Article on DontEnum and OOJS

D

dhtmlkitchen

I have an article online, but it's not officially linked yet. This
will be the first public link.

I've found tons of errors myself, and there's probably plenty more --
I just didn't find them yet.

I'd really like to get some review on this. Both technical and
editorial, but mostly technical.

The code examples are all plain green. I'll beautify them later (i do
this in my editor w/a regexp find/rep)

http://dhtmlkitchen.com/learn/js/enumeration/
=========================++++++++++++++++++++
+==========================

Enumeration and Object Oriented JavaScript

Who is this Article For?
This article is primarily intended for JavaScript library authors,
language implementors, language contributors.

Anyone who reads this article thoroughly will gain a deeper
understanding of how enumeration works in JavaScript, and how to deal
with language complications and cross browser issues.

For the TG1 group (who is writing ES4), my recommended changes from
ES3 to are notated as struck out and highlighted. My recommendations
reflect what I think should change from ECMAScript rev 3 to ECMAScript
rev 4.
Synopsis

This article is about the for in statement, how it is broken in IE,
and how to work around the problem.

Prototype Chain
The for in loop works with the prototype chain. In order to understand
how the for in loop really works, it is necessary to thoroughly
understand the prototype chain.

Constructor chaining affects the prototype chain, which affects
enumeration.
Borrowing Properties

Borrowing properties is taking properties of a supplier object and
using those properties as properties to receiver object. This is
usually done using a for in loop. Some examples demonsotrate
complications with the prototype chain and the problems caused by an
IE bug that I call the JScript DontEnum bug.

Borrowing properties is related to constructor chaining. Both can be
used as a solution to the problem of adding new behavior to a class.
The two are used sometimes interchangeably, and sometimes together.

Examples
I have included some real world code (YUI) and some examples of my
own. Most of the examples are abbreviated and to the point. The goal
is to explain JavaScript and programming techniques, which can be used
to create a real program. The examples display an understanding how
the prototype chain and enumeration come into play in real-world code.

This article is quite long. The top level sections should be read all
the way through, with the exception of the explanation of Creating a
Subclass, and Borrow methods using an Array. These sections don't have
any dependencies.

You can save paper by printing at a smaller scale (60%), and double
sided (if your printer supports that). Or, you can just read it on
your computer (I bet you knew that).

Tests
There are many live tests that you can and should run. The tests use
eval to run the textContent of the pre element in each example.

The tests show three things:

* How JavaScript works
* Browser bugs
* Issues with JavaScript

If a test does not have a browser table, it is because the test is
known to pass in every browser. If you encounter such test does not
pass, please email me.

If there are any false statements, or if you get a different result
than what is listed in a browser table, please send an email to me.

Browsers tested:
* Internet Explorer 7 (Windows)
* Mozilla Firefox 2 (Mac, Windows)
* Opera 9.2 (Mac, Windows)
* Safari 2 (Mac)
* Safari 3 (Windows, Webkit on Mac)

Many of the examples use String.nl

String.nl = /*@cc_on!@*/false ? "\r\n" : "\n";

Internet Explorer handles whitespace in unexpected ways. This caused
several problems with the examples in this article that use innerText
and innerHTML with the pre element (other elements could not be
supported). The odd line of code above is a workaround.

Each test is inside a pre. The test code is run in a closure to keep
the global namespace clean.

(function(){ // Closed testF scope.
// Example test code
})();

There is a small templating system that performs evaluation on
supplanted code marked by ${/*code*/}. This was necessary to
facilitate white box testing This is viewable in the source of this
document. The templating system uses a setInnerHTML function to deal
with Internet Explorer's whitespace problem.
 
D

Dr J R Stockton

In comp.lang.javascript message <[email protected]
legroups.com>, Fri, 19 Oct 2007 07:43:00, "(e-mail address removed)"
The code examples are all plain green. I'll beautify them later (i do
this in my editor w/a regexp find/rep)

http://dhtmlkitchen.com/learn/js/enumeration/


I will not be reading it, because :

You have overridden my chosen settings for font face and size.
Your chosen font is sans-serif, and too small, and will not Zoom in IE6
(OK, a heading Zooms; but having part-Zoom is silly).
You force the width so that the lines are too long and have too many
words.

I did try, though, and had a thought re .toFixed examples, which also
applies in the newsgroup FAQ and elsewhere : the example inputs to
..toFixed could be changed to be such that Javascript stores them
exactly. It will then be more honestly obvious what the point at issue
is. 0.07 can become 0.0625. Since 1.1255 cannot be stored exactly, but
only as (I think)
+1.1254999999999999449329379785922355949878692626953125
it should round down.

For ease of reading, a decimal point should always have at least one
digit on each side, if possible.

The pages seem so big that there probably should be a warning for dial-
up users but I could be wrong there as Open Office is downloading
currently.
 
