(d7e493e) Draw the command's own usage line while the prompt is open, and bump to 0.33.7

The reader knows the grammar approximately: that mv takes a status, not
whether it is a positional or --status, and that set has a flag per section,
not whether the criteria one is --acceptance or --criteria. In a shell fish
answers that as the line is typed. On the board there was nothing, and
getting it wrong cost a rejected line and a guess.

usage_of walks the typed verb against the clap definition and renders its
usage, so a flag added to a command is described here the day it is added.
A table of synopses kept beside it would be a second grammar, and it would
go stale silently: nothing but a reader typing that verb would ever see the
two disagree. The Usage: label is dropped, being the name of a block in a
help page of several where this is one line beside the line it describes.

It takes the keys' slot. esc and enter are two keys a reader learns once,
against a synopsis that is different for every verb; a refusal still wins
over both while it is up, being what was just said about the line that is
still on the row. refusal is the half that says what is wrong where this
says what would have been right.

A verb that resolves to nothing draws nothing and complains about nothing,
since the line is half typed most of the time the row is looked at. A
group's child is walked when the second word names one, so campaign set
describes set; while that word is half typed the group's own usage stands.

Grey, and cut with an ellipsis rather than wrapped: the footer is one row
and a second would reflow the columns every time a verb resolved. There is
always at least the keys' width to cut into, because that width comes off
the row before the line is windowed, so the usage never takes a cell from
what is being typed.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Ansho Rei
2026-09-15 16:51:08 +02:00
co-authored by Claude Opus 5
parent 634a1f44c4
commit 298e00eb33
4 changed files with 250 additions and 9 deletions
+244 -7
View File
@@ -980,12 +980,57 @@ fn command_name(matches: &clap::ArgMatches) -> Option<String> {
})
}
/// The usage line clap renders for the command the typed line names, or nothing while the
/// verb names none.
///
/// **Clap's own rendering, off the definition the parse uses**, so a flag added to a
/// command is described here the day it is added. A table of synopses kept beside it would
/// be a second grammar, and it would go stale silently: nothing but a reader typing that
/// verb would ever see the two disagree.
///
/// A verb that resolves to nothing answers `None`. The line is half typed most of the time
/// it is looked at, and a complaint about `com` is noise on a row the reader is typing on;
/// `refusal` is where a line that was actually entered is answered.
///
/// A group's child is walked when the second word names one, so `campaign set` describes
/// `set` and not `campaign`. While that second word is still half typed the group's own
/// usage stands, which is the honest answer: `campaign <COMMAND>` is what has been typed
/// so far.
///
/// The `Usage: ` label is dropped and the rest is kept verbatim. The label names the block
/// in a help page of several; this row is one line beside the line it describes, and the
/// cells are worth more than the word.
fn usage_of(args: &[String]) -> Option<String> {
// Colour off for the same reason the parse turns it off: what this row draws is a
// ratatui span, and an escape inside one is printed rather than obeyed.
let cli = <Cli as clap::CommandFactory>::command().color(clap::ColorChoice::Never);
let command = cli.find_subcommand(args.first()?)?;
// The child by exact name, which is what clap will resolve too. A half-typed one
// matches nothing and leaves the group's usage on the row.
let command = match args.get(1).and_then(|word| command.find_subcommand(word)) {
Some(child) => child,
None => command,
};
let usage = command.clone().render_usage().to_string();
Some(
usage
.strip_prefix("Usage: ")
.unwrap_or(&usage)
.trim()
.to_string(),
)
}
/// The one line the row carries for a line clap would not take.
///
/// Clap's own message, because the grammar is clap's: a complaint written here would
/// describe a rule this file does not own. What is dropped is the usage block and the
/// `--help` line after it, which are a paragraph — the synopsis is d7e493e's, and until
/// then the row says what is wrong rather than what would have been right.
