Best Performance Javascript on IE6 for finding top and left usingoffsetTop and offsetLeft

[moha297]

I want to get the top and left values for a div on the screen.

I have been using the code to calculate the top and left values.

var total1 = 0;
var total2 = 0;
while(element){
total1+=element.offsetTop;
total2+=element.offsetLeft;
try{
element=element.offsetParent;
}catch(E){
break;
}
}

For the same DOM TREE this code is giving a performance reading of
30msec in IE8 and 80 to 200msec in IE6. I want to gain a considerable
performance improvement in IE6.

I am open to all ideas.

By the way I also tried

var total1 = 0;
var total2 = 0;
do{
total1+=element.offsetTop;
total2+=element.offsetLeft;
}while(element=element.offsetParent);

after reading a blog entry or two. The performace is similar.

I found that the bulk of the time is always spent in getting property
value rather than parsing the DOM TREE which less constantly from my
logs.(I might be wrong)

Also the code I have put here is on the fly....might have made
mistakes.

Am looking at performance gain. The floor is open to try anything only
constraint is just javascript, jquery, prototype etc are not an
option.

Thanks in advance!!

 

[Peter Michaux]

I want to get the top and left values for a div on the screen.

physical screen or upper-left corner of the page (which may be out of
view if the page is scrolled.)?

As Richard Cornford has mentioned here many times, this problem is not
solved in general. If your div has parents that scroll, have table
elements, is a button, etc, then the calculation of the div's upper-
left corner relative to the upper-left corner of the page is complex.

I have been using the code to calculate the top and left values.

var total1 = 0;
var total2 = 0;
while(element){
total1+=element.offsetTop;
total2+=element.offsetLeft;
try{
element=element.offsetParent;

}catch(E){
break;
}
}

You should not need a try-catch block to calculate the position. Using
feature detection once when this code first runs, or when the page
first loads, should be enough to know how to calculate the offset in
that browser for the remainder of the life of the page.

Also, perhaps someone else can comment on the possible slowness of the
try-catch. I rarely use them.

For the same DOM TREE this code is giving a performance reading of
30msec in IE8 and 80 to 200msec in IE6. I want to gain a considerable
performance improvement in IE6.

Why is 80 ms considered a performance problem?

I am open to all ideas.

By the way I also tried

var total1 = 0;
var total2 = 0;
do{
total1+=element.offsetTop;
total2+=element.offsetLeft;

}while(element=element.offsetParent);

after reading a blog entry or two. The performace is similar.

I found that the bulk of the time is always spent in getting property
value rather than parsing the DOM TREE which less constantly from my
logs.(I might be wrong)

"parsing the DOM TREE"? That is done when the page loads, not when
your calculation is occurring.

Also the code I have put here is on the fly....might have made
mistakes.

The better the code you post, the more valuable the responses will be.

Peter

 

[Thomas 'PointedEars' Lahn]

I want to get the top and left values for a div on the screen.

I have been using the code to calculate the top and left values.

var total1 = 0;
var total2 = 0;
while(element){
total1+=element.offsetTop;
total2+=element.offsetLeft;
try{
element=element.offsetParent;
}catch(E){

Should be `e' as it does not refer to a constructor.

break;
}
}

For the same DOM TREE this code is giving a performance reading of
30msec in IE8 and 80 to 200msec in IE6.

That is unsurprising since that try...catch statement does not do anything
useful here as the assignment is not going to fail:

If `element' does not refer to an object, the `while (element)' statement
prevents execution from ever reaching the problematic assignment.

If `element' does refer to an object, and that object does not have an
`offsetParent' property or its `offsetParent' property value is `undefined'
or `null', `element' will be assigned either `undefined' or `null'. Since
that converts to `false', the next iteration is not going to happen and the
case that one would attempt to access the `offsetParent' property of
`undefined' or `null'.

I want to gain a considerable
performance improvement in IE6.

I am open to all ideas.

In most cases you do not need to determine the absolute position of an
element in the first place. For example, if you want to move an element by
10 px towards the right margin and by 20 px towards the bottom margin, you
simply increase its current `left' and `top' px-measured style property
values by those distances (while keeping the unit). If you have not set
those properties before, use the element's computed style.

