benchmark: replace [].join() with ''.repeat()

[vsemozhetbyt]

Checklist

• tests and/or benchmarks are included
• commit message follows commit guidelines

Affected core subsystem(s)

benchmark

This change is mostly for readability as the edited fragments are not performance-wise concerned. However, a benchmark to compare both ways of long string creation is added to be on the safe side.

Results of the added benchmark with the last node.8.0.0-nightly20170402

es\string-repeat.js size=10 encoding="ascii" mode="Array" n=1000: 290,692.60419876396
es\string-repeat.js size=1000 encoding="ascii" mode="Array" n=1000: 102,836.05382685862
es\string-repeat.js size=1000000 encoding="ascii" mode="Array" n=1000: 135.32093280569134
es\string-repeat.js size=10 encoding="utf8" mode="Array" n=1000: 288,349.3871854649
es\string-repeat.js size=1000 encoding="utf8" mode="Array" n=1000: 94,707.65121331392
es\string-repeat.js size=1000000 encoding="utf8" mode="Array" n=1000: 115.63063143181027
es\string-repeat.js size=10 encoding="ascii" mode="repeat" n=1000: 1,551,397.8094262932
es\string-repeat.js size=1000 encoding="ascii" mode="repeat" n=1000: 695,483.8063550529
es\string-repeat.js size=1000000 encoding="ascii" mode="repeat" n=1000: 467,033.2877975878
es\string-repeat.js size=10 encoding="utf8" mode="repeat" n=1000: 1,659,899.8084475622
es\string-repeat.js size=1000 encoding="utf8" mode="repeat" n=1000: 661,734.5385725063
es\string-repeat.js size=1000000 encoding="utf8" mode="repeat" n=1000: 455,820.924368369

[vsemozhetbyt]

Linter CI: https://ci.nodejs.org/job/node-test-linter/7978/

[lpinca]

Why [...str].length instead of just str.length?

[vsemozhetbyt]

It is unicode-proof, it counts code points, not char codes:

> '\ud83d\udc0e'.repeat(10).length
20
> [...'\ud83d\udc0e'.repeat(10)].length
10

[lpinca]

I like this style but I think it's not widely used in the codebase.
Is it better to put the for body in the next line?

Also is it ok to use let in for loops like this now?

[vsemozhetbyt]

I will wrap to conform.

It seems let in loops is mostly optimized now. At least they are rather common in tests or benchmarks.

[vsemozhetbyt]

New linter CI: https://ci.nodejs.org/job/node-test-linter/7990/

[jasnell]

Landed in 74dc3bf

[MylesBorins]

lts?

[gibfahn]

ping @vsemozhetbyt , should this be backported to v6.x?

[vsemozhetbyt]

@gibfahn I am not sure. If we have some +1 I can try to backport though if this does not land cleanly.

[gibfahn]

I suspect that it's not worth trying to backport any more complex performance improvements to v6.x, all the benchmarks will have to be rerun etc.

Having said that, if anyone is willing to do the work, then by all means do backport perf improvements.