/// `--help` line after it, which are a paragraph: this row says what is wrong, and
/// `usage_of` is what says what would have been right, on the same row for as long as the
/// verb resolves. A refusal wins the slot while it is up, being what was just said about
/// the line that is still there.
///
/// `--help` is the exception. It parses as an error of the kind that means *print this*,
/// and what it would print is a page: caught here, or clap's help lands on the alternate
@@ -2520,6 +2565,19 @@ impl Board {
.map(str::to_string)
}
/// The usage line of the command the prompt's line names, for the row to draw in place
/// of the keys.
///
/// Derived on every draw and held nowhere, the way `completion` is: what is on the row
/// is a function of what is in the line, so the two cannot come to disagree and there
/// is no state to clear when the line is edited. The walk is over the clap definition
/// alone and reads nothing off the board, since a keypress must not go to disk.
fn usage(&self) -> Option<String> {
let prompt = self.prompt.as_ref()?;
let split = split_line(&prompt.line, prompt.cursor);
usage_of(&split.args)
}
/// `tab`: take what every candidate agrees on, and then step through them one press at
/// a time.
///
@@ -4766,7 +4824,13 @@ fn draw_footer(frame: &mut Frame, board: &Board, area: Rect) {
// and is asked first: an input cannot be covered, and it is the newest of the three
// things that can hold the row. Only one of the two prompts can be open at a time.
if let Some(prompt) = &board.prompt {
draw_prompt(frame, prompt, board.suggestion().as_deref(), area);
draw_prompt(
frame,
prompt,
board.suggestion().as_deref(),
board.usage().as_deref(),
area,
);
return;
}
@@ -4979,7 +5043,26 @@ const PROMPT_OPENING: &str = " :";
/// of the field rather than a field of its own, so a long command loses the offer instead
/// of pushing the keys off the row. The terminal cursor sits on its first cell, which is
/// where the cursor already was: what is under a cursor is still what will be typed over.
fn draw_prompt(frame: &mut Frame, prompt: &Prompt, suggestion: Option<&str>, area: Rect) {
///
/// **The usage line takes the keys' slot as soon as the verb resolves.** The reader knows
/// the grammar approximately — that `mv` takes a status, not whether it is a positional or
/// a flag — and in a shell fish answers that as the line is typed. `esc` and `enter` are
/// what the row gives up for it: they are two keys a reader learns once, and the synopsis
/// is the thing that is different for every verb. A refusal still wins over both, being
/// what was just said about the line that is still on the row.
///
/// It is grey, the chrome colour: it is not in the line, nothing here is a value, and the
/// reader has to tell it from what they typed at a glance. Cut to what the row leaves with
/// an ellipsis, never wrapped: the footer is one line, and a second row would reflow the
/// columns every time a verb resolved. There is always at least the keys' width for it,
/// since that width was taken off the row before the line was windowed.
fn draw_prompt(
frame: &mut Frame,
prompt: &Prompt,
suggestion: Option<&str>,
usage: Option<&str>,
area: Rect,
) {
let keys = key_hints(&[("↵", "run"), ("esc", "cancel")]);
let spent: usize = PROMPT_OPENING.chars().count()
+ KEY_GAP.chars().count()
@@ -4994,6 +5077,7 @@ fn draw_prompt(frame: &mut Frame, prompt: &Prompt, suggestion: Option<&str>, are
let (visible, at) = prompt_window(&prompt.line, prompt.cursor, room);
let mut spans = vec![Span::raw(format!("{PROMPT_OPENING}{visible}"))];
let mut drawn = PROMPT_OPENING.chars().count() + visible.chars().count();
// Only ever drawn where the cursor is at the end of the line, which is what `visible`
// then ends at, so this follows the cursor rather than landing in front of the tail of
@@ -5002,16 +5086,30 @@ fn draw_prompt(frame: &mut Frame, prompt: &Prompt, suggestion: Option<&str>, are
let left = room.saturating_sub(visible.chars().count());
let shown: String = rest.chars().take(left).collect();
if !shown.is_empty() {
drawn += shown.chars().count();
spans.push(Span::styled(shown, Style::default().fg(Color::Grey.into())));
}
}
match &prompt.refused {
Some(message) => spans.push(Span::styled(
match (&prompt.refused, usage) {
(Some(message), _) => spans.push(Span::styled(
format!("{KEY_GAP}{message}"),
Style::default().fg(Color::BrightRed.into()),
)),
None => {
(None, Some(usage)) => {
// What the row leaves after the line and the gap. Nothing at all on a terminal
// too narrow to hold a cell of it, the bargain the suggestion already takes.