However, if you insist, read this article, among others about this oft-
discussed problem in this newsgroup:

By the way I also tried

var total1 = 0;
var total2 = 0;
do{
total1+=element.offsetTop;
total2+=element.offsetLeft;
}while(element=element.offsetParent);

after reading a blog entry or two.

Do not believe into anything written.

The performace is similar.

It should be more efficient. Are you sure your benchmark is sound?

I found that the bulk of the time is always spent in getting property
value rather than parsing the DOM TREE which less constantly from my
logs.(I might be wrong)

You are. You do not know what "(to) parse" means to begin with.

Also the code I have put here is on the fly....might have made
mistakes.

It is hardly readable in the first place.

Am looking at performance gain. The floor is open to try anything only
constraint is just javascript, jquery, prototype etc are not an
option.

jQuery and Prototype are written in "javascript", though. But you are
correct avoiding them because of their bad code quality.


PointedEars

 

[GTalbot]

I want to get the top and left values for a div on the screen.

As Peter Michaux replied to you, your description of the problem for
which you require assistance is not accurate, too general. An URL
would have helped. And maybe, just maybe, you may be wrongly using
offsetLeft and offsetTop to get the left and top values of a div on
the screen. We can't be sure of this without a real webpage, URL.

I have been using the code to calculate the top and left values.

var total1 = 0;
var total2 = 0;
while(element){
total1+=element.offsetTop;
total2+=element.offsetLeft;
try{
element=element.offsetParent;

}catch(E){
break;
}
}

First, the while statement does not make sense.

You want an element (which is going to execute the controlled block)
that has offsetTop and offsetLeft values to do the controlled block.
Therefore, your while statement should be

while (element.offsetParent) {..controlled block..}
meaning as long as the current element being examined has an non-null
offsetParent...

Second, using a try.. catch does not perfection make sense from a
debugging perspective and from a property detection support. Let's say
the assignment fails because the current element being examined in
that while block does not have an offsetParent: why should it
generates an exception or an error object? Anyway, try.. catch is for
managing exceptions, for debugging purposes. At least, this is what I
would want to do when choosing to add a try...catch. And here, you do
not even try to identify the error message, error line, type of error,
etc.. So why resort to a try..catch block anyway?

Third, your local variable identifiers (total1, total2) are not
recommendable. You should always try to choose identifiers for
variables that are meaningful, intuitive, that helps debugging, code
maintenance, examining in debugging tools, that helps review by others
who may not be accustomed to your internal function logic (or help
review by yourself years later). This helps everyone and can make a
huge difference when the code is very long, complex, with many
intricated functions.

Here's how I did the same function 7 years ago:


var Element = evt.target ;
var CalculatedTotalOffsetLeft = CalculatedTotalOffsetTop = 0 ; while
(Element.offsetParent)
{
CalculatedTotalOffsetLeft += Element.offsetLeft ;
CalculatedTotalOffsetTop += Element.offsetTop ;
Element = Element.offsetParent ;
} ;

OffsetXForNS6 = evt.pageX - CalculatedTotalOffsetLeft ;
OffsetYForNS6 = evt.pageY - CalculatedTotalOffsetTop ;



http://www.gtalbot.org/DHTMLSection/WindowEventsNS6.html#screenLeft#NoteOffsetXY

http://www.gtalbot.org/DHTMLSection/WindowEventsIE6.html#screenLeft#NoteOffsetXY

For the same DOM TREE this code is giving a performance reading of
30msec in IE8 and 80 to 200msec in IE6. I want to gain a considerable
performance improvement in IE6.

IE6 <sigh .. Why do you need to support IE6?>. Imagine that people
are less and less using that browser and that IE8 implemented an
improved offsetParent, offsetLeft and offsetTop model.

With those numbers, don't you want to tell your IE6 users to upgrade
or to switch? It would solve many many problems...

"30msec in IE8 and 80 to 200msec" does not mean much if we do not know
on which machine (CPU, RAM, video card, etc) these results are
gathered from. 30msec is very long for a super-computer and 200msec is
very fast on a Pentium 1 90Mhz.

