Fix warnings for Elixir 1.16 and update Elixir version requirement

[wkirschbaum]

No description provided.

[wkirschbaum]

The order of these attributes changes between Elixir v1.14 and v1.15, but don't believe it has a significant impact.

[joshnuss]

Any reason to stop testing OTP 22.x, Elixir 1.6 - 1.11?

[wkirschbaum]

I don't think there is any harm to include older version if it does not impede future supported versions. Elixir 1.11 is not "supported" anymore: https://hexdocs.pm/elixir/1.16.2/compatibility-and-deprecations.html

Since upgrading Elixir should be very simple, it should be acceptable for new libraries releases to only support from 1.12.x.

I will add Elixir 1.11 to see if it passes the tests.

[wkirschbaum]

Adding 1.11 gives the following compilation errors:

== Compilation error in file lib/nimble_parsec/compiler.ex ==
Error: ** (CompileError) lib/nimble_parsec/compiler.ex:931: cannot find or invoke local _/1 inside match. Only macros can be invoked in a match and they must be defined before their invocation. Called as: _(/ / _ = range)
could not compile dependency :nimble_parsec, "mix compile" failed. You can recompile this dependency with "mix deps.compile nimble_parsec", update it with "mix deps.update nimble_parsec" or clean it with "mix deps.clean nimble_parsec"
    (stdlib 3.17.2.4) lists.erl:1358: :lists.mapfoldl/3
    (stdlib 3.17.2.4) lists.erl:1359: :lists.mapfoldl/3
    (stdlib 3.17.2.4) lists.erl:1358: :lists.mapfoldl/3
    (stdlib 3.17.2.4) lists.erl:1359: :lists.mapfoldl/3
    (elixir 1.11.4) expanding macro: Kernel.".."/2
    lib/nimble_parsec/compiler.ex:931: NimbleParsec.Compiler.newline_allowed?/1

so think it is better to set the minimum to elixir 1.12. wdyt?

[joshnuss]

Thanks @wkirschbaum!

Looks good, I left a small comment, but otherwise good to go.

Do you know why order of attributes in Map changed in newer version of Elixir?

[wkirschbaum]

@joshnuss I am not entirely sure, but according to the Map documentation: Key-value pairs in a map do not follow any order (that's why the printed map in the example above has a different order than the map that was created).

So either it has to be manually ordered after retrieval or we need to accept ordering won't be predictable.

Regarding, the 1.12 constraint on Elixir: it is because of a failure of the latest version of the nimble_parsec library, but also because Elixir 1.11 is not "supported" anymore, so probably not worth supporting it in this library anymore.

[joshnuss]

Thanks @wkirschbaum

OK, sounds good, let's set 1.12 as the minimum.

So either it has to be manually ordered after retrieval or we need to accept ordering won't be predictable.

My only concern is that downstream, people will have unit tests making the same assumption, and they will break. But the alternative is sorting, which will hurt perf

[wkirschbaum]

Its already a problem. Nothing in the PR changes the behaviour, we are just testing more versions to expose these kinds of issues.

[joshnuss]

Thanks @wkirschbaum!!