let left = (area.width as usize).saturating_sub(drawn + KEY_GAP.chars().count());
if left > 0 {
let shown = match usage.chars().count() <= left {
true => usage.to_string(),
false => with_ellipsis(usage, left),
};
spans.push(Span::raw(KEY_GAP));
spans.push(Span::styled(shown, Style::default().fg(Color::Grey.into())));
}
}
(None, None) => {
spans.push(Span::raw(KEY_GAP));
spans.extend(keys);
}
@@ -8199,6 +8297,132 @@ approach";
assert_eq!(at.y, SCREEN.height - 1, "the cursor left the prompt row");
}
/// The grammar, where the key list was, as soon as the verb resolves: the reader knows
/// `mv` takes a status and not whether it is a positional or a flag, which is what a
/// shell answers as the line is typed and the board answered with nothing.
#[test]
fn the_row_shows_the_usage_of_the_command_being_typed() {
let dir = searchable();
let mut board = one_board(&dir);
board.open_prompt();
// Nothing typed yet, so there is no command to describe and the keys have the slot.
let row = footer(&board, SCREEN.width, SCREEN.height);
assert!(row.contains("esc cancel"), "{row}");
type_line(&mut board, "mv ");
let row = footer(&board, SCREEN.width, SCREEN.height);
assert!(
row.contains("mv [OPTIONS] --status <STATUS> <TICKET>"),
"the usage is not on the row: {row}"
);
assert!(
!row.contains("esc cancel"),
"the keys and the usage both took the slot: {row}"
);
// And it is the grammar of the verb in the line, not of the one typed before it.
type_line(&mut board, "1111111");
board.prompt_key(ctrl('w'), SCREEN);
for _ in 0.."mv ".len() {
board.prompt_key(typed(KeyCode::Backspace), SCREEN);
}
type_line(&mut board, "comment ");
let row = footer(&board, SCREEN.width, SCREEN.height);
assert!(row.contains("comment"), "{row}");
assert!(!row.contains("--status"), "the previous verb's usage stayed: {row}");
// Closing the prompt gives the board's own key list back.
board.prompt_key(typed(KeyCode::Esc), SCREEN);
let row = footer(&board, SCREEN.width, SCREEN.height);
assert!(row.contains("quit"), "the key list did not come back: {row}");
assert!(!row.contains("--status"), "the usage outlived the prompt: {row}");
}
/// The resolve step on its own: what a verb names, what a group's child names, and what
/// a word that names nothing yet does.
///
/// The values are clap's own rendering off the definition the parse uses, which is why
/// this asserts on flags rather than on a synopsis written out here: a table of those
/// kept beside the command would go stale the first time a flag was added, and silently,
/// since nothing but a reader typing that verb would ever see the two disagree.
