v0.3.0

Added

• Plot kind qq: sorted values against normal quantiles, one series per array or by= group, with a dashed
  line through each one's quartiles. The quantiles come from Python's statistics, so no scipy is needed.
• Plot kind ridge: density curves stacked top to bottom, one per array or by= group, on one shared height
  scale; overlap= sets the tallest ridge's height in rows. mean and median mark each ridge.
• Plot kind strip: every value as a point, one column per array or by= group, jittered sideways so ties
  stay visible (the same way every run; jitter= sets the spread). Takes mean and ann like box.
• kx.grid(n, cols=3) makes a figure of n panels and returns them as a flat array of axes. Each panel gets
  the theme's figure size, empty cells are removed, and constrained layout leaves room for colorbars and outside
  legends. Other keywords go to plt.subplots.
• kx.save('fig.png') saves the current figure (or the one of a given ax or figure) and returns the path.
  A relative name goes to /kaggle/working on Kaggle and to the current folder elsewhere; missing folders are
  created. Without an extension the theme's format is used, so thesis saves a PDF.
• with kx.context(theme, palette=..., cmap=..., **rc): takes the same arguments as kx.use and puts the
  previous settings back when the block ends, also after an error.
• kx.parse(spec) returns a Spec with kind, theme and flags. Printed, it is the canonical spec string
  (kind, theme, then flags A–Z), and kx.plot takes a Spec wherever it takes a spec string. Mistyped tokens get
  the closest valid ones as a hint.
• kx.plot(df, 'x', 'y', 'z', spec) names DataFrame columns, and by= can name a column. A mistyped column
  gets the closest names as a hint.
• kx.plot(..., **kwargs) passes other keywords to the matplotlib call that draws the kind, over kx's defaults:
  kx.plot(x, y, 'line', linewidth=3). bins= sets the shared hist bins and width= (height= for
  barh) the space each category's bars fill.
• Flags for every kind
  • Layout: legout puts the legend outside on the right, keeping a kind's legend title. rot rotates the x tick
    labels 45°.
  • Axes: logxy (log on both), symlogy (symmetric log y, so negatives work), zero (the value axis includes 0)
    and eq (equal aspect).
  • Number format: pct (0.25 → 25%) and si (1500 → 1.5k) on the value axis, which is x for barh and y
    otherwise.
• Flags for some kinds
  • date for line, step, area, band and scatter: x is read as dates (text works too) and gets compact
    date ticks.
  • mk for line, step, area and band: a marker on every point.
  • sort for bar and barh: largest first, by y, or by the total with stack.
  • ann for bar and barh: the value on each bar, inside each segment with stack.
  • mean and median for hist, kde and ecdf: a dashed / dotted line at each series' mean / median.
    mean for box and violin: a diamond at each mean. ann writes the values.
  • nocb, sym and logc for scatter with a numeric z: no colorbar, colour limits symmetric around 0,
    log colour scale.
• Flags that ask for opposite things (pct and si, logy and logxy, zero and a log y, sym and logc)
  raise an error.

Changed

• The tight flag leaves a figure with constrained layout (such as kx.grid's) alone instead of failing
  when it has a colorbar.

Fixed

• leg and legout on a plot with nothing to label, such as scatter with a numeric z, no longer warn
  "No artists with labels found"; the plot just has no legend.

v0.2.0

Added

• Plot kinds
  • barh: horizontal bars, first category on top. y and z are grouped like bar.
  • kde: smooth density curves (Gaussian KDE in numpy, no scipy needed), one per array.
  • ecdf: empirical CDF, the share of values at or below each x, one step curve per array.
  • box: box plots side by side (median, quartiles, 1.5 IQR whiskers, outliers), one per array.
  • violin: violin plots side by side with the median and quartiles marked, one per array.
• Flags
  • stack now works with bar and barh: z is drawn on top of y, so the bar's end is the total.
  • stack, norm (density) and the new cum (cumulative) work with hist.
• kx.plot(..., by=groups) splits x into one curve, box or violin per group (for kde, ecdf, box, violin).
• cmocean colourmaps thermal, haline, deep and balance for kx.use(cmap=...).
• Crameri colourmaps (Scientific colour maps 8) batlow, lipari, hawaii, oslo, davos, vik, roma,
  bam, cork and cyclic romaO, vendored the same way.
• kx.use(cmap="<name>_r") reverses any colourmap in kx.cmaps().
• Palettes tollight (Paul Tol), set1 and paired (ColorBrewer) and tab20 (Tableau, 20 categories).

Changed

• hist puts every series on the same bins, so overlaid bars line up.

v0.1.0

Added

• PyPI package: pip install kx-plot, then import kx. The wheel ships the themes, palettes and demo CSVs.
• Palettes: Paul Tol (tolbright, tolvibrant, tolhc, tolmc), Petroff (petroff6, petroff8, petroff10),
  Tableau (tab10, tableaucb), ColorBrewer (set2, dark2) and ibm. Their licences are in
  kx/styles/palettes/LICENCES.md.

Changed

• kx.py is now the kx/ package, with the themes in kx/styles/ and the demo CSVs in kx/data/.
  kx.source() prints every file in the package.

Release v0.0.2

Added

Plot kinds

• step: like line, drawn as steps centred on each x.
• area: filled areas, overlapping and semi-transparent by default.
  • area-stack stacks z on top of y, so the top edge is the total.
  • area-norm draws a 100% stack with a percent y axis. With leg, the legend goes outside on the right.
• band: a y line with a shaded y ± z band in the same colour. z is required.
• scatter with a text, bool or pandas categorical z draws one colour per category with a legend titled after z. A numeric z still gets a colorbar.

Flags

• stack and norm. Both work with area only for now.

Themes

• science: one journal column (3.5 in), 8 pt serif, inward ticks on all four sides, minor ticks, Paul Tol bright colours, saves at 600 dpi.
• talk: slides (16:9), 18 pt sans, thick lines, Okabe-Ito colours.
• poster: 12 × 9 in, 26 pt sans, very thick lines, saves at 300 dpi.
• thesis: A4 text width (160 mm), Computer Modern 10/9 pt without needing LaTeX, saves as PDF when the file name has no extension.

Palettes

• Named palettes live in styles/palettes/<name>.txt, one hex colour per line. Any file added there is found automatically.
• okabe: the Okabe-Ito palette, 8 colours that stay distinguishable for every common type of colour blindness.
• kx.palettes() lists the palettes and kx.palette(name) returns a palette's colours.
• kx.use(theme, palette="okabe") accepts a palette name as well as a list of colours.

Colormaps

• kx.use(theme, cmap=...) sets the colormap for continuous colour.
• The supported matplotlib colormaps are viridis, magma, plasma, inferno, cividis, turbo, twilight, coolwarm, berlin, managua and vanimo.
• kx.cmaps() lists only the ones the installed matplotlib has. berlin, managua and vanimo need matplotlib 3.10 or newer.

Release v0.0.1

Initial Alpha release

Added

• Initial graphic capabilities
• demo preview showing available options