max / audiofiles

Measure the file list's context menus against the vocabulary The measurement `0341c7b5` asked for, recorded in `quasi/files.rs`'s header. Three menus, 545 lines, and they split three ways rather than failing as one. `Row::menu` already is a context menu -- its header argues, correctly, that a menu belongs in the description because one description has to become a right-click, a long-press and a key-driven list. The file list is a `Node::Table`, and `Cells` has no counterpart. That asymmetry has been fixed once before in this exact place: `Cells::selected` records that it was deliberately absent "until 2026-08-15", when a consumer arrived. The 14-entry row menu is that sentence about the next field along. The other two menus have no container at all. A menu over the selection and a menu over the surface are not per-row, and `Outcome::Over` makes both app-modal -- which is what the toolbar port already recorded of the save-as-collection popover and what this pass took for the Import menu. Filed in quasicoherent as two tasks, not one: `Cells::menu` is a member with a precedent and a measured consumer, and the anchored menu is a design question with three candidate shapes and four consumers. Neither is implemented here. What is missing is never the contents -- five of the sixteen entries across the two unanchored menus are described at addresses of their own already. It is the gesture and the anchor, both times.

Author: Max Johnson <me@maxj.phd> · 2026-08-17 21:08 UTC

Signed with PGP, not checked

Commit: 012027b5d2a1a897ac340759debdc3fe2b63e3e3

Parent: e794b0f

1 file changed, +44 insertions, -3 deletions

M crates/audiofiles-browser/src/quasi/files.rs +44 -3

			@@ -35,9 +35,50 @@
35	35		//! drag-cooldown state machine so an OS drag that ends outside the window does
36	36		//! not leave egui's pointer state stale. That is a host input problem and
37	37		//! nothing about it is a fact about a sample.
38		-	//! - The waveform, the inline rename, the context menu. Each is its own
39		-	//! screen or its own affordance; folding them in here would make the port
40		-	//! about size rather than about shape.
	38	+	//! - The waveform and the inline rename. Each is its own affordance;
	39	+	//! folding them in here would make the port about size rather than about
	40	+	//! shape.
	41	+	//!
	42	+	//! # The context menus, measured 2026-08-17, and they split three ways
	43	+	//!
	44	+	//! `ui/file_list_menus.rs` is 545 lines of three menus — one for the row under
	45	+	//! the cursor (14 entries), one for the selection (11), one for empty space (5)
	46	+	//! — and the question asked of them was whether a context menu can be described
	47	+	//! at all. It can, partly, and where it cannot is two members rather than one
	48	+	//! vague gap.
	49	+	//!
	50	+	//! `Row::menu` is exactly the first one, on the wrong container. Its own
	51	+	//! header says what it is for — "what it offers, reached by right-click on a
	52	+	//! pointer host, long-press on a touch one, and a key in a terminal" — and why
	53	+	//! that belongs to the description rather than to a renderer: one description
	54	+	//! has to become a context menu, an action sheet and a key-driven menu, and no
	55	+	//! single renderer can be where that is said. It is right, and the file list is
	56	+	//! a [`Node::Table`], whose [`Cells`] has no `menu`.
	57	+	//!
	58	+	//! That asymmetry has been corrected once before, in this exact place.
	59	+	//! `Cells::selected` says it was deliberately absent "on the grounds that no
	60	+	//! table asked for one and a member added because its sibling has it is a member
	61	+	//! with no consumer to tell us what it should mean", and that it "was correct
	62	+	//! until 2026-08-15". This is the same sentence about the next field along, and
	63	+	//! the consumer that ends it is here: 14 entries, branching on node type, with
	64	+	//! a submenu. Filed as `quasi:vocabulary:table-row-menu`.
	65	+	//!
	66	+	//! The other two have no container at all. A menu over the selection and a
	67	+	//! menu over the surface are not per-row, and nothing in the vocabulary holds
	68	+	//! acts back until a host asks for them except `Row::menu`. Described as
	69	+	//! [`Outcome::Over`](quasi_router::Outcome::Over) they become app-modal
	70	+	//! overlays, which is what [`toolbar`](super::toolbar) already recorded of the
	71	+	//! save-as-collection popover — "near enough and not exact" — and what
	72	+	//! [`importing`](super::importing) took for the Import menu as its second
	73	+	//! consumer. These are the third and fourth, and they are the ones that make the
	74	+	//! shape clear: an anchored menu is not a modal, and its subject is whatever it
	75	+	//! opened over. Filed as `quasi:vocabulary:anchored-menu`.
	76	+	//!
	77	+	//! Two of the eleven selection entries and three of the five background entries
	78	+	//! are described already, at addresses of their own —
	79	+	//! [`bulk`](super::bulk)'s three modals, [`naming`](super::naming)'s New Folder,
	80	+	//! [`importing`](super::importing)'s two doors. So what is missing is never the
	81	+	//! contents. It is the gesture and the anchor, both times.
41	82		//! - Virtual scrolling. Recorded in the findings note as renderer policy
42	83		//! from the start: windowing rows the app already holds is a performance
43	84		//! technique, not a described fact.