Dotted keys can't define sub-tables within tables indirectly defined in header form

[Validark]

I believe based on the online library tester that this library fails to handle this valid TOML document:

[fruit.apple.texture]  # you can add sub-tables
smooth = true

[fruit]
apple.color = "red"
apple.taste.sweet = true

I haven't actually used this library outside of the browser, so all I got for you is a playground link: https://godbolt.org/z/T59E7d

Related: toml-lang/toml#771

[marzer]

The behaviour is correct; that's not legal TOML. You can't mix [table.headers] and dotted.key.value = "pairs" like that.

The way to do this correctly would be:

[fruit.apple.texture]
smooth = true

[fruit.apple]
color = "red"
taste.sweet = true

Dotted keys are effectively just shorthand for creating nested inline tables.

[Validark]

My mistake, I didn't scroll all the way down in toml-lang/toml#771 to see that this issue is still being debated.

[marzer]

Aye, it's a pretty confusing part of the spec. Given the choice personally I'd remove dotted.key.value = "pairs" from the language entirely.

[awvwgk]

I'm intrigued by this example, maybe I should rather ask this to the discussion in toml-lang/toml#769, but it seems more related to this particular implementation to me.

Why would this TOML document be invalid

[fruit.apple.texture]  # you can add sub-tables
smooth = true

[fruit]
apple.color = "red"
apple.taste.sweet = true

but changing the order of the tables be valid instead if dotted keys were shorthands for inline tables? https://godbolt.org/z/ss9xsP

[fruit]
apple.color = "red"
apple.taste.sweet = true

[fruit.apple.texture]  # you can add sub-tables
smooth = true

[marzer]

but changing the order of the tables be valid

Well, yeah, it shouldn't be. I need to revisit this part of the code (and the spec!), evidently.

[marzer]

This is fixed now, though won't find it's way into the online parser until whenever that next gets updated; I'm not the maintainer of that project.

(cc @bobfang1992)