String concat speed

[rjacoby]

In the Strings section the standard is:

"Strings longer than 80 characters should be written across multiple lines using string concatenation."

I understand the clarity of code - but the performance is 80-100% worse. jsPerf

The need for such long strings is debatable anyway, but isn't this kind of a steep performance penalty for code style?

[hshoff]

Hi Rafi, thanks for the checking out the style guide and including a jsPerf.

We think that if you are fine-tuning your app and you find that the bottleneck is string concatenation, then by all means move to a single line and leave a comment saying "this was slowing down the app, moved to a single line" kind of thing.

Ideally, we would avoid this mess all together and keep really long strings out of our apps, but sometimes it comes up.

This is one of those cases why we call it a "mostly reasonable approach to javascript", not everything was decided by a jsPerf, we just made a style call and went with clarity over performance.

My gut feeling is there will be many other things to optimize for performance before you get down to long string concatenation, but I haven't come across this problem before.

I'll leave this issue open for now and hopefully some other folks will chime in with an opinion and we can correct it if necessary.

🍻

[reissbaker]

It's a good point that it's slower, but I agree with @hshoff. If you have a rare case where it's a bottleneck, definitely do what it takes to get rid of it (and include a comment!), but in general we aim for readability over raw speed.

[rjacoby]

Totally agree with @hshoff and @reissbaker on this as an less-common bottleneck. Maybe just note that it shouldn't be overused?

Honestly it just popped out at me b/c of general concat performance in other languages (Java, Ruby).

[hshoff]

@rjacoby good call. I'll make a note and add the jsPerf.

[michaelmior]

You really shouldn't have to worry about this in your code and this should be handled by whatever tool you use to minify your JS for production. For example, uglifyjs will turn

var x = "a" + "b";

into

var x="ab";

[bridge-rabish]

I used this method for long string

['<div>',
    '<p>this is test</p>',
'</div>'].join('')

[gustavopaes]

Use join() it's a bad idea.
http://jsperf.com/ya-string-concat/10

[Nemoden]

So, why exactly strings concatenation is preferred by the Guide if it's 95% slower than joining it with the break symbol ()? It doesn't add much readability and extremely slow.

[nkbt]

@Nemoden maybe because of readability. When string concatenation becomes a bottleneck in the app, then it can be easily converted to the fastest way available.

[ghost]

Ok

[yonghu86]

well,it's not length.

[michaelmior]

I think concatenation is much more readable for the sole fact that you can keep the correct level of indentation. As I noted in my previous comment, I think concatenation of static strings should be removed by build tools so performance should not really ever be an issue.

[ibc]

I think concatenation is much more readable for the sole fact that you can keep the correct level of indentation.

Or you can use a proper editor that properly indents extra lines if they belong to the same line.

[ivanjonas]

It seems that the latest versions of Chrome have made performance a moot point. Here's a screenshot of revision 6 results.

[ljharb]

Nowaways it's preferred to use an ES6 template string literal - which transpiles to a long string (with or without breaks, i forget). This entire issue seems like a moot point now :-)

[GrayedFox]

Not quite a moot point.... basically, referencing variables inside of a template literal is still much slower than string concatenation, however without referencing a variable (and simply using a template literal for line breaks) seems to have no performance impact... https://jsperf.com/es6-string-literals-vs-string-concatenation/4

[ljharb]

But also, "much slower" won't make any difference unless you're building millions of strings per second. It's identically fast for most use cases.