If nodeA is an area HTML element which has a map HTML element
somewhere in the ancestor chain, then nodeA.offsetParent returns the
nearest ancestor map HTML element... but that is not the case in IE 7;
IE8 corrected this.

http://www.gtalbot.org/BrowserBugsSection/MSIE8Bugs/CSSOM-offsetParent-prop..html#FourthTest

If an element has no offsetParent, then its offsetLeft value must be 0
and its offsetTop value must be 0 ... but that is not the case in IE6.

http://www.gtalbot.org/BrowserBugsSection/MSIE6Bugs/OffsetValues.html

I am open to all ideas.

Post an URL. Make sure your webpage is using valid markup code
(including a doctype declaration, preferably declaring a strict DTD),
uses valid CSS code.

http://validator.w3.org/
http://jigsaw.w3.org/css-validator/

Also, read the comp.lang.javascript FAQ on posting code:

http://www.jibbering.com/faq/#posting

By the way I also tried

    var total1 = 0;
    var total2 = 0;
do{
  total1+=element.offsetTop;
    total2+=element.offsetLeft;

}while(element=element.offsetParent);

An assignment in a while clause is not recommendable, at least,
definitely not my recommendation. Some assignment may succeed and
return 0 (and be resolved as false while the assignment is successful
and correct).

Same thing with assignment in a if clause:
if(a = b)
may succeed but the value may be resolved as false because b == 0.

Some other regular posters may help me here on this precise issue.

after reading a blog entry or two. The performace is similar.

Why is performance important to you, with regards to offsetTop and
offsetLeft and with regards to IE 6?
Please elaborate.

(There is such a thing has having an overexcessive number of
positioned containers (like nested tables)... in a very bloated
webpage)

I found that the bulk of the time is always spent in getting property
value rather than parsing the DOM TREE which less constantly from my
logs.(I might be wrong)

Post an URL according to the constraints I gave you.

And if you are open to all ideas, then add a IE6nomore or IE6RIP
button in your webpage so that people get the message that IE6 is
buggy, not recommendable, etc. should upgrade or switch.

Also the code I have put here is on the fly....might have made
mistakes.

Well, whose fault is it then? Are you asking to get code answers on
the fly as well... with possible mistakes too?

"
The better the code you post, the more valuable the responses will
be.
"
Thank you Peter Michaud for speaking up my mind.

Gérard

 

[GTalbot]

For the same DOM TREE this code is giving a performance reading of
30msec in IE8 and 80 to 200msec in IE6. I want to gain a considerable
performance improvement in IE6.

Those "30msec in IE8 and 80 to 200msec in IE6" numbers are entirely
dependent and relative to the depth of the DOM tree of the tested
webpage and to the CPU+RAM of the tested machines. The best possible
(short term and long term) proactive solution for anyone/everyone
involved is still to always use updated web-capable softwares and not
to use very buggy, unreliable, non-trustworthy software like IE6.

I am open to all ideas.

By the way I also tried

    var total1 = 0;
    var total2 = 0;
do{
  total1+=element.offsetTop;
    total2+=element.offsetLeft;

}while(element=element.offsetParent);

Making an assignment in a while statement is not recommendable;
definitely not my recommendation. The while statement should be a
condition, an expression resulting into a boolean condition (and only
that). The while statement should not be an assignment: you want to
avoid side effects here.

Correct and coherent is:

do
{
total1+=element.offsetTop;
total2+=element.offsetLeft;
element = element.offsetParent;
/* moves upward in the offsetParent containment hierarchy: this
assignment must succeed because the while statement must have been
true */
}
while(element.offsetParent);
/* first test if the current element being actually examined has an
offsetParent before entering the controlled block */

Gérard

 

[David Mark]

As Peter Michaux replied to you, your description of the problem for
which you require assistance is not accurate, too general. An URL
would have helped. And maybe, just maybe, you may be wrongly using
offsetLeft and offsetTop to get the left and top values of a div on
the screen. We can't be sure of this without a real webpage, URL.




First, the while statement does not make sense.

You want an element (which is going to execute the controlled block)
that has offsetTop and offsetLeft values to do the controlled block.
Therefore, your while statement should be

while (element.offsetParent) {..controlled block..}

Just don't pass orphaned elements. I imagine that's what that
mysterious try-catch was about.

meaning as long as the current element being examined has an non-null
offsetParent...

And it better not have typeof "unknown" either.

Second, using a try.. catch does not perfection make sense from a
debugging perspective and from a property detection support.

