cue/format/printer: fix indentation for clauses

[FogDong]

Signed-off-by: Fog Dong [email protected]

This PR reset the allow for clauses at the position 0.

Fixes #1544

[FogDong]

PTAL😀 /cc @myitcv

[myitcv]

I'll defer to @rogpeppe here :)

[mvdan]

what about adding a check in writeWhitespace to never allow starting a file with a whitespace character? that seems to never make sense.

an alternative is why allowed begins with blank being set when printing the first token; it seems like we should start with just ignore, the zero value.

[FogDong]

what about adding a check in writeWhitespace to never allow starting a file with a whitespace character? that seems to never make sense.

an alternative is why allowed begins with blank being set when printing the first token; it seems like we should start with just ignore, the zero value.

That makes more sense, thanks for the review, I'll fix it as suggested.

[FogDong]

PTAL 😄 /cc @mvdan

[mvdan]

@Lyoness and myself are reviewing this together :) give us a couple of days if that's okay.

[mvdan]

I had a hunch that we could instead investigate why f.allowed had blank set when we were printing the first clause token. I debugged that a bit with @Lyoness and we ended up with a different fix, which you can find at https://review.gerrithub.io/c/cue-lang/cue/+/538624.

What do you think, @FogDong?

[FogDong]

LGTM, you guys know cue better than me after all. @mvdan

Still, I think it's better for the community if we can have some more discussion before another PR. Although it took some time to investigate this issue and made my first PR, it's time to close it.

[mvdan]

Certainly - and the last thing I want to do is to discourage further contributions :)

I did try to nudge the discussion in another direction in an earlier comment:

an alternative is why allowed begins with blank being set when printing the first token; it seems like we should start with just ignore, the zero value.

However, it was hard for me to say whether that idea would turn out to be a good fix without trying it. It's far from ideal if I ask you to implement a solution without me knowing ahead of time whether it will work or not - I'm new to the cue/format codebase as well.

So I thought it would be best to just try and see what happens. I'm still not sure whether Paul or Marcel will say that the second solution is better - if not, perhaps we'll come back here.