[stevespringett]

Closes #655

[bakkot]

Thanks for the contribution. This seems like something which should be opt-in, and also should generate a warning since you'll often want manual remediation. I've pushed up commits doing both (+ a test). If this still works for you I'll land.

[stevespringett]

@bakkot opt-in sounds reasonable as does displaying a warning.

[michaelficarra]

So we're just flattening the numbering and not the actual structure of the emu-clauses? That seems bad for doing things like targeting exactly-N-times nested emu-clauses in CSS.

[bakkot]

I don't see why it would be? From what I understand the ISO standards only specify a limit on the numbers themselves, not on the presentation of clauses. Changing how we number clauses doesn't seem to me like it ought to interact with CSS or anything else.

[michaelficarra]

I really don't like the auto-collapsing in this PR. I would strongly advise against it. The computer doesn't know, when collapsing a 6-time-nested clause, whether it's more appropriate to collapse the leaf into its parent or any one of the other 5 parent nodes along the way into their parent instead. That's a decision that the author should make. If it were up to me, I would make this a warning only, which could be caused to fail the build using --strict and corrected by a human.

[michaelficarra]

I don't see why it would be? From what I understand the ISO standards only specify a limit on the numbers themselves, not on the presentation of clauses.

The clause numbering should always coincide with the nesting of emu-clauses.

[bakkot]

This PR does not do any auto-collapsing. (I originally thought it did but I was misreading.) It only touches the numbering.

The clause numbering should always coincide with the nesting of emu-clauses.

That's a valid preference but unless you intend to use this feature I would prefer to defer to the preferences of the people who are actually going to use this. The warning is still emitted and enforceable with --strict, it just also makes the numbering compliant with ISO for people who are not enforcing the warning.

[michaelficarra]

I prefer not to include features that I think are both broken in implementation and footguns, but as you point out, I have no plans to use the feature myself, so I'm not going to get in your way.

[bakkot]

Version 21.4.0 is out including the new --max-clause-depth flag.