#[test]
fn a_usage_is_clap_s_own_and_a_nested_command_describes_itself() {
let usage = |line: &str| {
let words: Vec<String> = line.split_whitespace().map(str::to_string).collect();
usage_of(&words)
};
let mv = usage("mv").expect("mv names a command");
assert!(mv.starts_with("mv "), "the label survived: {mv}");
assert!(mv.contains("--status <STATUS>"), "{mv}");
assert!(mv.contains("<TICKET>"), "{mv}");
// The question the ticket is about: which flag the criteria section is under.
assert!(usage("set").is_some_and(|set| set.contains("--acceptance")));
// A group's child describes itself rather than its parent, once both words are
// there. `campaign set` takes a campaign; `campaign` alone takes a command.
let group = usage("campaign").expect("campaign names a command");
assert!(group.contains("<COMMAND>"), "{group}");
let child = usage("campaign set").expect("campaign set names a command");
assert!(child.contains("<CAMPAIGN>"), "{child}");
assert!(!child.contains("<COMMAND>"), "the parent's usage was drawn: {child}");
// A second word that names no child leaves the group's own usage up, which is what
// has been typed so far.
assert_eq!(usage("campaign se"), Some(group.clone()));
// And a verb that names nothing has nothing to say about it.
assert_eq!(usage("com"), None);
assert_eq!(usage(""), None);
assert_eq!(usage_of(&[]), None);
}
/// A line half typed is the state the row is looked at in, so a verb that resolves to
/// nothing draws nothing and complains about nothing.
#[test]
fn a_half_typed_verb_shows_no_usage_and_no_complaint() {
let dir = searchable();
let mut board = one_board(&dir);
board.open_prompt();
type_line(&mut board, "com");
let row = footer(&board, SCREEN.width, SCREEN.height);
assert!(row.contains("esc cancel"), "the keys lost the slot: {row}");
assert!(refused(&board).is_none(), "a half typed verb was refused");
assert!(!row.contains("Usage"), "a label reached the row: {row}");
}
/// The footer is one row, whatever the usage is: a second row would reflow the columns
/// every time a verb resolved, which is the same reason a notice of several lines is a
/// box rather than more footer.
#[test]
fn a_usage_wider_than_the_row_is_cut_and_leaves_the_columns_alone() {
let dir = searchable();
let mut board = one_board(&dir);
// `set` has the longest synopsis in the tool, a flag per section and one per scalar.
let narrow = 60;
let before = render(&board, narrow, SCREEN.height);
board.open_prompt();
type_line(&mut board, "set ");
let after = render(&board, narrow, SCREEN.height);
assert_eq!(before.len(), after.len(), "the row count changed");
assert_eq!(
before[..before.len() - 1],
after[..after.len() - 1],
"the columns moved to make room for the usage"
);
let row = after.last().expect("no rows rendered");
assert!(row.chars().count() <= narrow as usize, "the row ran over: {row}");
assert!(row.contains('…'), "a usage too wide for the row was not cut: {row}");
assert!(row.contains(" :set "), "the line lost its cells to the usage: {row}");
}
/// The line is edited where the cursor is, not at its end, which is the difference
/// between a command line and the search prompt's append-only query.
#[test]
@@ -8969,8 +9193,21 @@ approach";
let row = footer(&board, SCREEN.width, SCREEN.height);
assert!(!row.contains(":mv"), "the head of the line held the row: {row}");
// The slot at the end of the row keeps its width whatever is typed into the line,
// which is what taking it off the row before windowing is for. `mv` resolves, so
// what holds it is the usage; a verb that resolves to nothing leaves the keys.
assert!(row.contains("mv [OPTIONS]"), "a long line pushed the usage off: {row}");
board.dismiss(SCREEN);
board.open_prompt();
type_line(&mut board, &format!("nosuchverb {}", "a".repeat(200)));
let row = footer(&board, SCREEN.width, SCREEN.height);
assert!(row.contains("esc cancel"), "a long line pushed the keys off: {row}");
board.dismiss(SCREEN);
board.open_prompt();
type_line(&mut board, &format!("mv {}", "a".repeat(200)));
let mut terminal = Terminal::new(TestBackend::new(SCREEN.width, SCREEN.height)).unwrap();
terminal.draw(|frame| draw(frame, &board)).unwrap();
assert!(