String.substring vs String[].split

Question

I have a comma delaminated string that when calling String.split(",") it returns an array size of about 60. In a specific use case I only need to get the value of the second value that would be returned from the array. So for example "Q,BAC,233,sdf,sdf," all I want is the value of the string after the first ',' and before the second ','. The question I have for performance am I better off parsing it myself using substring or using the split method and then get the second value in the array? Any input would be appreciated. This method will get called hundreds of times a second so it's important I understand the best approach regarding performance and memory allocation.

-Duncan

The problem with my profiler is that I don't know where all the char[] arrays I see in the profiler come from. What is the use case to profile this and get metrics? — Duncan Krebs
– Duncan Krebs, Commented Dec 21, 2012 at 21:25
I didn't mean profile the number of char arrays created. Profile the time it takes to execute a somewhat realistic benchmark, before and after the "optimization", to see if it makes any noticeable difference. — user395760
– user395760, Commented Dec 21, 2012 at 21:27

Sergey Kalinichenko · Accepted Answer · 2012-12-21 21:36:49Z

43

Since String.Split returns a string[], using a 60-way Split would result in about sixty needless allocations per line. Split goes through your entire string, and creates sixty new object plus the array object itself. Of these sixty one objects you keep exactly one, and let garbage collector deal with the remaining sixty.

If you are calling this in a tight loop, a substring would definitely be more efficient: it goes through the portion of your string up to the second comma ,, and then creates one new object that you keep.

String s = "quick,brown,fox,jumps,over,the,lazy,dog";
int from = s.indexOf(',');
int to = s.indexOf(',', from+1);
String brown = s.substring(from+1, to);

The above prints brown

When you run this multiple times, the substring wins on time hands down: 1,000,000 iterations of split take 3.36s, while 1,000,000 iterations of substring take only 0.05s. And that's with only eight components in the string! The difference for sixty components would be even more drastic.

edited Dec 21, 2012 at 21:36

answered Dec 21, 2012 at 21:24

Sergey Kalinichenko

729k85 gold badges1.2k silver badges1.6k bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

user395760 Over a year ago

That is, of course, assuming it's actually performance critical and you couldn't achieve more speed up in other ways. Programmers have a tendency to make wildly inaccurate guesses about that.

Duncan Krebs Over a year ago

I appreciate all the answers. Seems like the theme is substring and you explained it the best.

Mr. Polywhirl Over a year ago

I wrote a method to retrieve a token at a desired index, pastebin.com/R9Z6uW6H

Jigar Joshi · Accepted Answer · 2012-12-21 21:23:46Z

4

ofcourse why iterate through whole string, just use substring() and indexOf()

answered Dec 21, 2012 at 21:23

Jigar Joshi

241k42 gold badges409 silver badges446 bronze badges

1 Comment

user395760 Over a year ago

The possibility of off-by-one errors? The increased amount of less obvious code?

fge · Accepted Answer · 2012-12-21 21:25:29Z

3

You are certainly better off doing it by hand for two reasons:

.split() takes a string as an argument, but this string is interpreted as a Pattern, and for your use case Pattern is costly;
as you say, you only need the second element: the algorithm to grab that second element is simple enough to do by hand.

answered Dec 21, 2012 at 21:25

fge

122k35 gold badges266 silver badges340 bronze badges

Comments

MrSmith42 · Accepted Answer · 2012-12-21 21:28:06Z

2

I would use something like:

final int first = searchString.indexOf(",");
final int second = searchString.indexOf(",", first+1);
String result= searchString.substring(first+1, second);

answered Dec 21, 2012 at 21:28

MrSmith42

10.2k6 gold badges41 silver badges52 bronze badges

3 Comments

Duncan Krebs Over a year ago

Thanks, what is the purpose of declaring the index values as final?

MrSmith42 Over a year ago

It's just a code convention I am used to. I make all variables final which are only assigned once.

Duncan Krebs Over a year ago

Thank I you upvoted! Please upvote my question I want to break 600!

Justin Niessner · Accepted Answer · 2012-12-21 21:24:28Z

1

My first inclination would be to find the index of the first and second commas and take the substring.

The only real way to tell for sure, though, is to test each in your particular scenario. Break out the appropriate stopwatch and measure the two.

answered Dec 21, 2012 at 21:24

Justin Niessner

246k40 gold badges416 silver badges546 bronze badges

Collectives™ on Stack Overflow

String.substring vs String[].split

5 Answers 5

3 Comments

1 Comment

Comments

3 Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

5 Answers 5

3 Comments

1 Comment

Comments

3 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related