Support `Decimal` and `MutableMapping` codecs, add a multiline-`Array` indent control

[goodboy]

In an attempt to resolve the feature request issues i submitted:

• 288

• 289

• 290

---

Summary:

• add a Array._multiline_indent: str and expose it through a indent: str kwarg to .multiline().
  • main purpose is to be able to style the indent level for user preferred indentation of say a multiline array of tables B)
• add a branch to the .items.item() encoder to handle decimal.Decimal and match against MutableMapping instead of dict.
• tests for all of it!

---

Possibly still todo?

• [ ] Do i need to submit more for Decimal and MutableMapping changes?
  • currently i added 2 new tests for decimal encoding in the best places i could find.
  • don't know if mut map tweaks require anything extra other then maybe some tests maybe using a custom type? ( i feel like that is just testing the protocol and thus stdlib stuff 😂 )
• [ ] tests for the inline Array indent parameter:
  • probably in tests_items.py i presume?

[goodboy]

Bah, going through rebase to master 🏄🏼

[goodboy]

There hopefully i got all the latest changes in now B)

[goodboy]

Also, if this patch is likely to never be accepted or needs to be changed to be accepted please let me know asap since we'd much rather not have to maintain our own fork :)

[frostming]

a better way to support this and similar cases is to allow passing a custom encoder to item() function.

i don't think we need round trip

[goodboy]

@frostming ah ok, I'm fine to change it to do that.

(also sorry i missed your comment don't have my email hooked up rn for GH 😂)

is to allow passing a custom encoder to item() function.

can you show an example that might already exist for doing this btw in the code base and for which feature addition out of the issues I made does this apply? All?

---

i don't think we need round trip

Also can you clarify wym by that?

[frostming]

You can convert a decimal to a float representation in TOML, but can't convert it back, and Decimal isn't listed as the data types defined by TOML spec, so it's not worth to expose a new api and item type for it.

Instead, I'd propose something like this:

import tomlkit
a = Decimal('34.56')

def my_encoder(o):
    if isinstance(o, Decimal):
        return tomlkit.float_(o)
    raise ValueError(f"Invalid type {type(value)}")

item = tomlkit.item(a, encoder=my_encoder)   # a Float object
v = {"a": Decimal("34.56")}
print(tomlkit.dumps(v, encoder=my_encoder)
# a = 34.56

In this way users can convert custom types, not only Decimal, to TOML items in tomlkit. And it is clearer that users can't convert it back(parse anything as Decimal)

[goodboy]

so it's not worth to expose a new api and item type for it. In this way users can convert custom types, not only Decimal, to TOML items in tomlkit. And it is clearer that users can't convert it back(parse anything as Decimal)

@frostming works for me. So I guess i'll just drop all the Decimal related changes then and rebase this branch a bit.

Any other quiffs and/or do you have any suggestions for the tests I still need to do?

[frostming]

@goodboy yes, I've proposed a change in #296

[goodboy]

@frostming thanks for the follow up!

i will hopefully get back to this PR this week. Will check out that landed PR and change our stuff to match first, then come back here and drop the unneeded history.

Still kinda looking for a little guidance on how thorough the new tests for the multi-line arrays should be? i have some basic ideas but don't want to push too deep if y'all aren't that concerned about it.