getNodesByType

/*global document*/
//a function to get DOM nodes by nodeType property.
//If you do not supply a value I will give you every DOM node.
//
//example:
// var allComments = document.getNodesByType(8);
// or
// var allComments = document.getNodesByType("COMMENT_NODE");
//
//The accepted string values are the actual node type names, so that the
//typeValue argument can be supplied dynamically from other code.
//
//
// Please try http://prettydiff.com/ for all your web development needs!
//
//
//Keep in mind that the following node types are valid in the W3C DOM
//standards, but have been deprecated in the WHATWG DOM specification.
//
// 2 - ATTRIBUTE_NODE
// 4 - CDATA_SECTION_NODE
// 5 - ENTITY_REFERENCE_NODE
// 6 - ENTITY_NODE
// 12 - NOTATION_NODE
//
//This means all node types are still valid in the standard, but the
//deprecated types may not be retrievable from certain DOM
//implementations.
//
document.getNodesByType = function (typeValue) {
    "use strict";
    var types = 0,
        valueTest = (typeof typeValue === "string") ? typeValue.toUpperCase() : "";
    switch (typeValue) {
    case "ALL":
        types = 0;
        break;
    case "ELEMENT_NODE":
        types = 1;
        break;
    case "ATTRIBUTE_NODE":
        types = 2;
        break;
    case "TEXT_NODE":
        types = 3;
        break;
    case "CDATA_SECTION_NODE":
        types = 4;
        break;
    case "ENTITY_REFERENCE_NODE":
        types = 5;
        break;
    case "ENTITY_NODE":
        types = 6;
        break;
    case "PROCESSING_INSTRUCTION_NODE":
        types = 7;
        break;
    case "COMMENT_NODE":
        types = 8;
        break;
    case "DOCUMENT_NODE":
        types = 9;
        break;
    case "DOCUMENT_TYPE_NODE":
        types = 10;
        break;
    case "DOCUMENT_FRAGMENT_NODE":
        types = 11;
        break;
    case "NOTATION_NODE":
        types = 12;
        break;
    }
    if (isNaN(valueTest) === false && (valueTest.length === 1 || valueTest === "10" || valueTest === "11" || valueTest === "12")) {
        types = Number(valueTest);
    }
    if (valueTest === "" && (typeValue === 0 || typeValue === 1 || typeValue === 2 || typeValue === 3 || typeValue === 4 || typeValue === 5 || typeValue === 6 || typeValue === 7 || typeValue === 8 || typeValue === 9 || typeValue === 10 || typeValue === 11 || typeValue === 12)) {
        types = typeValue;
    }
    return (function () {
        var root = document.documentElement,
            output = [],
            child = function (x) {
                var a = x.childNodes,
                    b = a.length,
                    c = 0;
                for (c = 0; c < b; c += 1) {
                    if (a[c].nodeType === types || types === 0) {
                        output.push(a[c]);
                    }
                    if (a[c].nodeType === 1) {
                        child(a[c]);
                    }
                }
            };
        if (types === 1 || types === 0) {
            output.push(root);
        }
        child(root);
        return output;
    }());
};

I'm not sure if this is faster or not. Or if it's even different but I figured I would show you the code I made before erasing it.

var nodetype = ["NULL", "ELEMENT", "ATTRIBUTE", "TEXT", "CDATA", "ENTITY_REFERENCE", "ENTITY", "RROCESSING_INSTRUCTION", "COMMENT", "DOCUMENT", "DOCUMENT_TYPE", "DOCUMENT_FRAGMENT", "NOTATION"];

function getchildrennodes() {
    var b = [];
    b = toArray(arguments, 0);
    for (var c = 0; c < b.length; c++) {
        if (b[c].hasChildNodes && b[c].hasChildNodes()) {
            b = b.concat(toArray(b[c].childNodes))
        }
    }
    return b
}
function toArray(d) {
    var c = -1,
        b = d.length,
        a = Array(b);
    while (++c < b) {
        a[c] = d[c]
    }
    return a
};