Right. It's just hiding some other problem (code passing an orphaned
element).

Let's say
the assignment fails because the current element being examined in
that while block does not have an offsetParent: why should it
generates an exception or an error object?

It definitely can.

Anyway, try.. catch is for
managing exceptions, for debugging purposes. At least, this is what I
would want to do when choosing to add a try...catch.

Right. There should not be a try-catch here.

And here, you do
not even try to identify the error message, error line, type of error,
etc.. So why resort to a try..catch block anyway?

Exactly. It's just covering up a problem.

Third, your local variable identifiers (total1, total2) are not
recommendable. You should always try to choose identifiers for
variables that are meaningful, intuitive, that helps debugging, code
maintenance, examining in debugging tools, that helps review by others
who may not be accustomed to your internal function logic (or help
review by yourself years later). This helps everyone and can make a
huge difference when the code is very long, complex, with many
intricated functions.

Here's how I did the same function 7 years ago:

var Element = evt.target ;
var CalculatedTotalOffsetLeft = CalculatedTotalOffsetTop = 0 ; while
(Element.offsetParent)
{
  CalculatedTotalOffsetLeft += Element.offsetLeft ;
  CalculatedTotalOffsetTop += Element.offsetTop ;
  Element = Element.offsetParent ;

} ;

OffsetXForNS6 = evt.pageX - CalculatedTotalOffsetLeft ;
OffsetYForNS6 = evt.pageY - CalculatedTotalOffsetTop ;

http://www.gtalbot.org/DHTMLSection/WindowEventsNS6.html#screenLeft#N...

http://www.gtalbot.org/DHTMLSection/WindowEventsIE6.html#screenLeft#N...

That doesn't take borders into account. If you _must_ measure all the
way to the document origin (virtually never), use
getBoundingClientRect.

IE6 <sigh ..  Why do you need to support IE6?>.

Because lots of corporate users are stuck with it? And because
there's virtually no difference with IE7. And then there are the
browsers that copied IE6 oddities...

Imagine that people
are less and less using that browser and that IE8 implemented an
improved offsetParent, offsetLeft and offsetTop model.

Yes, less people are using IE6. IE6-8 (and perhaps earlier) all have
getBoundingClientRect (and other browsers have copied). But it is
much simpler and less problematic to measure to the origin of a
positioned ancestor.

With those numbers, don't you want to tell your IE6 users to upgrade
or to switch? It would solve many many problems...

Because you can't tell users to upgrade their browsers. Some can't
upgrade them. Some don't know what a browser _is_.

"30msec in IE8 and 80 to 200msec" does not mean much if we do not know
on which machine (CPU, RAM, video card, etc) these results are
gathered from. 30msec is very long for a super-computer and 200msec is
very fast on a Pentium 1 90Mhz.

If nodeA is an area HTML element which has a map HTML element
somewhere in the ancestor chain, then nodeA.offsetParent returns the
nearest ancestor map HTML element... but that is not the case in IE 7;
IE8 corrected this.

http://www.gtalbot.org/BrowserBugsSection/MSIE8Bugs/CSSOM-offsetParen...

Skip the image maps and you'll be fine.

[...]

Why is performance important to you, with regards to offsetTop and
offsetLeft and with regards to IE 6?
Please elaborate.

(There is such a thing has having an overexcessive number of
positioned containers (like nested tables)... in a very bloated
webpage)

Yes. And there is such a thing as making one of them
position:relative.

Post an URL according to the constraints I gave you.

And if you are open to all ideas, then add a IE6nomore or IE6RIP
button in your webpage so that people get the message that IE6 is
buggy, not recommendable, etc. should upgrade or switch.

Hell no. You _never_ insult the user's browser. They may not be able
(or know how) to upgrade.

 

[David Mark]

physical screen or upper-left corner of the page (which may be out of
view if the page is scrolled.)?

As Richard Cornford has mentioned here many times, this problem is not
solved in general. If your div has parents that scroll, have table
elements, is a button, etc, then the calculation of the div's upper-
left corner relative to the upper-left corner of the page is complex.

It's been done. It's just not an advisable cross-browser design to
rely on such complex code when simpler options are available.

 

[GTalbot]

Just don't pass orphaned elements.

