I haven't yet figured it out... Probably will have a querrying API or smth.

I have a bunch of ideas about file watches & update(), flush(), reload(), assign_file()

1. dotted.paths.are.not.allowed
2. [[ lists_of_tables_are_not_allowed ]]
3. "quoted keys" = "are not allowed"

Why?

Because they are annoying.

[table]
they = "exists"

[and]
work = [ "normally", "-ish" ]

Keys can, technically, be whatever... This makes the statement "utoml is toml subset" false, but also:

= = 5 # is totally valid
# so...

Also, duplicate keys are valid (by default). In the same file: they get overwritten in order In different files: they get overwritten in the order the files were parsed, e.g., via parse_userfile

There are 2 types of text:

basic_string = "- Simple, \n- Escapable\n text"
and = "
    multiline
    by
    default
"

And

literal_string = 'where \ is just a character.'
so = '\u1234' # doesn't work

TOML's multiline strings ("""text""") are also allowed for better syntax highlighting.
In non-literal strings use \ to escape text & TOML's \uXXXX and \UXXXXXXXX for unicode characters.

Equivalent to TOML.

a = 123
b = 0xDEADBEEF
c = 123_456_789

Equivalent to TOML.

a = 123.456
b = 123e2
c = -inf

Equivalent to TOML.

a = true
b = false

Equivalent to TOML.

a = 2025-11-23
b = 12:30:00
c = 2025-11-23 12:30:00+02:00

Equivalent(-ish) to TOML.

list = [ 1, 2, 3 ]
yeah = [ [ 'a', 'b' ],,,, 5, ]

Similar to TOML.

table = {
    b = { c = 1, d = 2 },
    e = [ 1, 2 ]
}

You cannot have dotted paths:

a = { terrible.better.yet.horrible = 999 }
also.invalid.here = {}
[and.here]