
High-level API for common chart types (scatter, line, bar, histogram, pie, donut, spider, parallel coordinates)
Low-level API for full gnuplot control
LiveView components for real-time charts
Explorer DataFrame integration
Nx tensor support
ML visualizations (confusion matrix, ROC curves, loss curves, embeddings)
SVG, PNG, and ASCII output formats
Async/parallel rendering
Binary data optimization for large datasets