Module fmt/spec

Format specifications — the ${value:spec} vocabulary (STD_API_AUDIT D3.10).

The grammar is Rust's, minus dynamic width/precision:

spec := [[fill]align][+][#][0][width][.precision][kind]
align := "<" | ">" | "^"
kind  := "x" | "X" | "b" | "o"

Example

{ FormatSpec } :: import "std/fmt/spec";

s := FormatSpec.parse(">8.2");
s.pad(`3.14159`, Alignment.Left);   // "    3.14"

Widths and precisions are RUNE counts, not byte counts — _apply_width, pad and pad_numeric all measure with chars().count(), so é occupies one column and a truncation can never cut a rune in half. That is the one place in the string world where the unit is deliberately not bytes: a format spec describes a COLUMN, and D4's byte basis would have made accented rows come out ragged.

Stability

unstable — two shapes are still open, and both are breaking to change.

The grammar is missing Rust's DYNAMIC width and precision ({:w$}, {:.p$}), as the module header says. Yo has no format macro to collect the extra arguments, so adding them means inventing a spelling (spec.with_width(n)? a second pad overload?) rather than extending the parse table.

And parse is TOTAL: it never reports a bad spec, it ignores what it does not recognise, so "{:.2x" degrades silently to something a caller did not ask for. That is the right default for interpolation — a typo in a log line should not abort the program — but it also means a spec typo is invisible forever, and giving parse a Result later is breaking. Freezing waits on deciding both, not on a release count.