D

dhtmlkitchen

In comp.lang.javascript message <[email protected]
legroups.com>, Fri, 19 Oct 2007 07:43:00, "(e-mail address removed)"
<[email protected]> posted:





I will not be reading it, because :

You have overridden my chosen settings for font face and size.
Your chosen font is sans-serif, and too small, and will not Zoom in IE6
(OK, a heading Zooms; but having part-Zoom is silly).
You force the width so that the lines are too long and have too many
words.

ctrl+/- works fine in IE7 and all other browsers. I don't have IE6, so
I couldn't try it.

Zoom is amazing in Opera!

I did try, though, and had a thought re .toFixed examples, which also
applies in the newsgroup FAQ and elsewhere : the example inputs to
.toFixed could be changed to be such that Javascript stores them
exactly. It will then be more honestly obvious what the point at issue
is. 0.07 can become 0.0625. Since 1.1255 cannot be stored exactly, but
only as (I think)
+1.1254999999999999449329379785922355949878692626953125
it should round down.
1.1254999999999999449329379785922355949878692626953125 === 1.1255; //
true

A more useful, pragmatic result of 1.1255.toFixed(3) would be 1.126,
not 1.125. I believe that is the intent of the spec, but it falls
short, unfortunately. I think Safari had the right approach. It might
not be 100% compliant, but gets results that don't require mental
binary conversions.



There are 10 types of people in the world: Those who understand
binary, and those who don't...

For ease of reading, a decimal point should always have at least one
digit on each side, if possible.

The pages seem so big that there probably should be a warning for dial-
up users but I could be wrong there as Open Office is downloading
currently.
Nah, no warnings. The pages are only about 10-12k after gzip. The
server response does seem pretty slow, though. I'm going to get a java
profiler for that.

I broke out a few examples into new pgs. I'll probably break out a few
more pgs. It seems that the pages are too long. I should break them
out and summarize.
 
D

Dr J R Stockton

In comp.lang.javascript message <[email protected]
oglegroups.com>, Sat, 20 Oct 2007 20:36:23, "(e-mail address removed)"
ctrl+/- works fine in IE7 and all other browsers. I don't have IE6, so
I couldn't try it.

Zoom is amazing in Opera!

Yes. But there are still a fixed number of words per line, so zooming
in soon reaches the screen width. The whole approach is too dirigiste.
1.1254999999999999449329379785922355949878692626953125 === 1.1255; //
true

Yes, but it remains true for substantial changes in the LHS : in
particular, 1.12550000000000005 === 1.1255 .
A more useful, pragmatic result of 1.1255.toFixed(3) would be 1.126,
not 1.125. I believe that is the intent of the spec, but it falls
short, unfortunately. I think Safari had the right approach. It might
not be 100% compliant, but gets results that don't require mental
binary conversions.

Since 1.1255 is stored under-value, 1.1255.toFixed(3) should give 1.125.
IE6 gives 1.126, Opera 9.24 and Firefox 2.0.0.8 give 1.125 .

What should be said is that, if always getting the exact final digit
matters, one should only use quantities, including intermediates, that
can be represented exactly in IEEE Doubles. Input and output strings
can be adjusted accordingly, as strings.


Please read the FAQ about news reply formatting.

Consider Section 3(a) of the Uniform Time Act of 1966 (15 U.S.C.
260a(a)) !


You have :

var day = 1000 * 60 * 60 * 24;
var yesterday = new Date(new Date() -day );
var today = new Date();
var tomorrow = new Date(new Date() - (-day) );

The second and fourth of those lines are wrong, unless the code is not
to be used with certain date/time/place combinations. Without curing
that fault, the fourth line is better written as

var tomorrow = new Date(+new Date() + day);

One can avoid the creation of a second Object by
var D = new Date() ; D.setTime(+D + Day)
or
var D = new Date() ; D.setDate(D.getDate()+1)


In Opera, Ctrl-Alt-V finds errors.
 
P

Peter Michaux

I have an article online, but it's not officially linked yet. This
will be the first public link.

I've found tons of errors myself, and there's probably plenty more --
I just didn't find them yet.

I'd really like to get some review on this. Both technical and
editorial, but mostly technical.

The code examples are all plain green. I'll beautify them later (i do
this in my editor w/a regexp find/rep)

http://dhtmlkitchen.com/learn/js/enumeration/

The article is very long and it seems like the article could be broken
into several smaller ones focused on one particular recommendation.
For example, the article talks about broken dontEnum in JScript and
also about the location of the JSON methods. Those seem like very
separate issues.

Peter
 
D

dhtmlkitchen

On Oct 19, 12:43 am, "(e-mail address removed)" <[email protected]>
wrote:

The article is very long
Yes, 12 long pages + 2 popup examples! I really didn't want it to be
more than 3-4 pages.

and it seems like the article could be broken
into several smaller ones focused on one particular recommendation.
For example, the article talks about broken dontEnum in JScript and
also about the location of the JSON methods. Those seem like very
separate issues.
You're right.

I'm trying to get Nicholas to write a contrasting proposal. That way,
I could have a simple link;

* Alterate JSON Proposal

Instead of the lengthy (and irrelevant) JSON discussion.

Time for a thread on that... :)
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

No members online now.

Forum statistics

Threads
473,755
Messages
2,569,536
Members
45,012
Latest member
RoxanneDzm

Latest Threads

Top