Case of illogical domain suggestion

[demark]

Hi there,

i found case of illogical domain suggestion:

domains = ["ua.com", "ui.com"], input val is user@uo.com, suggestion: user@ua.com, but if you look at keyboard image - it's obvious that i've mispelled for domain ui.com

I clearly understand that built-in algorithm doesn't solve all cases, but i suggest to think about how to solve similar cases.

[tombouctou]

What about hacking a js implementation of keyboard distance:
http://cpansearch.perl.org/src/KRBURTON/String-KeyboardDistance-1.01/KeyboardDistance.pm ?

[derrickko]

Very interesting, that would actually be really cool (and hopefully more effective). Open to a pull request for it ;)

[Kreker]

var email_checker = function(email) {
//original email
this.email=email;
}
email_checker.prototype = {
email : false, //user email
potential_servers : [], //server list for incorrect email
username : false, //email username
//keyboard list
keyboard : [
['q', 'w', 'e', 'r', 't', 'y', 'u', 'i', 'o', 'p'],
['a', 's', 'd', 'f', 'g', 'h', 'j', 'k', 'l'],
['z', 'x', 'c', 'v', 'b', 'n', 'm']
],
//post servername list
get_server_names : {"yahoo":1, "google":1, "hotmail":1, "gmail":1, "me":1, "aol":1, "mac":1,
"live":1, "comcast":1, "googlemail":1, "msn":1,
"facebook":1, "verizon":1, "sbcglobal":1, "att":1, "gmx":1, "mail":1, "ymail":1, "ya":1, "yandex":1, "vk":1, "rambler":1, "list":1, "inbox":1, "tut":1},

/*
* Return email servername and save email username
* @return string
*/
get_server_from_email : function(email) {
    var parts = email.split('@');
    this.username = parts[0];
    var parts = parts[1].split('.');
    return parts[0];
},

/*
* Add potential server if servername matched servername with replaced letter 
* @param int key  -  letter-key in original email
* @param int current_row  -  row on the keyboard to be searched
* @param int keypos  -  keypos on the keyboard to be searched
* @param array server_name  -  splitted servername
* @return void
*/
server_checkout : function(key, current_row, keypos, server_name) {

    var new_letter = current_row[keypos];
    if (typeof old_letter != 'undefined')

    if (new_letter) {
        old_letter = server_name[key];
        server_name[key] = new_letter; //replace letter

        new_srv_name = server_name.join('');

        if (this.get_server_names[new_srv_name]) //server found!
            this.potential_servers.push(new_srv_name);

        server_name[key] = old_letter;
    }
},

/*
* Check email post server
* @return bool/object
*/
check :  function() {
    var server = this.get_server_from_email(this.email);

    if (this.get_server_names[server]) 
        return true;


    variants = [];
    //search variants
    var server_spl = server.split('')

    for (key = 0; key < server_spl.length; key++) { //checking every symbol

        //search keypos
        keypos = {row:0,pos:0};
        for (row in this.keyboard) {
            for (col in this.keyboard[row]) {
                if (this.keyboard[row][col] == server_spl[key]) {
                    keypos = {row:row,pos:col};
                }
            }   
        }
        //adjacent rows diapason
        var start = (keypos['row'] - 1 >= 0) ? keypos['row'] - 1 : 0;
        var end = (keypos['row'] + 1 <= 2) ? Number(keypos['row']) + 1 : 2;

        for (cur_row = start; cur_row <= end; cur_row++) { //search in adjacent rows
            //before key
            this.server_checkout(key, this.keyboard[cur_row], (keypos['pos'] - 1), server_spl);
            //current key in other row
            if (keypos['row'] != cur_row)
                this.server_checkout(key, this.keyboard[cur_row], keypos['pos'], server_spl);

            //after key
            this.server_checkout(key, this.keyboard[cur_row], (Number(keypos['pos']) + 1), server_spl);

        }
    }

    return this.potential_servers;
}

}
var res = new email_checker('kreker20@hmail.com').check();
if (res === true)
console.log('Mailserver correct');
else
console.log('Maybe '+res.join(' or '));

[Kreker]

fck parser.
expect suggestions

[hpshelton]

The issue here is that the implementation doesn't allow for two domains to be the same distance apart:

dist = this.stringDistance(domain, domains[i]);
if (dist < minDist) {
    minDist = dist;
    closestDomain = domains[i];
}

Add console.log(dist + ' ' + domains[i]); in that section and you will see that the distance calculated for both "ua.com" and "ui.com" is 2, but the algorithm returns only the first suggestion. Reverse the two domains in the list and the correct suggestion is returned.

Two options I see are that the search could be extended to be "<=" and to return an array of multiple equidistant domains to the implementing application to choose amongst, or the keyboard distance can be used to return the closest if the first search returns multiple equidistant domains.

[hpshelton]

An initial implementation is in one of my branches based on the Perl implementation referenced above. Mailcheck will fall back to the keyboard distance if the first pass returns multiple equidistant domains.

I'll make a pull request after some touch-up and more testing.

[derrickko]

@hpshelton That's awesome. Really looking forward to it.