I do not understand your "Just don't pass orphaned elements." .. or
I'm not sure I understand what you mean to say.

If the currently tested element has no offsetParent, then the
controlled block is not executed and the execution continues, carries
on, goes out of the while loop. Isn't that what is sought here?

It definitely can.

Yes, you're right. If the assignment fails, then there should be an
exception created. I got mixed up with something else.

Sometimes the assignment succeeds but the resulting expression is
evaluated as false.

if(a = b) {... controlled block ...};

if b == 0, then the controlled block may not be executed even though
the assignment was successfully executed. So, it shouldn't be what the
coder expected.

That doesn't take borders into account.  If you _must_ measure all the
way to the document origin (virtually never), use
getBoundingClientRect.

Borders. This is news to me. I'll have to verify this some day.

Because lots of corporate users are stuck with it?  And because
there's virtually no difference with IE7.

As far as offsetParent, offsetLeft and offsetTop are involved, there
is very little difference between IE6 and IE7: only 1 difference,
IIRC.
There are some/more differences between IE 6 and IE 7 with regards to
positioniseverything.net bugs

Explorer Exposed
http://www.positioniseverything.net/explorer.html

 And then there are the
browsers that copied IE6 oddities...


Yes, less people are using IE6.  IE6-8 (and perhaps earlier) all have
getBoundingClientRect (and other browsers have copied).  But it is
much simpler and less problematic to measure to the origin of a
positioned ancestor.

When the ancestor is positioned (non-static), yes. But what happens
when the elements are within a table or within nested tables? I think
you still have to resort to offsetParent.

Because you can't tell users to upgrade their browsers.

I try to invite them to upgrade and try to address their intelligence
at all times. Ultimately, it's all up to them to decide. But I don't
hide (or don't try to hide) to them that IE6 is very buggy,
unreliable, not-trustworthy, etc. I certainly want them to know and
understand that web-capable softwares (or any kind of softwares)
should be using the latest stable available version for all kinds of
reasons: security, bug fixes, stability, speed, accessibility and
usability features, etcetctectcc.

 Some can't
upgrade them.  

Often, these people can still/nevertheless install an alternate
browser.

In June 11th 2009 in the US, television signal stopped being for
analog tv; about 15% of users could no longer get local tv. In Canada,
the deadline for getting ready for digital-only tv signal is august
2011. It's roughly the same for other countries regarding the
transition to digital tv.
If you have to buy a new tv because of technological reasons (and FCC
and government rules), then I don't understand why corporate users can
not upgrade their browser softwares. They certainly were warned in the
past that their applications shouldn't be entirely dependent on
Microsoft products and Microsoft Windows platform.

The top 20 IT mistakes to avoid, November 19, 2004
11. Developing Web apps for IE only
http://www.infoworld.com/d/developer-world/top-20-it-mistakes-avoid-314?page=0,3

Some don't know what a browser _is_.  

Those (like seniors) are the most difficult portion of the market.
There's very little you can do ... unless you're a friend of them and
you visit them at home.

[snipped]

Hell no.  You _never_ insult the user's browser.  They may not be able
(or know how) to upgrade.

If your message is not agressive or on purpose bashing the browser
itself, if you invite diplomatically to upgrade or switch and if you
address their intelligence, then those who can may do so. Those who
can not .. whatever the reasons .. will not.

It is objectively still in the user's best interests to always use the
most updated stable release (for countless of reasons) of any software
(s)he may be using.

regards, Gérard

 

[David Mark]

I do not understand your "Just don't pass orphaned elements." .. or
I'm not sure I understand what you mean to say.

Orphaned (removed from the document) elements can become ActiveX
objects behind the scenes. IIRC, orphaning by innerHTML replacement
is a sure bet.

if (typeof element.offsetParent == 'unknown') {
(element.offsetParent); // Boom
}

So you wouldn't normally pass such an element to such a function
(offset position makes no sense for orphans). The try-catch hides
such mistakes.

If the currently tested element has no offsetParent, then the
controlled block is not executed and the execution continues, carries
on, goes out of the while loop. Isn't that what is sought here?

See above. Just evaluating it is enough to trigger an exception.

Yes, you're right. If the assignment fails, then there should be an
exception created. I got mixed up with something else.

