should generate_graph_children() filter out abs deps from other cycles

[oliver-sanders]

There's an if statement in the code that filters out absolute dependencies from other cycles.

It's not entirely clear why this statement was added, the corresponding statement in generate_graph_parents was removed, should this one be removed too?

See this comment:

What about this equivalent bit in generate_graph_children()?

cylc-flow/cylc/flow/taskdef.py

Lines 49 to 51 in 4d6430a

	if is_abs and trigger.get_parent_point(point) != point:
	# If 'foo[^] => bar' only spawn off of '^'.
	continue

Originally posted by @MetRonnie in #6103 (comment)

• Are there any consequences to this statement? Are there graphs where this could cause a bug?
• Are there any consequences to removing this statement? E.G. Does it impact performance, causing double counting?

@dwsutherland, any insights welcome.

[dwsutherland]

Not entirely sure.. Because the children of [^] are potentially infinite, where the parents are not?
Don't think I was the original author (although possible?)

[oliver-sanders]

Not entirely sure.. Because the children of [^] are potentially infinite, where the parents are not?

That makes good sense!

[oliver-sanders]

With the change to grah-parents, abs deps seem to be working as expected, @MetRonnie, please close if you're happy with this explanation.

[MetRonnie]

Ok I've managed to figure out a case where this continue is hit

        P1 = foo
        R1/2 = foo[1] => pub

Without this continue, it results in edges that shouldn't exist (only 1/foo -> 2/pub should exist):