refactor: switch to *Python*'s official tomllib, as toml cannot parse inline table values properly

[stevenxxiu]

Thanks for this library.

I use Prospector often, which makes use of this library. In Prospector I often use inline tables, like:

[tool.pdm.scripts]
post_install = {shell = '''
    mv "${PDM_PROJECT_ROOT}/__pypackages__" "${PDM_PROJECT_ROOT}/nushell/";
    ln --symbolic "${PDM_PROJECT_ROOT}/nushell/__pypackages__" "${PDM_PROJECT_ROOT}"
'''}

This cannot be parsed by the toml library, which this library uses. As you can see in https://github.com/uiri/toml/blob/0.10.2/toml/decoder.py#L654:

def load_inline_object(self, line, currentlevel, multikey=False,
                       multibackslash=False):
    candidate_groups = line[1:-1].split(",")
    # ...

The parsing is really basic and is clearly going to tail. It assumes inline tables are all in one line.

This file was last updated 2020/10/30 so seems rather dead.

This bug is also mentioned at uiri/toml#348, quite some time ago.

Python now has a native Toml library, so it makes sense to switch to that. The API seems quite compatible, as I only had to switch the library, and all tests still pass.

[nat-n]

Hi, I just arrived here from encountering and then wanting to solve the same problem 😄 I see you got here first, but your solution is not quite right.

tomllib is only available in the standard library for python >=3.11.

Since this project declares compatibility with python >=3.7, you also need to add a dependency on tomli. It's exactly the same package as tomlib for python <3.11

You'll need to do something like this for full compatibility: https://github.com/nat-n/poethepoet/blob/903ed9af0dd313a6da32f24ec4357101435976d8/poethepoet/config.py#L5