Checking for empty string performance tests

Search the Internet on .NET performance and you'll be sure to find a large number of String manipulation tips. People use various criteria to give a recommendation such as speed, ease of implementation, coolness and personal preference so you will often find contradictory advice. In my opinion being informed is important so I devised a series of tests that compare various operations. Once you see the numbers your can use your own judgement to form an opinion.

In this articles I will put to test two methods of testing for an empty string:

• Compare string with String.Empty.
• Check string length. If the length is 0 then the string is empty.

I will run my test application on .NET v1.1.4322, .NET v2.0.50727 and then on Mono v1.2.2.1 on the same machine with the various frameworks deployed side by side.

Test application:

Each set of tests start with testing the String.Empty constant. After that progressively longer random strings are generated starting from 0 to 10000 and the test is executed for each one.

The operation I am benchmarking is very short so I will run the same test 10,000,000 times to get a clearly visible difference between the various platforms.

Test results:

Please note that the exact test results will depend on your computer performance.

String Length Length test Comparison test 0 (String.Empty) 0.0781250 0.4062500 0 0.0781250 2.2343750 1000 0.0781250 2.4375000 2000 0.1093750 2.5937500 3000 0.1093750 2.7031250 4000 0.1250000 2.8437500 5000 0.1406250 2.9843750 6000 0.1250000 3.0468750 7000 0.1406250 3.0937500 8000 0.1406250 3.2968750 9000 0.1562500 3.6875000 10000 0.1875000 3.3593750

Table 1 String comparison speed in .NET v1.1.4322 - test is executed 10,000,000 times.

String Length Length test Comparison test 0 (String.Empty) 0.3281250 0.7968750 0 0.3437500 2.2343750 1000 0.3750000 1.7500000 2000 0.3593750 1.7968750 3000 0.3593750 1.7500000 4000 0.3437500 1.7031250 5000 0.4062500 1.8281250 6000 0.3437500 1.8437500 7000 0.3750000 1.7187500 8000 0.3593750 1.7812500 9000 0.3593750 1.7500000 10000 0.3593750 1.7500000

Table 2 String comparison speed in .NET v2.0.50727 - test is executed 10,000,000 times.

String Length Length test Comparison test 0 (String.Empty) 0.2970000 1.9840000 0 0.4220000 2.0620000 1000 0.4220000 2.2660000 2000 0.4210000 2.3120000 3000 0.4370000 2.3130000 4000 0.4220000 2.3280000 5000 0.4380000 2.3590000 6000 0.4380000 2.3290000 7000 0.4380000 2.3440000 8000 0.5000000 2.3280000 9000 0.4370000 2.3600000 10000 0.4530000 2.3900000

Table 3 String comparison speed in Mono v1.2.2.1 - test is executed 10,000,000 times.

Here's what it looks like if we plot a graph with the data.

Figure 1 String comparison speed in .NET v1.1.4322 - test is executed 10,000,000 times.

Figure 2 String comparison speed in .NET v2.0.50727 - test is executed 10,000,000 times.

Figure 3 String comparison speed in Mono v1.2.2.1 - test is executed 10,000,000 times.

Cross platform comparisons:

String Length .NET v1.1.4322 .NET v2.0.50727 Mono-1.2.2.1 0 (String.Empty) 0.0781250 0.3281250 0.2970000 0 0.0781250 0.3437500 0.4220000 1000 0.0781250 0.3750000 0.4220000 2000 0.1093750 0.3593750 0.4210000 3000 0.1093750 0.3593750 0.4370000 4000 0.1250000 0.3437500 0.4220000 5000 0.1406250 0.4062500 0.4380000 6000 0.1250000 0.3437500 0.4380000 7000 0.1406250 0.3750000 0.4380000 8000 0.1406250 0.3593750 0.5000000 9000 0.1562500 0.3593750 0.4370000 10000 0.1875000 0.3593750 0.4530000

Table 4 Cross platform string length test speed - test is executed 10,000,000 times.

String Length .NET v1.1.4322 .NET v2.0.50727 Mono-1.2.2.1 0 (String.Empty) 0.4062500 0.7968750 1.9840000 0 2.2343750 2.2343750 2.0620000 1000 2.4375000 1.7500000 2.2660000 2000 2.5937500 1.7968750 2.3120000 3000 2.7031250 1.7500000 2.3130000 4000 2.8437500 1.7031250 2.3280000 5000 2.9843750 1.8281250 2.3590000 6000 3.0468750 1.8437500 2.3290000 7000 3.0937500 1.7187500 2.3440000 8000 3.2968750 1.7812500 2.3280000 9000 3.6875000 1.7500000 2.3600000 10000 3.3593750 1.7500000 2.3900000

Table 5 Cross platform string compare test speed - test is executed 10,000,000 times.

Here's what it looks like if we plot a graph with the data.

Figure 4 Cross platform string length test speed - test is executed 10,000,000 times.

Figure 5 Cross platform string compare test speed - test is executed 10,000,000 times.

Conclusions:

Testing string length is faster than comparing with String.Empty on all platforms.

As expected string length is not a significant factor but using String.Empty will yield better performance for a comparison test in both versions of .NET.

Interestingly .NET v1.1.4322 is faster than .NET v2.0.50727 for length tests but it is slower for compare tests. In fact even Mono-1.2.2.1 is faster than .NET v1.1.4322 for the string compare test.

Download code

Comments: tb@tbiro.com