src: cache the result of GetOptions() in JS land by joyeecheung · Pull Request #24091 · nodejs/node

joyeecheung · 2018-11-04T21:05:36Z

Instead of calling into C++ each time we need to check the value
of a command line option, cache the option map in a new
internal/options module for faster access to the values in JS land.

Instead of calling into C++ each time we need to check the value of a command line option, cache the option map in a new `internal/options` module for faster access to the values in JS land.

nodejs-github-bot · 2018-11-04T21:05:37Z

@joyeecheung build started: https://ci.nodejs.org/blue/organizations/jenkins/node-test-pull-request-lite-pipeline/detail/node-test-pull-request-lite-pipeline/1490/pipeline

joyeecheung · 2018-11-04T21:06:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/18331/

addaleax

LGTM with 1 comment

addaleax · 2018-11-04T21:09:48Z

lib/internal/bootstrap/loaders.js

      return id.startsWith('internal/') ||
-          (id === 'worker_threads' &&
-           !internalBinding('options').getOptions('--experimental-worker'));
+          (id === 'worker_threads' && !config.experimentalWorker);


Can we avoid going back to process.config? 😕

There are also two other reference to config in here, I think a better option would probably be to just specialize something for NativeModule here - maybe just directly generate the whole map when the process starts and pass it to the loader bootstrapper here and let NativeModule short-circuit the internal/options implementation. ~~That also opens up the opportunity to emit all the warning for experimental modules in one place here instead of emitting them everywhere across the code base.~~

EDIT: wait, we emit them regardless of the cli options..so never mind I guess

joyeecheung · 2018-11-05T08:20:51Z

Fixed linter and the bootstrap module list test: https://ci.nodejs.org/job/node-test-pull-request/18336/

joyeecheung · 2018-11-05T14:05:11Z

Resume CI: https://ci.nodejs.org/job/node-test-pull-request/18342/

lib/internal/options.js

refack · 2018-11-05T16:09:10Z

Instead of calling into C++ each time

I get that motivation, but couldn't we make it simpler? I'd be happy to share your reasoning.

joyeecheung · 2018-11-05T22:26:43Z

@refack See #24091 (comment) my reasoning is we could just eliminate calling into C++ at all eventually by creating the map before bootstrapping loaders so it will be made available in loafers.js as an argument directly. This PR just moves the calls to a central place that can be manipulated in loaders.js

refack · 2018-11-05T23:24:35Z

Resume: https://ci.nodejs.org/job/node-test-commit/22901/
(previous fail was infra, should be resolved)

joyeecheung · 2018-11-06T00:15:15Z

Just FYI there are others who disagree (that's one of the reasons test/parallel/test-bootstrap-modules.js exists

@refack Just to explore about this point a bit further - I think the reason why test/parallel/test-bootstrap-modules.js exists is to make sure we load less stuff during the bootstrap if they are not really necessary. In this case, the options have to be loaded during the bootstrap no matter what since it is conceptually a dependency of the bootstrap process (whereas things like TCPWrap really are not and they are the main targets that test/parallel/test-bootstrap-modules.js tests against).
In that regard, the length of the module load list is not a very effective measure, as the compilation of a very simple module like options.js that does not depend on too many other modules only incurs a very small overhead (whereas other bigger modules may incur a much bigger overhead by running code synchronously and even invoking bindings that do much more). But again, we don't actually have a better measure to evaluate the "weight" of these native modules..so that remains to be investigated.

refack · 2018-11-06T00:29:40Z

I think the reason why test/parallel/test-bootstrap-modules.js exists is to make sure we load less stuff during the bootstrap if they are not really necessary.

