digdag should fail on multiple operators within a task

[danielnorberg]

I.e., the below workflow definition should fail instead of silently just executing echo bar.

run: +foo
+foo:
  sh>: echo bar
  py>: baz

Also, it might be worth considering if multiple identical keys should fail. In the below workflow definition the second sh> operator implicitly overrides the first operator when the yaml is parsed because yaml.

run: +foo
+foo:
  sh>: echo bar
  sh>: echo baz

[rbparrish]

From how I think about the syntax, I feel multiple identical operators should also fail. Only one primary operator per task (not including the case of _check: & _error: usage, or multiple tasks within a task)

[danielnorberg]

Yeah, tricky thing is that the yaml parsing happens before validation, and at that point the parser has already discarded all preceding identical keys.

[danielnorberg]

Turns out though that the yaml parser can, at gunpoint, be persuaded into disallowing duplicate keys by injecting some validation logic. Kind of icky but it gets the job done.

[frsyuki]

SnakeYAML should have an option that throws an exception. It's good for all users. SnakeYAML is not on github (BitBucket!) but the author is still active.

[danielnorberg]

Yeah, let's submit a SnakeYAML PR.

[danielnorberg]

I opened an issue in the SnakeYAML project to propose adding this validation to SnakeYAML itself: https://bitbucket.org/asomov/snakeyaml/issues/337/option-to-disallow-duplicate-keys

Meanwhile, it seems that the very same issue has been addressed in the past by implementing a custom constructor, similarly to what my current PR does: https://jira.spring.io/browse/SPR-12318

Should we go ahead with my PR until we have managed to get unique key validation into SnakeYAML proper?