Available Now!

String Concatenation Benchmark

These are the largely unedited results of a quick and dirty benchmark I produced to determine whether string concatenation is faster using the String.concat() method or a StringBuffer/Builder. By inspecting the source code, it is clear that the concat operation will be more efficient for two strings, but it is not obvious that the concat operation is faster for three.

I tested a few different techniques for concatenating strings, these were:

concat

Calling String.concat().

buffer

Using a StringBuffer.

builder

Using a StringBuilder.

buffer (LK)

Using a StringBuffer that had been pre-set to the correct length.

builder (LK)

Using a StringBuilder that had been pre-set to the correct length.

Incidentally, the LK stands for "Length Known". Note that Java's implicit String concatenation operator is not tested because the aim of this benchmark is to assist in identfying the fastest way to implement the operator. To identify the number of strings concatenated by a particular test, a number is appended to the name given above. So "buffer 2" means that two strings were concatenated using a StringBuffer.

Each of the techiques was tested against a collection of strings of variable length with a pre-determined average (some techniques may work better than others on short or long strings). For more details of the implementation, just read the source code.

The test was run under Debian(4.0) Linux(2.6.15.1) on an Intel (P4@3.4Ghz) - all times quoted are in milliseconds. Java was executed without any command-line switches. Over an average of three runs, the results were as given below. The spreadsheet (OpenDocument format) used to record this data and generate the charts is available for download.

It appears (on this preliminary inspection at least) that the concat() method is fastest for combining 3 strings in all circumstances. Suprisingly, string concatenation appears to be slower in Java 1.6 than 1.5. These result may need to be verified.