ADR 0140: The styles a capture can have¶

Status¶

Accepted

Context¶

ADR 0007 set the pillar: syntax is one hue, and the categories are told apart by weight, slant and opacity rather than by colour. It gave a table — keyword bold, type italic, string and number at ~78%, comment at ~45%.

Everything in that table was a literal in EditorViewport. ADR 0048 later gave types and strings a real hue each and made those two configurable, which left the arrangement backwards: the two deliberate exceptions to the pillar could be configured, and the pillar itself could not. Someone who found the comments too faint had no way to say so.

Decision¶

Four keys, in the same table every other setting lives in:

key	default
`syntax_keyword_bold`	`true`
`syntax_type_italic`	`false`
`syntax_comment_opacity`	`57`
`syntax_number_opacity`	`78`

Opacity is a percentage rather than an alpha byte, because a config file is read by people. The defaults reproduce the previous literals exactly: 57% of 255 is 145, which is the value ADR 0048 picked over ADR 0007's 45% because 45% measured 3.64:1, below WCAG AA. That reasoning now sits next to the default it produced.

syntax_type_italic defaults off, which is a divergence from ADR 0007's table rather than a restoration of it. Types carry a hue now, and slant on top of colour is louder than either alone. The key exists for people who want ADR 0007's original scheme back.

The metrics have to agree with the font¶

fontForCapture() says what a run is drawn with, and metricsForCapture() says what it is measured with — for run widths and for the caret. ADR 0124 is about what happens when those two disagree: the caret sits somewhere the text is not. So italic needed a cached QFontMetrics of its own, on the same branch condition as the font, and both functions now test the same config flag rather than the same hardcoded capture.

Consequences¶

No new tests, and that is a deliberate call rather than an omission. The schema side is already covered by tests that do not need to know these keys exist: one asserts every documented key's default is what a fresh config returns, another that the starter file mentions every key. Both covered these four the moment they were added.

What those cannot check is whether the editor draws differently, so that was verified by looking: types italic and keywords plain under one config, comments at full strength under another, and the default rendering unchanged against the same file. Glyph spacing is correct in each, which is the metrics-agreement check above.

This closes the last of the paper cuts ROADMAP had accumulated.