CI, scripts and agents#

Notebooks run in CI#

By default a failing command shows its error and the notebook goes on, as with !command. For a notebook that CI runs, stop at the first failure and bound the time of long commands:

%scikitplot_config raise_on_error=true timeout=900

A failure then raises CommandFailedError (its .result holds the exit status and the output), shown in two lines. A command stopped by the timeout has exit status 124 and result.timed_out.

Setting	Default	Meaning
`raise_on_error`	`false`	a non-zero exit raises `CommandFailedError`
`timeout`	`none`	seconds a subprocess command may run
`isolate`	`false`	run native commands in a subprocess too
`stream`	`true`	show subprocess output as it comes

Without IPython#

The engine behind the magic is a function, for scripts, tests and agents:

import io

from scikitplot._ipymagic import MagicSettings, run_command

result = run_command("info --format json", out=io.StringIO())
assert result.ok
record = result.to_dict()  # argv, command, mode, exit_code, stdout, data ...
assert record["mode"] == "in-process"

strict = MagicSettings(raise_on_error=True, timeout=60)
result = run_command(["env", "--format", "json", "doctor"], settings=strict,
                     out=io.StringIO())
assert "backends" in result.data

Every run is also logged at DEBUG on the scikitplot._ipymagic._bridge logger: command, exit status, time, where it ran and the size of its output (never the output itself).