`orca.timeline`

A run's Dista timeline and its workers over time, drawn as SVG on the server for the run's page.

The timeline is dista's, in Chrome's trace format: a process_name event names each worker's lane (pid), and each task is an X event on thread 1 (running) or 2 (waiting for inputs), with ts and dur in microseconds. Each worker is one lane; a task is a bar coloured by its name, and the time a task waited for its inputs a thin bar under it. Perfetto shows the same file in full; this is the glance.

The worker count comes from orca.instances, which an ec2 run has: one step per instance started or ended.

Structs

`Bar`

struct Bar

Implements Copyable, Movable

Fields: lane: Int; start: Float64: Seconds from the timeline's first event.; length: Float64; name: String; waiting: Bool: Waiting for inputs, rather than running.

`Timeline`

struct Timeline

Implements Movable

Fields: lanes: List[String]; bars: List[Bar]; end: Float64: Seconds from the first event to the last one's end.; left_out: Int

Functions

`read_timeline`

def read_timeline(text: String) -> Timeline

`timeline_svg`

def timeline_svg(t: Timeline) -> String

One lane per worker, its tasks as bars, under a time axis.

`workers_svg`

def workers_svg(changes: List[Tuple[Float64, Int]], total: Float64) -> String

The number of workers over a run, from changes (seconds from the run's start, and +1 or -1), as a step line.

Aliases

comptime MAX_BARS = 20000

Bars drawn at most; a timeline with more says how many it left out.