Added compact JSON format

[mitsuhiko]

This adds a compact JSON format. For up to 120 characters the snapshot is put in a single line, beyond that multiple lines are emitted.

Refs #287

[mitsuhiko]

I had to bump up to 1.56 due to matklad/once_cell#201

[mitsuhiko]

@irevoire does this address your issue?

[irevoire]

Well, it's better as default behaviour, but don't you think being able to choose what's going to happen would be even better? 👀

[mitsuhiko]

What do you mean by "chose what's going to happen"?

[irevoire]

Having this PR as the default behaviour and then a compact and pretty macro I guess 🤔

[mitsuhiko]

You can always alias the macro yourself? Generally I think the compact behavior is likely inferior for most uses as it results in worse diffability.

[irevoire]

I can change the name, but I can't create a new macro that only uses the “default” view without forking insta, right?
After this PR, the best I'll be able to do is to use the compact view that stops working once I reach 120 chars?

---

Here is the kind of use-cases where it gets unreadable if you use the prettyfier;

    #[test]
    fn anything_and_index_deletion() {
        // The indexdeletion doesn't batch with anything that happens AFTER
        assert_smol_debug_snapshot!(autobatch(input_from([IndexDeletion, DocumentAddition])), @"Some(IndexDeletion { ids: [0] })");
        assert_smol_debug_snapshot!(autobatch(input_from([IndexDeletion, DocumentUpdate])), @"Some(IndexDeletion { ids: [0] })");
        assert_smol_debug_snapshot!(autobatch(input_from([IndexDeletion, DocumentDeletion])), @"Some(IndexDeletion { ids: [0] })");
        assert_smol_debug_snapshot!(autobatch(input_from([IndexDeletion, DocumentClear])), @"Some(IndexDeletion { ids: [0] })");
        assert_smol_debug_snapshot!(autobatch(input_from([IndexDeletion, Settings])), @"Some(IndexDeletion { ids: [0] })");

        // The index deletion can accept almost any type of BatchKind and transform it to an IndexDeletion
        // First, the basic cases
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentAddition, IndexDeletion])), @"Some(IndexDeletion { ids: [0, 1] })");
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentUpdate, IndexDeletion])), @"Some(IndexDeletion { ids: [0, 1] })");
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentDeletion, IndexDeletion])), @"Some(IndexDeletion { ids: [0, 1] })");
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentClear, IndexDeletion])), @"Some(IndexDeletion { ids: [0, 1] })");
        assert_smol_debug_snapshot!(autobatch(input_from([Settings, IndexDeletion])), @"Some(IndexDeletion { ids: [0, 1] })");

        // Then the mixed cases
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentAddition, Settings, IndexDeletion])), @"Some(IndexDeletion { ids: [0, 2, 1] })");
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentUpdate, Settings, IndexDeletion])), @"Some(IndexDeletion { ids: [0, 2, 1] })");
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentAddition, Settings, DocumentClear, IndexDeletion])), @"Some(IndexDeletion { ids: [1, 3, 0, 2] })");
        assert_smol_debug_snapshot!(autobatch(input_from([DocumentUpdate, Settings, DocumentClear, IndexDeletion])), @"Some(IndexDeletion { ids: [1, 3, 0, 2] })");
    }

Honestly, the more I think about it, and the more I think you're right, 120 chars are probably enough for everyone; I just find it strange to force a limit instead of letting people assume their bad choices with an always compact view and an always prettified version 😁

[mitsuhiko]

I think the problem if it does not have some sort of auto detection is that you are forced to use different types of assertion macros all the sudden if the thing you assert on gets too large. I'm not sure that's a good user experience.

[irevoire]

That's true.
Well, thanks for the PR and sorry for thinking too loud. I actually think it totally fixes my « issue »!