Hacker News Reader

Comment by ExpiredLink | Hacker News Reader

> Building a string with a series of concatenations like this is extremely inefficient

No. Not in Java since 1999 or so.

Factually incorrect. Creating new strings by concatenation requires reallocation because Strings are immutable in java. Using a Builder is over 700x faster:

Concat code: http://pastebin.com/CcezmJbf

Builder code: http://pastebin.com/Dd6kzcKa

Results, using JRE 7:

> time java concat data.blk Finished. java foo data.blk 170.08s user 1.52s system 102% cpu 2:48.12 total

> time java builder data.blk Finished. java builder data.blk 0.30s user 0.05s system 151% cpu 0.230 total

This is true, javac only replaces individual concatentations with use of a StringBuilder, so foo + bar + baz + "!" is fast, but performing a concatenation in a loop does multiple copies for each iteration. It (or the JIT) could conceivably do escape analysis to optimize simple cases like this, but it doesn't.

builder: http://hastebin.com/qadusoriqa.cs

concat: http://hastebin.com/nasukuxemi.cs

What about the time to write the program?

About the same.

In special cases multiple string concatenations can be optimized into using a StringBuilder calls. Not in the general case.

I'd say in the general case multiple string concatenations can be optimized. Not in special cases.

I built a string this way in java just 5 years ago. Still made it 100x times faster by using a StringBuilder.

So like 1ms down from 100ms? Either way a rounding error in the episode-of-LOST-average-duration timescale ;)

Why don't you give it a try yourself. Create a file with a high ratio of lines to length, and read it in using a loop and string concat, or builder. Do it with a builder first, when you get up to 1 minute, switch over to concat and see how long it takes.

Reply on news.ycombinator.com