Fix workflow-state command and xtrigger.

[hjoliver]

• handle --status and --output queries the same, with respect to polling vs one-off queries
• unified polling (for db not found, not connected, or condition not satisfied)
• close workflow-state: check latest instance for status #5143 (check latest instance of a task)
• make the workflow-state command and xtrigger flow-specific: add --flow option
• log warnings for failed xtrigger calls (currently these are silent except in debug mode)
• supersede workflow-state: query latest task instance #5144
• (alternate run-dir already done Fix workflow-state alternate run-dir #6031 )
• retain --message option (task message) as well as --output (trigger name) to avoid breaking existing polling workflows

---

Minor db change Change to the workflow database structure : Format of outputs column in task_outputs table has changed from Cylc 8.0.0 where it was a stringified list of task messages, to a stringified dictionary of the format {<output label>: <task message>} (like how it was in Cylc 7).

---

The command and xtrigger now take WORKFLOW//CYCLE/TASK:SELECTOR as an arg.

SELECTOR can be a status or an output. With --output, or if not a valid status, it is taken as an output; otherwise a status.

Any query will be polled for if no matching result is found.

A good test case:

[scheduling]
    cycling mode = integer
    initial cycle point = 2001
    final cycle point = 2003
[scheduling]
    [[xtriggers]]
        ws = workflow_state("%(workflow)s//%(point)s/foo:succeeded", offset="-P1"):PT2S
    [[graph]]
        P1 = """
           foo[-P1] => foo
           foo:x => bar & baz
        """
        R/2002/P1 = "@ws => baz"
[runtime]
    [[foo]]
        script = """
           cylc message -- "xxx"
        """
        [[[outputs]]]
           x = xxx
    [[bar]]
        script = """
          if [[ "$CYLC_TASK_CYCLE_POINT" == "$CYLC_WORKFLOW_FINAL_CYCLE_POINT" \
             && "$CYLC_TASK_SUBMIT_NUMBER" == "1" ]]
          then
              cylc trigger --flow=new $CYLC_WORKFLOW_ID//2002/foo
          fi
        """
    [[baz]]

Check List

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• Applied any dependency changes to both setup.cfg (and conda-environment.yml if present).
• Tests are included (or explain why tests are not needed).
• CHANGES.md entry included if this is a change that can affect users
• Cylc-Doc pull request opened if required at Some doc updates for 8.3.0.dev changes. cylc-doc#727
• If this is a bug fix, PR should be raised against the relevant ?.?.x branch.

[hjoliver]

Note also: #6030

[MetRonnie]

tests/k/xtriggers/01-workflow_state.t failing btw

[wxtim]

Seems reasonable; Given the amount on your plate I will start drafting tests for this so that we can ram it through.

I've added the following:hjoliver#50 - It's not a complete set of tests, but I've covered some of this. I'll do more tomorrow. If you want to get them in feel free to merge though.

[hjoliver]

One-line fix pushed for the function signature advice (we shouldn't advise adding all the null args with = None).

[hjoliver]

And another small commit handles "None" properly in xtrigger defintions, just in case someone does it.

[hjoliver]

and the same as the above but with alt_cylc_run_dir=None added to the xtrigger args causes it the xtrigger to not return True.

That would be because it was interpreting the string "None" as an alternate run-dir. My previous commit will prevent that too.

[wxtim]

I've had a really decent go at breaking this, and run out of ideas.

[hjoliver]

@MetRonnie - final comments addressed. Merge? 🙏

[MetRonnie]

Alright, think we're nearly there. However there is this still outstanding? hjoliver#54

[MetRonnie]

💫

[wxtim]

🚀