Sometimes the assignment succeeds but the resulting expression is
evaluated as false.

if(a = b) {... controlled block ...};

if b == 0, then the controlled block may not be executed even though
the assignment was successfully executed. So, it shouldn't be what the
coder expected.


Borders. This is news to me. I'll have to verify this some day.

Just search the archive or the Web. You have to include the borders
(clientLeft/Top) in most browsers (it's something that has to be
tested). Some older Opera's were broken in this regard.

As far as offsetParent, offsetLeft and offsetTop are involved, there
is very little difference between IE6 and IE7: only 1 difference,
IIRC.
There are some/more differences between IE 6 and IE 7 with regards to
positioniseverything.net bugs

There are a few. Most (if not all) are avoidable.

Explorer Exposedhttp://www.positioniseverything.net/explorer.html



When the ancestor is positioned (non-static), yes. But what happens
when the elements are within a table or within nested tables? I think
you still have to resort to offsetParent.

I didn't mean that you didn't have to use offsetParent at all. I mean
limit the number of "hops" and stop short of the body (unless it has
position:relative).

I try to invite them to upgrade and try to address their intelligence
at all times.

See, the typical user don't want to hear about their intelligence.
Thin ice.

Ultimately, it's all up to them to decide. But I don't
hide (or don't try to hide) to them that IE6 is very buggy,
unreliable, not-trustworthy, etc.

That's ridiculous. If your page is buggy and/or unreliable, the user
will (rightfully) blame _you_. No amount of blaming the browser will
move them (as well it should not).

I certainly want them to know and
understand that web-capable softwares (or any kind of softwares)
should be using the latest stable available version for all kinds of
reasons: security, bug fixes, stability, speed, accessibility and
usability features, etcetctectcc.


Often, these people can still/nevertheless install an alternate
browser.

Some can and some can't.

In June 11th 2009 in the US, television signal stopped being for
analog tv; about 15% of users could no longer get local tv. In Canada,
the deadline for getting ready for digital-only tv signal is august
2011. It's roughly the same for other countries regarding the
transition to digital tv.
If you have to buy a new tv because of technological reasons (and FCC
and government rules), then I don't understand why corporate users can
not upgrade their browser softwares.

You don't need to understand everything to be successful.

They certainly were warned in the
past that their applications shouldn't be entirely dependent on
Microsoft products and Microsoft Windows platform.

The top 20 IT mistakes to avoid, November 19, 2004
11. Developing Web apps for IE onlyhttp://www.infoworld.com/d/developer-world/top-20-it-mistakes-avoid-3...

That's another topic altogether. Some corporate users are simply
stuck with IE6/7 and will be for years. In most cases, it has nothing
to do with building IE-only websites.

Those (like seniors) are the most difficult portion of the market.

Lots of PC-savvy corporate users are clueless about browsers and/or
disallowed from making decisions about which browsers to use.

There's very little you can do ... unless you're a friend of them and
you visit them at home.

Sounds like a very hard way to go. They won't let you waltz into
corporate sites changing browsers anyway. I don't think you have time
to visit every "senior" at home either. So tackle the problem from
the other end.

[snipped]

Hell no.  You _never_ insult the user's browser.  They may not be able
(or know how) to upgrade.

If your message is not agressive or on purpose bashing the browser
itself, if you invite diplomatically to upgrade or switch and if you
address their intelligence, then those who can may do so. Those who
can not .. whatever the reasons .. will not.

Nobody wants to hear about it from your site, unless your site is
about browsers.

It is objectively still in the user's best interests to always use the
most updated stable release (for countless of reasons) of any software
(s)he may be using.

See above for various reasons why not.

 

[Garrett Smith]

Should be `e' as it does not refer to a constructor.

Right. One more consideration is that IE (JScript, actually) will add
the identifier not to catch-block's scope, but to the function's
variable object.

In an event handler callback, it is common to use - e - for the event
parameter.

When a function with variable - e- reatches a catch block and the catch
block has `e`, then in IE, the function's variable `e` refers to the
Error in IE.

Example:

document.body.onclick = function(e) {
e = e || window.event;
try {
throw new Error("catch scope bugs!");;
} catch(e) { //<-- XXX JScript bug, replaced e's value with error.
} finally {
alert(e.message || "Pass:" + e.type);
}
};

IE elerts: "catch scope bugs!"

