async_hooks: remove experimental onPropagate option

[jasnell]

Refs: #46374

The onPropagate option for AsyncLocalStorage is problematic for a couple of reasons:

1. It is not expected to be forwards compatible in any way with the upcoming TC-39 AsyncContext proposal.
2. It introduces a non-trivial O(n) cost invoking a JavaScript callback for every AsyncResource that is created, including every Promise.

While it is still experimental, I recommend removing it while we can revisit the fundamental use cases in light of the coming AsyncContext proposal.

/cc @nodejs/async_hooks

[Qard]

LGTM, though @Flarna may have some thoughts as I believe they're depending on this interface. Not sure how critical it is but they may want a viable alternative proposed before this can be removed.

[vdeturckheim]

lgtm - aligned with @Qard here

[GeoffreyBooth]

I like the idea of aligning our development of this with WinterCG.

I don’t really accept that it’s our responsibility to develop replacements for any users depending on an experimental API. If they want a new feature in Node, they should propose and contribute it.

[jasnell]

If they want a new feature in Node, they should propose and contribute it.

Well, to be fair, that's exactly what @Flarna did :-) ... There's no rush pushing this out. We can give Flarna time to revisit if they want.

[Qard]

To be clear, I'm fine removing it as long as @Flarna doesn't feel it's too disruptive. I'm not aware of anyone else using it.

[Flarna]

2. It introduces a non-trivial O(n) cost invoking a JavaScript callback for every AsyncResource that is created, including every Promise.

Actually the O(n) cost is not introduced by onPropagate. If set each resource will trigger one call. O(n) will only happen if there are n ALS instances using onPropagate.
The O(n) complexity is already there independent of onPropagate because ALS uses the init hook to propagate context internally in all ALS instances.

Anyhow, following TC-39 sounds reasonable. Any hints that TC-39 will accept the AsyncContext proposal anytime soon? As far as I know the proposal is a few years old meanwhile and has still not reached stage 1. Or is there something ongoing which is not visible on the TC-39 propsals github repo?

[jasnell]

Anyhow, following TC-39 sounds reasonable. Any hints that TC-39 will accept the AsyncContext proposal anytime soon? As far as I know the proposal is a few years old meanwhile and has still not reached stage 1. Or is there something ongoing which is not visible on the TC-39 propsals github repo?

Work is underway now. With @jridgewell, @legendecas, and @littledan heading it up.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/49267/

[jridgewell]

Anyhow, following TC-39 sounds reasonable. Any hints that TC-39 will accept the AsyncContext proposal anytime soon?

We'll have a better idea of how long it will take after the TC39 meeting this week.

For now, I think it's better to think of this as aligning with the implementation strategy that AsyncContext will take, which will make everything considerably faster. It's possible that V8 will add a special hook for Node to control how the propagation works (they same way they've added support for async_hooks), but that has the potential to impact the performance guarantees that AsyncContext requires to be part of the core language.

As far as I know the proposal is a few years old meanwhile and has still not reached stage 1. Or is there something ongoing which is not visible on the TC-39 propsals github repo?

I've been meeting with delegates for a few weeks getting feedback on the proposal. We've gotten approval from the security folks, which was a major blocker the last time this was presented. We also have a private matrix channel to discuss that will become public when we get accepted to Stage 1.

[legendecas]

The O(n) complexity is already there independent of onPropagate because ALS uses the init hook to propagate context internally in all ALS instances.

Value propagation can be optimized to avoid the complexity with the number of AsyncLocalStorage instances, as the propagation doesn't calling into user code. See https://github.com/legendecas/proposal-async-context/blob/master/src/index.ts#L7 for an example.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/49270/

[nodejs-github-bot]

Commit Queue failed

- Loading data for nodejs/node/pull/46386
✔ Done loading data for nodejs/node/pull/46386
----------------------------------- PR info ------------------------------------
Title async_hooks: remove experimental onPropagate option (#46386)
Author James M Snell (@jasnell)
Branch jasnell:remove-asynclocalstorage-onpropagate -> nodejs:main
Labels async_hooks, author ready
Commits 1
- async_hooks: remove experimental onPropagate option
Committers 1
- James M Snell PR-URL: https://github.com/nodejs/node/pull/46386
Refs: https://github.com/nodejs/node/issues/46374
Reviewed-By: Stephen Belanger Reviewed-By: Vladimir de Turckheim Reviewed-By: Geoffrey Booth Reviewed-By: Benjamin Gruenbaum Reviewed-By: Gerhard Stöbich Reviewed-By: Chengzhong Wu ------------------------------ Generated metadata ------------------------------
PR-URL: https://github.com/nodejs/node/pull/46386
Refs: https://github.com/nodejs/node/issues/46374
Reviewed-By: Stephen Belanger Reviewed-By: Vladimir de Turckheim Reviewed-By: Geoffrey Booth Reviewed-By: Benjamin Gruenbaum Reviewed-By: Gerhard Stöbich Reviewed-By: Chengzhong Wu --------------------------------------------------------------------------------
ℹ This PR was created on Fri, 27 Jan 2023 22:08:13 GMT
✔ Approvals: 6
✔ - Stephen Belanger (@Qard): https://github.com/nodejs/node/pull/46386#pullrequestreview-1273599442
✔ - Vladimir de Turckheim (@vdeturckheim): https://github.com/nodejs/node/pull/46386#pullrequestreview-1273608580
✔ - Geoffrey Booth (@GeoffreyBooth) (TSC): https://github.com/nodejs/node/pull/46386#pullrequestreview-1273724192
✔ - Benjamin Gruenbaum (@benjamingr): https://github.com/nodejs/node/pull/46386#pullrequestreview-1273975929
✔ - Gerhard Stöbich (@Flarna): https://github.com/nodejs/node/pull/46386#pullrequestreview-1276087836
✔ - Chengzhong Wu (@legendecas) (TSC): https://github.com/nodejs/node/pull/46386#pullrequestreview-1276289905
✔ Last GitHub CI successful
ℹ Last Full PR CI on 2023-01-31T02:24:58Z: https://ci.nodejs.org/job/node-test-pull-request/49270/
- Querying data for job/node-test-pull-request/49270/
✔ Last Jenkins CI successful
--------------------------------------------------------------------------------
✔ No git cherry-pick in progress
✔ No git am in progress
✔ No git rebase in progress
--------------------------------------------------------------------------------
- Bringing origin/main up to date...
From https://github.com/nodejs/node
* branch main -> FETCH_HEAD
✔ origin/main is now up-to-date
- Downloading patch for 46386
From https://github.com/nodejs/node
* branch refs/pull/46386/merge -> FETCH_HEAD
✔ Fetched commits as 088e470dcdaa..3c265fa0aace
--------------------------------------------------------------------------------
[main 60f40eb7a8] async_hooks: remove experimental onPropagate option
Author: James M Snell Date: Fri Jan 27 14:03:51 2023 -0800
3 files changed, 6 insertions(+), 79 deletions(-)
delete mode 100644 test/async-hooks/test-async-local-storage-stop-propagation.js
✔ Patches applied
--------------------------------------------------------------------------------
✖ Git found no trailers in the original commit message, but 'Refs: https://github.com/nodejs/node/issues/46374' is present and should be a trailer.

https://github.com/nodejs/node/actions/runs/4065850372

[jasnell]

Landed in dc90810