| 47 |
47 |
|
//!
|
| 48 |
48 |
|
//! # Findings
|
| 49 |
49 |
|
//!
|
| 50 |
|
- |
//! **1. The column description exists twice, in two vocabularies.** `build.rs`
|
| 51 |
|
- |
//! holds `TASK_COLUMNS` as `makeover_layout::Column` and emits the narrowing CSS
|
| 52 |
|
- |
//! and the JSON the frontend tests check themselves against; [`COLUMNS`] here
|
| 53 |
|
- |
//! holds the same seven columns as `quasi_router::screen::Column` because the
|
| 54 |
|
- |
//! runtime one carries an address and the build-time one cannot. The two agree
|
| 55 |
|
- |
//! today and nothing checks that they do. Not a vocabulary gap — the overlap is
|
| 56 |
|
- |
//! name, width and priority, which a build script could emit as a Rust const for
|
| 57 |
|
- |
//! this module to take the address half onto. Left alone because inventing that
|
| 58 |
|
- |
//! generator while porting the screen is two changes wearing one commit.
|
|
50 |
+ |
//! **1. The column description exists twice, in two vocabularies, until the
|
|
51 |
+ |
//! flip.** `build.rs` holds `TASK_COLUMNS` as `makeover_layout::Column` and
|
|
52 |
+ |
//! emits the narrowing CSS and the JSON the frontend tests check themselves
|
|
53 |
+ |
//! against; [`COLUMNS`] here holds the same seven as
|
|
54 |
+ |
//! `quasi_router::screen::Column`, because the runtime one carries an address
|
|
55 |
+ |
//! and the build-time one cannot. Not a vocabulary gap, and not a thing to fix:
|
|
56 |
+ |
//! `TASK_COLUMNS` serves the JS task list only and deletes with it, so the
|
|
57 |
+ |
//! duplication resolves by subtraction. Nothing checks that the two agree and
|
|
58 |
+ |
//! nothing should — a generator bridging them, or a test asserting parity,
|
|
59 |
+ |
//! would be permanent machinery in service of a frontend being removed. Add a
|
|
60 |
+ |
//! column in both by hand in the meantime.
|
| 59 |
61 |
|
//!
|
| 60 |
62 |
|
//! **2. A table cannot say there is more.** [`Node::List`] carries a
|
| 61 |
63 |
|
//! [`Rest`](quasi_router::screen::Rest) and `Node::Table` carries nothing of the
|
| 155 |
157 |
|
|
| 156 |
158 |
|
/// The table, left to right.
|
| 157 |
159 |
|
///
|
| 158 |
|
- |
/// The same seven columns as `build.rs`'s `TASK_COLUMNS`, in the same order,
|
| 159 |
|
- |
/// with the same widths and the same four marked sortable — see finding 1 above
|
| 160 |
|
- |
/// for why the description exists in two places and what would close it. The
|
| 161 |
|
- |
/// priority a column has when room runs out is the build script's business,
|
| 162 |
|
- |
/// because it is what generates the narrowing CSS, so only the width is
|
| 163 |
|
- |
/// restated here; a renderer with no stylesheet reads it off the same order.
|
|
160 |
+ |
/// Seven columns, four of them sortable. This is the description that survives
|
|
161 |
+ |
/// the flip; `build.rs`'s `TASK_COLUMNS` says the same thing for the JS table
|
|
162 |
+ |
/// and goes away with it (finding 1 above). The priority a column has when room
|
|
163 |
+ |
/// runs out is the build script's business, because it is what generates the
|
|
164 |
+ |
/// narrowing CSS, so only the width is stated here; a renderer with no
|
|
165 |
+ |
/// stylesheet reads it off the same order.
|
| 164 |
166 |
|
const COLUMNS: [(&str, Width, Option<TaskSortColumn>); 7] = [
|
| 165 |
167 |
|
(
|
| 166 |
168 |
|
"description",
|