Add tests from @iarna/toml-spec-tests in prescribed JSON format

[pyrmont]

Summary

This PR adds almost¹ all of the tests from @iarna's toml-spec-tests repository but:

1. uses JSON rather than YAML for the output files;
2. encodes the JSON in the format described in the JSON encoding document; and
3. deletes the YAML files since they are no longer necessary.

All credit for writing the original TOML tests belongs to @iarna and the contributors to her repository.

Background

I created the tests for my own use in developing Tomlin, a TOML parser in Janet. For my purposes, I test and compare the results using a Janet script. As a result, I have not tested this with the Python script in this repository. Nevertheless, since the time-consuming work is in writing the JSON files in the prescribed format, I wanted to share what I'd done.

Notes

This PR conflicts with #5. While that PR was made first, given that it includes YAML output files, I would argue that this PR is to be preferred.

¹ The nested QA tests (here and here) break Janet's JSON encoder/decoder and I have not written JSON output files for them.

[pyrmont]

As part of the PR, I've also removed the mixed-line-ending and trailing-whitespace hooks from .pre-commit-config.yaml. These hooks conflict with tests that are checking how a TOML parser handles various characters.

[pradyunsg]

Thanks for the PR! ^.^

Looks like I know what I'm reviewing all of Saturday morning!

[hukkin]

Thanks pyrmont, this was incredibly useful to me!

Here's two missing JSON files (note that these are NOT the nested tests that you mentioned, seems like in total 4 files were missing)

• valid/qa/qa-scalar-literal-multiline-40kb.json
• valid/string/spec-string-escape-4.json

[hukkin]

Oh, and I've got one question about the encoding used for escaped characters: why are some characters double escaped in JSON?

E.g.

{"a":
  {"type":"string","value":"\\n"}}

The actual value we want is just a newline. However in the JSON we have a backslash plus a newline.

On the other hand we have

{"a":
  {"type":"string","value":"\""}}

Both the actual value and the JSON encoded value are just a single quote character.

This seemed inconsistent, but maybe I'm missing something obvious?

[pyrmont]

@hukkinj1 Thanks for the kind words and thanks for spotting the missing files! Not sure what happened there.

And you were right, I just completely screwed up the JSON encoding. To be honest, I'm still confused so I'm not 100% sure this is correct but I think you were right and that the escape sequences in the string values don't need to be escaped. I've fixed that now and it's hopefully correct!

[hukkin]

The escapes seem good to me now. Thanks once more!

[CAM-Gerlach]

Hey @pradyunsg have you gotten a chance to review this, or is there anything blocking this PR from being merged? It is specifically mentioned in @hukkin 's PEP 680 and elsewhere.

As part of the PR, I've also removed the mixed-line-ending and trailing-whitespace hooks from .pre-commit-config.yaml. These hooks conflict with tests that are checking how a TOML parser handles various characters.

FYI, instead of removing those hooks completely from the whole repo, you can exclude files by type (with exclude_types) or by regex (with exclude).

[encukou]

According to a strict reading of the MIT licence, the copyright notice should appear exactly as in the original, so something like:

Suggested change

	Copyright (c) 2021 Pradyun Gedam and Contributors
	Copyright (c) 2021 Pradyun Gedam and Contributors
	Copyright Rebecca Turner <me@re-becca.org>

would be more appropriate than adding a new note at the bottom.

[CAM-Gerlach]

(I am not a licensed attorney, and the following should not be taken as formal legal advice.)

I concur with what you state above, as modifying the license text in this manner while not preserving the copyright notice intact, at its location in the original, is not an accepted approach for what is intended here.

To note, I do offer an alternate suggestion, as multiple copyright lines of text, rather than the single one matching the form of the one specified as variable in the SPDX definition of the MIT license, often breaks programmatic license detection and validation, including GitHub's. An alternative would be placing both complete copyright notices on the same line with clear lexical separation between them, like this:

Suggested change

	Copyright (c) 2021 Pradyun Gedam and Contributors
	Copyright Rebecca Turner <me@re-becca.org>; Copyright (c) 2021 Pradyun Gedam and Contributors

Given the original copyright notice is preserved and set off with a semicolon, both follow the form of standalone copyright notices, and this one line is permitted to vary without changing the legal effect of the license, it would seem equivalent separation is achieved without resorting to a line break.

[arp242]

Hey @pradyunsg have you gotten a chance to review this, or is there anything blocking this PR from being merged?

AFAIK the status is that https://github.com/BurntSushi/toml-test is the de-facto standard: toml-lang/toml#585 (comment)

So ideally, toml-test should be moved here, overwriting this.

[pyrmont]

I'm not sure what the delay has been but I'll close this PR as it's been outstanding for more than a year at this point and after a cursory examination, the tests in @BurntSushi's repository look more comprehensive.