The convention I have adopted is to name exception identifier as - ex -
and event as - ev -.

 

[David Mark]

Right. One more consideration is that IE (JScript, actually) will add
the identifier not to catch-block's scope, but to the function's
variable object.

In an event handler callback, it is common to use - e - for the event
parameter.

When a function with variable - e- reatches a catch block and the catch
block has `e`, then in IE, the function's variable `e` refers to the
Error in IE.

Example:

document.body.onclick = function(e) {
   e = e || window.event;
   try {
     throw new Error("catch scope bugs!");;
   } catch(e) { //<-- XXX JScript bug, replaced e's value with error.
   } finally {
     alert(e.message || "Pass:" + e.type);
   }

};

IE elerts: "catch scope bugs!"

The convention I have adopted is to name exception identifier as - ex -
and event as - ev -.

JSLint lets you know about such ambiguities before they bite you.

Error:

Problem at line 5 character 12: 'e' is already defined.

} catch(e) { //<-- XXX JScript bug, replaced e's value with error.

Implied global: window 1,2,7

Also, you've got an "unreachable" semi-colon and a couple of implied
globals (the alert one is troubling).

 

[Peter Michaux]

It's been done.

I've never seen such code and whoever claims "it does everything" is
probably not correct.

It's just not an advisable cross-browser design to
rely on such complex code

The implication of that statement is that the code likely doesn't do
everything and so code that does everything doesn't really exist.

when simpler options are available.

Agreed.

Peter

 

[Peter Michaux]

Should be `e' as it does not refer to a constructor.

If we are now off topic, some folks like to write globals in all
capitals. Since that looks like it might be an (implied) global then
captial 'E' may be appropriate.

That is unsurprising since that try...catch statement does not do anything
useful here as the assignment is not going to fail:

The assignment will not fail but getting a host object's property may.

In most cases you do not need to determine the absolute position of an
element in the first place.

That is absolutely ;-) correct. I think I've only needed absolute
position once in the past couple years.

Peter

 

[David Mark]

I've never seen such code and whoever claims "it does everything" is
probably not correct.

I'm not saying the last one I wrote does it all, but it covers most
(if not all) of the issues mentioned.

The implication of that statement is that the code likely doesn't do
everything and so code that does everything doesn't really exist.

Not exactly. I mean that even if it does work 100%, the code is quite
complex with multiple forks (shouldn't a first choice for a design).

 

[Peter Michaux]

If we are now off topic, some folks like to write globals in all
capitals. Since that looks like it might be an (implied) global then
captial 'E' may be appropriate.

Nevermind. I really don't use try-catch enough.

Peter

 

[Peter Michaux]

I'm not saying the last one I wrote does it all, but it covers most
(if not all) of the issues mentioned.

The case that Richard Cornford refers to is the case where the code
truly does it all. Even the full "etc" I mentioned.

Peter

 

[Garrett Smith]

[...]

The convention I have adopted is to name exception identifier as - ex -
and event as - ev -.

JSLint lets you know about such ambiguities before they bite you.

JSLint has so much imposition of the author's viewpoint's conveyed as
"errors" that it is hard to find the *real* errors. I would call a real
error someting like assignment to undeclared identifier.

I do not follow Doug's coding conventions. I do take advantage of YUI
warnings, and do pay attention to Firebug's "DOM" tab for globals that
got created.

I would like a lint tool that can find errors or be configured to handle
things like:-

document.addEventListener("click", documentClicked, false);

function documentClicked(ev){ }

JSLint doesn't allow that perfectly valid, easily understandable program
to pass.

Error:

Problem at line 5 character 12: 'e' is already defined.

} catch(e) { //<-- XXX JScript bug, replaced e's value with error.

Implied global: window 1,2,7

JSLint thinks that is an error. Even when "assume a browser" is checked.

Also, you've got an "unreachable" semi-colon and a couple of implied
globals (the alert one is troubling).

What is troubling about using alert?

JSLint parses window.alert(); or alert(); as an Error. See I just don't
get that. Now I can't figure out what makes that an error, yet:-

document.parentWindow.alert(1);

- is perfectly fine in JSLint. No error whatsoever. I don't get it.

