Benchmarks for 2 string libraries
It was fun reading K&R once again. Out of curiosity i set out to implement a substring search function in go. It wasn’t an easy task for a newbie programmer like me. Due to multiple frustrations i decided to peek at the go standard string library to see how it is done.
The standard strings library has a method called “Contains” that does what i need:
e.g
strings.Contains(str, subs) –> returns bool
However, i noticed that strings.Contains() calls another funciton called Index(): func Index(s, sep string) int
I immediately noticed that this function made the len(s) call 3 times. I curiosly added the following code:
m := len(s)
and replaced all 3 len(s) called with m.
Committed code can be found here: https://github.com/mesb/go/commit/5a56f4ea49b14f2251f64d7e38fbd8f892e13ac1
Thanks to some hospitable people on the #go-nuts IRC channel, i learned to do benchmarks on such experiments.
Here is the test directory: https://github.com/mesb/tests
and below are the benchmark results
Test Results
| ns/op customized string lib | ns/op std string lib |
|---|---|
| 1308 | 1639 |
| 1295 | 1663 |
| 1307 | 1726 |
| 1290 | 1748 |
| 1246 | 1617 |
| 1286 | 1617 |
| 1261 | 1673 |
| 1241 | 1651 |
| 1312 | 1633 |
| 1284 | 1747 |
Average ns/op
| customized lib | standard lib |
|---|---|
| 1283 | 1671.4 |
In effect, the golang len(string) has to be made faster.
