transform: performance like it matters

[soypat]

Lot of noise in this PR, sorry about that. What matters is that

• I'm using a 256byte table for complement calculation instead of map
• I'm avoiding heap allocations where possible by using a byte slice and unsafe conversion.

// ReverseComplement takes the reverse complement of a sequence.
// ReverseComplement expects ASCII input.
func ReverseComplement(sequence string) string {
    n := len(sequence)
    newSeq := make([]byte, n)
    for i := 0; i < n; i++ {
        newSeq[i] = complementTable[sequence[n-i-1]]
    }
    // This is how strings.Builder works with the String() method. If Mr. Go says it's safe...
    return *(*string)(unsafe.Pointer(&newSeq))
}

[gitpod-io]

[soypat]

Test is failing because I included a test for the complement table and it seems it was doing something I did not expect. 'U's complement is 'A'?

Edit: Updated tests. I learned that Uracil binds with Adenine in RNA sequences :)

[Koeng101]

A few things to change (single letter variables) - mostly interested in benchmarks. How much faster is it?

[TimothyStiles]

I can fix variables and stuff but I'm also interested in benchmarks! Would be happy to start on a real benchmark suite for Poly @soypat.

[soypat]

@Koeng101 @TimothyStiles I'll fix the variable names and add better documentation today- but I'd like to note that i as a generic counter in small scopes has been shown to be as readable or even more readable than a full length name. See Beniamini et al..

It'd be awesome to have a benchmark suite like you say to get a real taste of how performance changes affects those using the library 😄, more than happy to help out however I can. Been looking at refactoring the fasta package 👀.

Benchmarks for 32byte string input (click)

Before

goos: linux
goarch: amd64
pkg: github.com/TimothyStiles/poly/transform
cpu: Intel(R) Core(TM) i5-8265U CPU @ 1.60GHz
BenchmarkReverseComplement-8   	 1639549	       714.7 ns/op	     224 B/op	       3 allocs/op
BenchmarkComplement-8          	 3199111	       383.4 ns/op	      48 B/op	       1 allocs/op
BenchmarkReverse-8             	 4345659	       325.0 ns/op	     176 B/op	       2 allocs/op
PASS

After

BenchmarkReverseComplement-8   	18890924	        60.14 ns/op	      32 B/op	       1 allocs/op
BenchmarkComplement-8          	25396846	        55.45 ns/op	      32 B/op	       1 allocs/op
BenchmarkReverse-8             	19949305	        55.67 ns/op	      32 B/op	       1 allocs/op

[TimothyStiles]

@Koeng101 if complement base can't find a complement for the given nucleotide what should it return? N?