console: refactor to use rest params and template strings

[jasnell]

Checklist

• tests and code linting passes
• a test and/or benchmark is included
• the commit message follows commit guidelines

Affected core subsystem(s)

console

Description of change

Overall cleanup in code, eliminate reliance on arguments.

Benchmarks show that as of v8 5.0.71.32, using rest params + apply has good performance. The spread operator is not yet well optimized in v8

misc/console.js method=restAndSpread concat=1 n=1000000: 374779.38359
misc/console.js method=restAndSpread concat=0 n=1000000: 375988.30434
misc/console.js method=argumentsAndApply concat=1 n=1000000: 682618.61125
misc/console.js method=argumentsAndApply concat=0 n=1000000: 645093.74443
misc/console.js method=restAndApply concat=1 n=1000000: 682931.41217
misc/console.js method=restAndApply concat=0 n=1000000: 664473.09700

[mscdex]

It would be nice to see manual string concatenation added to the benchmarks. I suspect that would be the "fastest" (at least compared to template strings the last I checked).

[jasnell]

It's there.. see https://github.com/nodejs/node/pull/6233/files#diff-32494b677ab738804f3e70d72fa804adR31 .. specifically, the concat=1 config above means doing manual concat whereas concat=0 means use template strings

[jasnell]

And you're right, they edge out template strings just a bit... but not by much.

[mscdex]

@jasnell There is only benchmarking of util.format() and template strings currently. I meant also adding benchmarks for testing raw concatenation like 'this is ' + arguments[0] + ' of ' + arguments[1].

[jasnell]

oh! misunderstood... I was referring to the concat in util.format(...args) + '\n'.

[Fishrock123]

It may be good to leave these (but not the benchmarks) for new contributors to deal with, since it's pretty easy to instruct people to change stuff to use es6.

[jasnell]

Why? I'm working in code already anyway (#6174). Is there a reason not to land this PR?

[Fishrock123]

Why?

since it's pretty easy to instruct people to change stuff to use es6.

Is there a reason not to land this PR?

I meant the type of changes, as exciting as it may be for us. I haven't checked this PR, but so long as perf is good we should be fine I suppose.

[jasnell]

@mscdex ... updated the benchmark to include a rest params + concat only example and you're right, it is the fastest approach, but not by a whole lot. Using rest params with apply, util.format and concat for the \n seems to be a reasonable choice here.

misc/console.js method=restAndSpread concat=1 n=1000000: 431909.61616
misc/console.js method=restAndSpread concat=0 n=1000000: 380620.89756
misc/console.js method=argumentsAndApply concat=1 n=1000000: 695168.75332
misc/console.js method=argumentsAndApply concat=0 n=1000000: 720588.97836
misc/console.js method=restAndApply concat=1 n=1000000: 726704.47661
misc/console.js method=restAndApply concat=0 n=1000000: 725573.25529
misc/console.js method=restAndConcat concat=1 n=1000000: 783515.12632

[jasnell]

@mscdex ... how does this look to you now?

[mscdex]

What is the reason for using a stream in the benchmarks?

[jasnell]

@mscdex ... because I didn't want to write out to console but still wanted to keep the scenario as close to console.log and console.warn as possible.

[mscdex]

LGTM

[jasnell]

Landed in 33c242e

[Trott]

Looks like this landed without running through CI. There's a linting error. Fix is in #6315

[MylesBorins]

already fixed in cff2a13

[jasnell]

It appears I have a bug in my landing script :-/ ... sometimes it runs lint, other times it doesn't. Not quite sure why. Ah well, that's what I get for making the landing script overly complicated ;-) ... looks like I'm back to manual linting

[jasnell]

@Trott ... what would you think about a git pre-commit hook that enforces linting before commit? Obviously folks would have to install it locally to use it but having it available in tree would be worthwhile

[Trott]

@jasnell Not sure. People are supposed to run make test before submitting a PR and that will run the linting, so... ¯_(ツ)_/¯

On the other hand, tools over rules and all that.

[jasnell]

heh... well, I pulled the faulty make lint stuff out of my landing script and went back to the tried and true pre-push git hook. That'll teach me to get fancy with it.