antzucaro / matchr Goto Github PK

Thanks for your great work! 💯

In my opinion, BMPM2 is by far the widely best phonetic matcher on earth:

https://stevemorse.org/phonetics/bmpm2.htm

It is well organized and easy to implement:

https://stevemorse.org/phoneticinfo.htm

Please have a look and consider porting from PHP to Go 🥇

matchr.JaroWinkler("dr", "driveway", true)
--
panic: runtime error: index out of range
github.com/antzucaro/matchr.jaroWinklerBase(0xc82056cf71, 0x2, 0xc82056cfd1, 0x8, 0x101, 0x3fe8000000000000)
    /home/vagrant/workspace/go/src/github.com/antzucaro/matchr/jarowinkler.go:100 +0x59d
github.com/antzucaro/matchr.JaroWinkler(0xc82056cf71, 0x2, 0xc82056cfd1, 0x8, 0xb69801, 0xc8202a8418)
    /home/vagrant/workspace/go/src/github.com/antzucaro/matchr/jarowinkler.go:134 +0x50

As a hotfix, I've changed line :100 to explicitly checking r1 & r2 have an index at i like this:
for i = 0; i < j && len(r1) > i && len(r2) > i && r1[i] == r2[i] && nan(r1[i]); i++ {

Hi there! I was wondering why there is a hardcoded max length of 4 for double metaphone? (I've also noticed other implementations limit this to 4). Is there a particular design decision behind this?

Hi!
Could you create a release?

Have you chosen a license for this project? I'd love to use your Smith-Waterman implementation in a GPL-licensed bioinformatics project of mine:

https://github.com/plsql/jh-bio

Let me know.

Thanks!

If NYSIIS function receives a string with only numbers or numbers and symbols it will panic. The function should probably return an empty string instead (""). Line 24 generates a panic because the input is empty if the string is numbers or symbols only.

I have a fix with tests ready for this in a local branch if you want. Or you can use this code right about line 24.

	// if no characters are left return blank
	if len(input) == 0 {
		return ""
	}

Test cases:

	{"2002", ""},
	{"1/2", ""},
	{"", ""},

I don't know if you're interested, but there's a much easier way of doing Smith-Waterman that involves having an extra row and column at the 0th indexes, whose scores are all zero. You only need a single body loop:

https://github.com/plsql/jh-bio/blob/unique-kmers/bioutils/alignment.go#L63

For example, comparing "heh" and "hhh" will use the below matrix:

0.0  0.0  0.0  0.0  

0.0  1.0  1.0  1.0  

0.0  0.5  0.5  0.5  

0.0  1.0  1.5  1.5 

Hi.

I have tried this library and compared it with https://github.com/adrg

In some cases we experience differences in the results:

package main

import (
	"fmt"

	"github.com/adrg/strutil/metrics"
	"github.com/antzucaro/matchr"
)

func main() {
	r2 := "wilson kjell"
	r1 := "wilson mathias"
	fmt.Printf("matchr long distance:%f\n", matchr.JaroWinkler(r1, r2, true))
	fmt.Printf("matchr short distance:%f\n", matchr.JaroWinkler(r1, r2, false))

	m := metrics.NewJaroWinkler()
	fmt.Printf("adrg:%f\n", m.Compare(r2, r1))
}
// matchr long distance:0.694444
// matchr short distance:0.694444
// adrg:0.816667

https://go.dev/play/p/z2IQsqYjIDQ

What is correct distance between these strings?

The origninal implementation (strcmp95) called from perl gives us 0.83523

Thank you.

In util.go, the function charAt is never used and can be removed completely. However, if you intend to use it, you may want to change it.

Currently, if the provided index is out of bounds, the function returns 0. However Go, unlike C, allows null characters (\u0000) in strings. It's therefore impossible to distinguish between an out-of-bound index and a null character.

You could fix this by returning -1 instead, since the rune type is just an alias for int32.