Skip to content

Figure provenance and reproduction

The eleven checked-in SVGs are scientific teaching figures produced by tools/generate_guide_figures.py from real scalar API evaluations. They are not generated artwork. The generator first executes every assertion in the nine worked scenarios, then draws the models and annotations. The scenario code and numerical expectations are in examples/guide.py.

Reproduce the assets

From a checkout with FuzzyRoutines installed, use the pinned documentation toolchain on CPython 3.14:

python -m pip install -r docs/requirements-plots.txt
python tools/generate_guide_figures.py --output-directory docs/site/content/en/assets/figures
python tools/generate_guide_figures.py --output-directory docs/site/content/en/assets/figures --check

--check compares every expected SVG byte for byte and writes no files. The documentation CI runs it on Python 3.14 with this pinned toolchain. SVG IDs use a fixed hash salt, metadata omits the generation date, and labels use a fixed font family. Reproduction is scoped to that toolchain; byte identity across different plotting-library or font versions is not promised.

For editorial review, an explicit --preview-directory PATH additionally writes PNGs to the chosen directory. The default writes only SVGs below --output-directory. No command accesses the network after dependencies have been installed. A missing dependency, stale figure, failed scenario assertion, or write failure produces a nonzero exit status.

Matplotlib and its dependencies, including NumPy, belong only to the documentation toolchain. Installing or importing FuzzyRoutines does not require them. The default example script imports neither.

Verify scientific content independently

SVG byte comparison detects stale assets; it does not prove their mathematics. tests/test_guide_figures.py independently checks all 401 coordinates of each of the 36 plotted curves against elementary piecewise or exponential formulas. It also checks scatter coordinates, six sensor bars, thresholds, the four-node centroid calculation with exact rational arithmetic, and title/legend layout. The reference calculations do not call the library's membership functions. Documentation CI installs the plotting dependencies and runs these tests against the installed scalar API before comparing the SVG files.

Interpret the pictures correctly

  • Continuous model lines use 401 display samples; they are illustrations, not proofs of exact geometry or a quadrature policy used by Centroid.
  • Alpha-cut dots show the stated nine-point grid; outlined dots indicate the exhaustive five-point discrete universe's qualifying coordinates.
  • Scale diagnosis uses eleven coordinates including endpoints. The plot's smooth curves and the diagnostic observations are different kinds of evidence.
  • The normalization figure draws only discrete points, with no interpolating line or implied intermediate universe coordinates.
  • The centroid picture contrasts analytical moments with a separate user-side four-point trapezoidal calculation. Neither the displayed line density nor that coarse calculation configures the library's adaptive integration.
  • The operator picture fixes the second input grade at 0.6. Its horizontal axis varies the first grade; neither axis is a physical measurement.

Every image has a descriptive Markdown alternative and SVG title/description. Its accompanying page gives the numerical result in text, so colour perception or image availability is not required to understand the example. Model parameters and labels are illustrative rather than empirical calibration data. The Universal Fuzzy Scale figure specifically preserves the historical coefficients and compares the historical and modern evaluators.

The plots use Matplotlib's SVG backend and savefig metadata support.