Main reason sure, but it's also been quoted as a braier for adding new (trivial) modules - #23081 (review)

Again, I'm slightly in favor... Encapsulation/Performance is just a balance we should keep in mind.

refack · 2018-11-06T00:30:53Z

ubuntu-on-arm had a flake, so resume again: https://ci.nodejs.org/job/node-test-commit/22902/

Instead of calling into C++ each time we need to check the value of a command line option, cache the option map in a new `internal/options` module for faster access to the values in JS land. PR-URL: nodejs#24091 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Gus Caplan <me@gus.host> Reviewed-By: Refael Ackermann <refack@gmail.com>

Trott · 2018-11-08T04:41:07Z

Landed in f895b5a

Instead of calling into C++ each time we need to check the value of a command line option, cache the option map in a new `internal/options` module for faster access to the values in JS land. PR-URL: #24091 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Gus Caplan <me@gus.host> Reviewed-By: Refael Ackermann <refack@gmail.com>

Instead of calling into C++ each time we need to check the value of a command line option, cache the option map in a new `internal/options` module for faster access to the values in JS land. PR-URL: nodejs#24091 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Gus Caplan <me@gus.host> Reviewed-By: Refael Ackermann <refack@gmail.com>

Instead of calling into C++ each time we need to check the value of a command line option, cache the option map in a new `internal/options` module for faster access to the values in JS land. PR-URL: #24091 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Gus Caplan <me@gus.host> Reviewed-By: Refael Ackermann <refack@gmail.com>

src: cache the result of GetOptions() in JS land

22446f1

Instead of calling into C++ each time we need to check the value of a command line option, cache the option map in a new `internal/options` module for faster access to the values in JS land.

joyeecheung requested a review from addaleax November 4, 2018 21:05

nodejs-github-bot added the lib / src Issues and PRs related to general changes in the lib or src directory. label Nov 4, 2018

addaleax approved these changes Nov 4, 2018

View reviewed changes

fixup! src: cache the result of GetOptions() in JS land

977cb18

cjihrig approved these changes Nov 5, 2018

View reviewed changes

refack reviewed Nov 5, 2018

View reviewed changes

lib/internal/options.js Show resolved Hide resolved

refack added process Issues and PRs related to the process subsystem. cli Issues and PRs related to the Node.js command line interface. labels Nov 5, 2018

devsnek approved these changes Nov 5, 2018

View reviewed changes

refack approved these changes Nov 5, 2018

View reviewed changes

refack added the author ready PRs that have at least one approval, no pending requests for changes, and a CI started. label Nov 5, 2018

Trott closed this Nov 8, 2018

BridgeAR mentioned this pull request Nov 14, 2018

Release proposal: v11.2 #24350

Merged

beevelop mentioned this pull request Nov 15, 2018

2018-11-15 Version 11.2.0 (Current) @BridgeAR beevelop/docker-nodejs#232

Closed

codebytere added the backported-to-v10.x label Dec 13, 2018

MylesBorins mentioned this pull request Dec 22, 2018

v10.15.0 proposal #25176

Merged

This was referenced Dec 26, 2018

2018-12-26 Version 10.15.0 'Dubnium' (LTS) @MylesBorins beevelop/docker-nodejs#254

Closed

2018-12-26 Version 10.15.0 'Dubnium' (LTS) @MylesBorins beevelop/docker-android-nodejs#276

Closed

Uh oh!

Conversation

joyeecheung commented Nov 4, 2018

Uh oh!

nodejs-github-bot commented Nov 4, 2018

Uh oh!

joyeecheung commented Nov 4, 2018

Uh oh!

addaleax left a comment

Choose a reason for hiding this comment

Uh oh!

addaleax Nov 4, 2018

Choose a reason for hiding this comment

Uh oh!

joyeecheung Nov 4, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

joyeecheung commented Nov 5, 2018

Uh oh!

joyeecheung commented Nov 5, 2018

Uh oh!

Uh oh!

refack commented Nov 5, 2018

Uh oh!

joyeecheung commented Nov 5, 2018

Uh oh!

refack commented Nov 5, 2018

Uh oh!

joyeecheung commented Nov 6, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

refack commented Nov 6, 2018

Uh oh!

refack commented Nov 6, 2018

Uh oh!

Trott commented Nov 8, 2018

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

joyeecheung Nov 4, 2018 •

edited

Loading

joyeecheung commented Nov 6, 2018 •

edited

Loading