Bad recurrence exclusion warning

[hjoliver]

See https://cylc.discourse.group/t/cannot-have-more-than-1-repetition-for-zero-duration-recurrence/705/13

[scheduler]
    cycle point format = %Y
[scheduling]
    initial cycle point = 3000
    [[graph]]
        P1Y ! ^ = "foo[-P1Y] => foo"
[runtime]
    [[foo]]

Validation:

$ cylc validate bug
WARNING - Cannot have more than 1 repetition for zero-duration recurrence 3000.
Valid for cylc-8.3.0.dev

The warning refers to the exclusion point ^.

[oliver-sanders]

@hjoliver, is the issue that the error message is wrong (i.e. "Must have more than 1 repetition..."), or would we rather let Cylc silently ignore the null-recurrence?

[hjoliver]

Maybe I've misunderstood how exclusions are supposed to work, but it seems to me that this:

P1Y !^ = ...

should represent a single-point (^) exclusion - not a broken recurrence exclusion - in which case there should not be any warning message. ??

[oliver-sanders]

I think I thought that P1Y was R1 for some reason.

Yes, P1Y!^ should work, T00!^ is a documented example, possibly a more recent breakage?

[oliver-sanders]

This warning is coming from parsing the exclusion itself, you'll get the same warning if you do:

[scheduling]
  [[graph]]
    ^ = foo

Because "^" gets replaced with "3000" and "3000" is not a valid expression.

The correct way of writing it is like this:

[scheduling]
  [[graph]]
    R1/^ = foo

And for the exclusion example:

[scheduling]
  [[graph]]
    P1Y ! R1/^ = foo

[oliver-sanders]

When we parse an exclusion, we first attempt to parse it as a sequence, then fall back to parsing it as a point.

It's this first attempt at parsing which causes the zero-duration warning. I've raised a PR to suppress this logging for exclusion parsing.

[hjoliver]

The correct way of writing it is like this: ... And for the exclusion example:

[scheduling]
  [[graph]]
    P1Y ! R1/^ = foo

According to our documentation that is correct but not necessary, because single point exclusions are also supposed to be correct.

... [However, going by your last comment and new PR, you've come to the same conclusion yourself 👍 ]

[oliver-sanders]

However, going by your last comment and new PR, you've come to the same conclusion yourself

...and being the author of that functionality...