Trait OutputChannel

Source

pub trait OutputChannel: Send + Sync {
    // Required method
    fn op_output(&self, line: &str);

    // Provided methods
    fn log(&self, level: LogLevel, message: &str) { ... }
    fn raster(&self, frame: &str) { ... }
}

Expand description

The single conduit for the one user terminal. Exactly one impl is installed per run via install; producers submit to a bucket and only the impl touches an fd. Every method is non-blocking.

Push 1 defines the op-output bucket only; log / status / raster land in later pushes (SRD-87 §4).

Required Methods§

Source

fn op_output(&self, line: &str)

op-output bucket: one adapter-rendered op line. Reaches the terminal/file the channel owns — never around it.

Provided Methods§

Source

fn log(&self, level: LogLevel, message: &str)

log bucket: project one diagnostic line to the live terminal (SRD-87 §5). This is the output half only — the durable session.log write and the fold-ring append are L1 intake, done by observer::log_categorized before this is ever called, and the sink_active/min_level gate (whether a line reaches the live surface at all) stays with the observer. The default writes the colorized line to stderr — the fd the channel owns. A sink-active interactive run never reaches here (the sink renders from the ring); a console-owning run never reaches here (suppressed upstream); piped and bootstrap runs land here.

Source

fn raster(&self, frame: &str)

raster bucket: a fully-rendered, self-contained terminal frame (braille / ANSI cells) — e.g. the plotter’s canvas (SRD-87 §5). The producer owns layout (cursor controls, line endings); the channel owns the fd. The default writes the frame bytes raw to the stdout the channel owns — correct for a console-owning adapter (the plotter owns the screen) and for a piped plot redirected to a file.