chore(remap): Make del function only take one arg

[lucperkins]

At the moment, the del function can be used to delete any number of object fields. This capability, however, is inconsistent with an emerging core tenet of TRL, which is that functions should return a single value. If del is called on two fields and only one exists, it's not clear what the appropriate response should be. Furthermore, it's been decided that del should return the value of the field deleted rather than null, in order to "bubble up" the value of the field upon deletion and thus enable expressions like this:

.ts = to_timestamp(del(.timestamp))

Closes #5605

[lucperkins]

@JeanMertz @FungusHumungus There's a bit more work that needs to be done here to have del return the value, but I'm not sure about the desired behavior. What's the desired behavior for del(.non_existent_field)?

I'll finish this up when we get the desired end state sorted.

[binarylogic]

@lucperkins does this close an issue?

[lucperkins]

@binarylogic It might address #5566, but I'm going to hold off on pushing any further on this until there's consensus around expected behavior.

[jamtur01]

Also linked to #5605.

[jamtur01]

To move this forward: @binarylogic @lucperkins For this release we just need to answer: should calling del on a non-existing field error or return null?

[binarylogic]

For now, it should return null until we finalize the error handling story.

[StephenWakely]

It looks like this is always returning Null. It would be really handy if we could address #5150 here as well so that del returns the value of the field being deleted - only returning Null if the field doesn't exist.

I'm sure I saw somewhere you added a remove_and_get function to Object that handled this case, but I can't seem to find it now.

[lucperkins]

@FungusHumungus You're not just seeing things 😄 I did have that. I think I misunderstood the outcome of the discussion above and I now see that del should return null only if the field doesn't exist. I'll revert.

[lucperkins]

@FungusHumungus Okay, I've reverted that behavior. I've left remove_and_get unimplemented for now in LogEvent and Metric. As far as I can tell those implementations aren't currently necessary for VRL, but if that's incorrect, let me know and I'll dig in and supply them, though I may need a bit of help.

[JeanMertz]

This is looking good, but isn't quite what we want just yet. I left some notes on what I think we should do before we merge this PR.

[JeanMertz]

I updated the PR based on the latest consensus in #5633 (comment). This should be good for a final round of reviews before we merge it.

[jszwedko]

Looks good! Just a few small suggestions / questions.

[jszwedko]

Nice work!

[jszwedko]

TIL we support ifs as expressions 😄

fixed