checker.no_import.js

//Thai-Spellcheck-JS : github.com/wishawa/thai-spellcheck-js

//If you don't want to use import, comment this line and 
//import thaiSpellcheckerBackend from './thbrk.js';//thbrk.js is generated by emscripten (the WebAssembly SDK)

var Module;//This will store the WebAssembly Module object used to interact with backend
var breakerObject;//This will store the pointer to the loaded breaker, which is mostly the dictionary
var findBreaksFunctionWrapper;//This is a wrapper we use to call C function to find incorrect words

//The load function returns a Promise that resolves to the actual function that find misspellings
function loadThaiSpellchecker() {
    return new Promise(function(resolve, reject) {
        thaiSpellcheckerBackend().then(function(r) {//a promise that resolves to the WebAssembly Module object
            Module = r;//Save the Module object to a global variable

            findBreaksFunctionWrapper = Module.cwrap('th_brk_wc_find_breaks', 'number', ['number', 'number', 'number', 'number', 'number', 'number']);
            //[ThBrk *brk, const thchar_t *s, int pos[], size_t pos_sz, int inc[], int *inc_ret]
            breakerObject = Module.ccall('th_brk_new', 'number', ['string'], [null]);
            //[const char *dictpath], left empty (null) to use default

            resolve(checkThaiSpelling);
        });
    });
}


//input: a javascript string. ex. "ไข่ใก่ฟองนี้มีขะหนาดไหญ่"
//return: a 2D integer array of n rows and 2 columns, where n is the number of incorrectly spelled words
//the number in the first column is the index of the first character of a misspelled word
//the second number in the column is the index of the first letter of the word after the misspelled word
//ex. [[ 3,  6],
//     [14, 24]]
function checkThaiSpelling(text) {

    const l = text.length;

    const utfText = Module._malloc((l + 1) * 2);//allocate space for text for backend
    Module.stringToUTF16(text, utfText, (l + 1) * 2);//copy text to allocated space as UTF16

    const pos = Module._malloc((l + 1) * 4);//allocate space for word break positions
    const inc = Module._malloc((l + 1) * 4);//allocate space for incorrect word positions

    const incRet = Module._malloc(4);//allocate a single int for number of incorrect words
    const breakCount = findBreaksFunctionWrapper(breakerObject, utfText, pos, l+1, inc, incRet);//call the C (WASM) function
    //pos: indexes of the first character of every word
    //breakCount: number of words == length of pos
    //inc: indexes of the first character of every incorrect word

    Module._free(utfText);

    const incCount = Module.getValue(incRet, 'i32');//get number of incorrect words from allocated int
    //incCount: number of incorrect words == length of inc
    Module._free(incRet);


    var result = Array(incCount).fill(0).map(x => Array(2));//create empty 2D array of incCount rows and 2 columns

    var posIndex = 0;//cursor to where we are in the pos array
    var p;//for temporarily storing value from pos array
    for(var i=0; i<incCount; i++) {
        //copy incorrect word beginning index from C/WASM array to JS array
        result[i][0] = Module.getValue(inc + i*4, 'i32');
        p = l;
        //find the end of the incorrect word by finding the first word that begins after the incorrect word
        while(posIndex < breakCount && (p = Module.getValue(pos + posIndex*4, 'i32')) <= result[i][0]) {
            //p: the index of the first letter of a word
            //we want the lowest p that is higher than result[i][0]
            //TODO: see if binary search can help here
            posIndex++;
        }
        if(p <= result[i][0]) p = l;//if there is no word after the incorrect word, use the text length
        result[i][1] = p;
    }
    Module._free(pos);
    Module._free(inc);
    return result;
}

//export {loadThaiSpellchecker};//export for loading with import / dynamic import