Upgrade to prettier 3

[nene]

This is now working, but it relies on the --experimental-vm-modules NodeJS flag, the use of which generates bunch of warnings when running the tests.

The good part is that (besides having to convert to async-await syntax) no tests needed to change. So this new Prettier 3 seems to format everything exactly as Prettier 2 did.

Perhaps there is a better way though. Should try out @prettier/sync which should eliminate the need to litter all tests with async-await. Maybe it also won't need this NodeJS flag.

[nene]

Experimented with @prettier/sync, but didn't get it working. Getting an error:

    DataCloneError: function (text, options) {
            return (0, transformCst_1.transformCst)((0, sql_parser_cst_1.parse)(text, {
    ...<omitted>... } could not be cloned.

Possibly related to @prettier/sync#4

[karlhorky]

it relies on the --experimental-vm-modules NodeJS flag, the use of which generates bunch of warnings when running the tests

In Node.js v21.3.0+ there is a --disable-warning flag (some signals it may be backported to v18 and v20)

[karlhorky]

Alternative

To avoid having to work with and maintain Prettier-specific code, the prettier-plugin-sql-cst project could change to a generic formatter library (eg. nene/sql-formatter-cst), for consumption by prettier-plugin-sql as it was already tried here (comment hidden and marked as offtopic):

• https://github.com/un-ts/prettier/issues/291#issuecomment-1803811105

And then prettier-plugin-sql could maintain the Prettier-specific parts.

[JounQin]

@prettier/sync is not full functional, you can use https://github.com/un-ts/synckit to wrap its async API directly very easily. See also https://github.com/prettier/prettier-synchronized/issues/2

[nene]

As I wrote in the linked issue, this plugin can't be separated from Prettier. The whole point of this plugin is to make use of Prettier to perform the actual code layout.

[nene]

@karlhorky thanks for the pointers though to other possible workarounds/solutions.

[nene]

@JounQin attempted to use synckit, but this didn't work out for me. Here's a PR #16 if you're interested.

In the end, I think the only win would have been a bit shorter syntax when writing tests and not having to convert all existing tests to async. But I did the latter already. On the downside I'd have more tooling around the tests - another thing that could break.

So I guess I'll live with the async stuff.

[JounQin]

OK, I'll help to find the cause when I'm free.

---

Keep async is good if prettier has already supported.

[JounQin]

@nene I take a look at the code base, --experimental-vm-modules is only required for jest so I don't think that's a big problem right? And this is a limitation by jest itself:

https://jestjs.io/docs/ecmascript-modules

And I suggest to migrate to Vitest instead which supports ESM very well.