chore(remap): condition without subfields is set to Remap source

[StephenWakely]

Ref #4791

This updates the conditions such that if you don't specify a subfield it will take the value as the remap source.

So both these examples work and are equivalent:

[transforms.filter]
  type = "filter"
  inputs = ["stdin"]
  condition.type = "remap"
  condition.source = 'contains!(.message, "funky")'

  type = "filter"
  inputs = ["stdin"]
  condition = 'contains!(.message, "funky")'

check_fields conditions work and can be set as before. A deprecation warning is still emitted.

Signed-off-by: Stephen Wakely [email protected]

[binarylogic]

Thanks! Just making sure these tests are adequate. We've been bitten by these types of changes in the past. Would it make sense to add a behavior test for the new remap syntax?

[StephenWakely]

We've been bitten by these types of changes in the past. Would it make sense to add a behavior test for the new remap syntax?

I have already moved all the behaviour tests over to Remap in a previous PR, but I have left them specifying the type="remap. It seems the unit tests actually take an array of conditions so rather than specifying them as they currently are:

    [[tests.outputs.conditions]]
      type = "remap"
      source = '''
        .a.b == 123 && \
        .x.y == 456 && \
        .x.z == 789
      '''

you can do:

    conditions = ['''
        .a.b == 123 && \
        .x.y == 456 && \
        .x.z == 789
      ''']

I didn't like the way this looked at first, so I hesitated to change it. But I'm actually warming to it now that I type this comment. I'll move them over...

[bruceg]

I echo @pablosichert's test concerns, but otherwise LGTM

[pablosichert]

Some suggestions, otherwise LGTM

[fanatid]

Suggested change

	r#"Map(RemapConfig { source: ".nork == true" }"#,
	r#"Map(RemapConfig { source: ".nork == true" })"#,

Same for 2 tests above. Rest looks good 👍