 

[David Mark]

Thomas 'PointedEars' Lahn wrote:
moha297 wrote:

[...]

The convention I have adopted is to name exception identifier as - ex -
and event as - ev -.

JSLint lets you know about such ambiguities before they bite you.

JSLint has so much imposition of the author's viewpoint's conveyed as
"errors" that it is hard to find the *real* errors. I would call a real
error someting like assignment to undeclared identifier.

Nah, default settings yield about a dozen "errors" in My Library (out
of roughly 9,000 lines).

I do not follow Doug's coding conventions. I do take advantage of YUI
warnings, and do pay attention to Firebug's "DOM" tab for globals that
got created.

JSLint does far more than that.

I would like a lint tool that can find errors or be configured to handle
things like:-

document.addEventListener("click", documentClicked, false);

function documentClicked(ev){ }

JSLint doesn't allow that perfectly valid, easily understandable program
to pass.

Bad style. Makes it more confusing than it needs to be (for beginners
who may have to maintain it). And you can turn that one off I think.

JSLint thinks that is an error. Even when "assume a browser" is checked.

Well, I don't care for that one.

What is troubling about using alert?

Nothing troubling about using alert. Now, calling it unqualified in a
- finally - clause is another story.

JSLint parses window.alert(); or alert(); as an Error. See I just don't
get that.

He wants it to be fully qualified. There are cases where that message
is helpful (e.g. generalizing functions for use with frames).

Now I can't figure out what makes that an error, yet:-

document.parentWindow.alert(1);

- is perfectly fine in JSLint. No error whatsoever. I don't get it.

That's an implied global for sure. Or perhaps you mean with "assume a
browser"? Yes, I agree that assuming a browser should allow for the
user of window (it used to).

 

[Garrett Smith]

Thomas 'PointedEars' Lahn wrote:
moha297 wrote: [...]

[...]

function documentClicked(ev){ }

JSLint doesn't allow that perfectly valid, easily understandable program
to pass.

Bad style. Makes it more confusing than it needs to be (for beginners
who may have to maintain it). And you can turn that one off I think.

It is fine that way; nothing really confusing about it. It is
syntactically valid, standard code for any js engine.

A beginner might find it *odd* that a function could be referenced
before it appeared in source order. A beginner might not know about
entering an execution context, what a FunctionDeclaration was, what a
FunctionExpression was, but would just notice that "it works".

A beginner would not know a lot of things, including things about
unicode support, ===, typeof, Host object.

Nothing troubling about using alert. Now, calling it unqualified in a
- finally - clause is another story.

What difference does it make if it is finally clause?

He wants it to be fully qualified. There are cases where that message
is helpful (e.g. generalizing functions for use with frames).

What is it that you think he wants to be fully qualified?

That's an implied global for sure. Or perhaps you mean with "assume a
browser"? Yes, I agree that assuming a browser should allow for the
user of window (it used to).

It is odd because document is a property of the window.

 

[David Mark]

David Mark wrote:
Thomas 'PointedEars' Lahn wrote:
moha297 wrote:
[...]

[...]

function documentClicked(ev){ }
JSLint doesn't allow that perfectly valid, easily understandable program
to pass.

Bad style.  Makes it more confusing than it needs to be (for beginners
who may have to maintain it).  And you can turn that one off I think.

It is fine that way; nothing really confusing about it. It is
syntactically valid, standard code for any js engine.

You know exactly what I mean.

A beginner might find it *odd* that a function could be referenced
before it appeared in source order.

Especially if they are coming to JS from another language.

A beginner might not know about
entering an execution context, what a FunctionDeclaration was, what a
FunctionExpression was, but would just notice that "it works".

A beginner would not know a lot of things, including things about
unicode support, ===, typeof, Host object.

So what?

What difference does it make if it is finally clause?

Augmented scope of course. It's best to reference alert as a method
of the window object (that's what it is after all). Then these
questions don't come up.

What is it that you think he wants to be fully qualified?

The window reference of course. He wants to see it referenced as a
property of an object (e.g. the Global object). As for frames,
consider this:-

e = e || window.event; // No good for elements in other frames

So that "error" in JSLint helps to spot these situations as well.

It is odd because document is a property of the window.

In a browser (assuming a browser), the Global object has a - document
- property as well. At least, that is the underlying assumption here.