018-hirst-painting/docs/milestones/mil-003-visible-colours.md

MIL-003: Visible Colours

Metadata

Key	Value
ID	MIL-003
CrossReference	BC-001
Language	en
Domain	it

Version History

Date	Status	Author	Reviewer	Change	Commit
2026-10-08	Accepted	Jens Tirsvad Nielsen	S01	Initial version	188b2e9

---

Purpose

This gate decides whether the palette keeps only colours that stand out from the white background, so that every dot of the painting is clearly visible (objective O6 of the Business Case). The palette must contain no colour whose contrast with white is below 2.0, and the painting must still meet every earlier criterion.

Deliverable

A contrast measure and a faint-colour filter in the program's source file under src/, applied when the palette is extracted, together with their tests and one real run that shows the painting drawn from the filtered palette.

Go / No-Go Criteria

#	Criterion (objectively checkable)	Go	No-Go
1	The contrast measure gives 1.0 for white and 21.0 for black, and the filter keeps a colour only when its contrast with white is 2.0 or more (SC8)	The tests of the measure and the filter pass	Any of those tests fails
2	The palette of the reference image has no colour with a contrast below 2.0, still has at least 2 colours, and still has no white shade (SC8, SC3)	The printed palette with the contrast of every colour meets all three	A colour below 2.0, fewer than 2 colours, or a white shade
3	The earlier behaviour is unchanged: all tests pass, and ruff and mypy --strict are clean	All tests pass and both tools report no problem	A failing test or a reported problem
4	A real run still draws the painting in 30 seconds or less, with 100 dots, no trail and no turtle, and the window stays open until a click (SC1, SC5, SC6)	A timed run and a captured picture show all of it, and every dot is clearly visible	Any of them fails
5	The phase code has a review record against qc-programming-python with verdict Go (SC7)	RC-* verdict Go	Go-with-conditions or No-Go, or no record

Dependencies

Depends on	Reason
MIL-001	The filter is applied inside the palette function that phase delivered, beside its white-shade filter
MIL-002	The painting is drawn again with the filtered palette, and its tests must keep passing

Traceability

Business Case objective / KPI / user story	Reference
O6: every dot clearly visible, contrast with white of 2.0 or more	BC-001; criteria 1, 2 and 4, SC8
O2: palette without white shades	BC-001; criterion 2, SC3 (still holds)
O1, O4: the painting and its clean presentation	BC-001; criteria 3 and 4, SC1, SC5 and SC6
O5: delivery through the framework, code reviewed against qc-programming-python	BC-001; criterion 5, SC7

No use case or user story applies: every task is a technical step, so the Tasks column "Needs its own Use Case/User Story?" says No throughout.

Ownership

Role	Stakeholder ID (SA)
Owner	S01
Approving reviewer	S01

Target Date

2026-10-10 — inside the one-week limit of BC-001 (2026-10-07 to 2026-10-14), and leaves four days of slack.

Tasks

#	Task	Summary	Needs its own Use Case/User Story?	Reference
1	Add the contrast measure	Add a function that returns the contrast ratio of a colour against white, as the Web Content Accessibility Guidelines (WCAG) 2.x define it from relative luminance: 1.0 for white and 21.0 for black. It is the measure that objective O6 and criterion SC8 of the Business Case rest on.	No
2	Add the faint-colour filter	Add a named constant for the limit, 2.0, and a function that drops every colour whose contrast with white is below it, keeping the order of the rest. It sits beside the white-shade filter, which stays as it is, so SC3 keeps its own check.	No
3	Add tests for the measure and the filter	Add tests under tests/ with known values (white 1.0, black 21.0, a mid grey) and with the boundary: a colour just below 2.0 is dropped and one just above is kept. They pin the limit before the real image is involved, as the white-shade tests do for 240.	No
4	Apply the filter when the palette is extracted	Make extract_palette apply the faint-colour filter after the white-shade filter, update its tests, and add one that checks the palette of the reference image: no colour below 2.0 and at least 2 colours. Print the palette with each contrast; the measurement on 2026-10-08 predicts 22 colours.	No
5	Check the painting again with a real run	Run main on a real window, time it and capture the picture: 100 dots in 10 rows of 10, no trail, no turtle cursor, the window open until a click, and every dot clearly visible on the white background. This rechecks SC1 to SC8 together.	No
6	Review the phase code against qc-programming-python	Check the code of this phase against framework/qc/qc-programming-python.md and record the result as an RC-*; a Go verdict is Go criterion 5 (SC7).	No

---