optimizing code...

**Evertjan.** · Jul 20 '05, 02:35 PM

Re: optimizing code...

Ian Richardson wrote on 16 feb 2004 in comp.lang.javas cript:[color=blue]
> I'm writing a large Javascript application (uncompressed source around
> 400K) which is doing almost all the initialisation it needs to in a
> just-in-time manner. However, I have included an option for almost all
> of this to be done when the application first starts.
>
> Under this circumstance, and since the last few nightly builds of
> Mozilla, I've been getting:
>
>
> Script warning:
>
> A script on this page is causing mozilla to run slowly. If it continues
> to run, your computer may become unresponsive.
>
> Do you want to abort the script?
> [....][/color]

Script engines, that is interpreted language emulators, also called
interpreters, are not up to fast execution. Why not make a programme in a
language that will be compiled? Then the setup time end the dynamic
execution time will be orders of magnitude better.

If you compress the source code of your javascript script, the setup time
will be even longer, I suppose.

--
Evertjan.
The Netherlands.
(Please change the x'es to dots in my emailaddress)

**Ian Richardson** · Jul 20 '05, 02:36 PM

Re: optimizing code...

Evertjan. wrote:
[color=blue]
> Ian Richardson wrote on 16 feb 2004 in comp.lang.javas cript:
>[color=green]
>>I'm writing a large Javascript application (uncompressed source around
>>400K) which is doing almost all the initialisation it needs to in a
>>just-in-time manner. However, I have included an option for almost all
>>of this to be done when the application first starts.
>>
>>Under this circumstance, and since the last few nightly builds of
>>Mozilla, I've been getting:
>>
>>
>>Script warning:
>>
>>A script on this page is causing mozilla to run slowly. If it continues
>>to run, your computer may become unresponsive.
>>
>>Do you want to abort the script?
>>[....][/color]
>
>
> Script engines, that is interpreted language emulators, also called
> interpreters, are not up to fast execution. Why not make a programme in a
> language that will be compiled? Then the setup time end the dynamic
> execution time will be orders of magnitude better.[/color]

The only real option is to rewrite the whole thing in Java... and I'm
not looking forward to that!! I'd have to learn Java, first...
[color=blue]
> If you compress the source code of your javascript script, the setup time
> will be even longer, I suppose.[/color]

Taking out all the redundant comments and whitespace does actually save
a bit of time, but nothing like enough. That's why I'm looking to
optimize the code as much as possible.

Ian

**Richard Cornford** · Jul 20 '05, 02:36 PM

Re: optimizing code...

"Ian Richardson" <zathras@chaos. org.uk> wrote in message
news:c0pvpv$19j l07$1@ID-99375.news.uni-berlin.de...
<snip>[color=blue]
>What I need is:
>1. A means of being able to forcibly skip this message[/color]

The are no mechanisms for disabling the message from javascript. In so
far as it can be prevented it is usually done by splitting the code up
into chunks and executing each in sequence with setTimeout so the
browser gets a chance to catch up with what it needs to do, and reset
the counters it uses to measure the time a script has been running.
[color=blue]
>2. A faster form of loop than for(x in list)[/color]

Of all of the - for - loop options - for(var prop in obj) - is certainly
slowest but nobody can tell if any alternatives would be better (even
applicable) without knowing the context of its use.
[color=blue]
>3. A faster sort than a quicksort for a SELECT list[/color]

There is a big difference between a sort that tries to operate on the
options in a select element and sorting a data structure that it used to
build an options list. It is unlikely that anything useful can be said
on the subject witho9ut being able to see the implementation.
[color=blue]
>4. A faster form of search - I'm currently using binary search[/color]

A binary search, properly implemented, is hard to beat for speed.
[color=blue]
>5. A fast way of getting one OPTION at any point in a SELECT list
> moved to the top of the list. I'm currently using a method akin
> to a bubble sort.[/color]

That is a bit strange as moving an item to the top of a list only
requires that two items are considered/moved (the one that is to go to
the top of the list and the one already there). That would imply that
there are other criteria in the specification for the task.
[color=blue]
>[Lists can be up to 1000 items...][/color]

One thousand items is not that bad, you can't expect instantaneous
results form the browser/javascript combination but there is often room
for improvement.
[color=blue]
>I know I might be asking for the moon on a stick, but any help
>would be greatly appreciated.[/color]

With javascript performance is often a trade off, usually exchanging
increased memory use for performance. Generally you should aim to avoid
doing the same thing more than once wherever practical. Apart form that
there are lots of little things that done one way can inhibit
performance when accumulated in a big script, but are difficult to
generalise about and best addressed by discussing actual code. There is
also a lot of interest on this group in optimum implementations .

Unfortunately a 400Kb post would probably be unwelcome, a URL to an
example page (with scripts) would be better, but you may not feel like
rendering your work to date public. But you will not get much help with
the specifics without showing at leas some of the relevant code (and its
context).

Richard.

**Jim Ley** · Jul 20 '05, 02:36 PM

Re: optimizing code...

On Mon, 16 Feb 2004 14:13:14 -0000, "Richard Cornford"
<Richard@litote s.demon.co.uk> wrote:
[color=blue][color=green]
>>2. A faster form of loop than for(x in list)[/color]
>
>Of all of the - for - loop options - for(var prop in obj) - is certainly
>slowest but nobody can tell if any alternatives would be better (even
>applicable) without knowing the context of its use.[/color]

It is however dangerous and best avoided.
[color=blue][color=green]
>>3. A faster sort than a quicksort for a SELECT list[/color]
>
>There is a big difference between a sort that tries to operate on the
>options in a select element and sorting a data structure that it used to
>build an options list.[/color]

And you definately only want to do the 2nd one.
[color=blue][color=green]
>>4. A faster form of search - I'm currently using binary search[/color]
>
>A binary search, properly implemented, is hard to beat for speed.[/color]

The questions all seem to not be about javascript, but about how to
deal with lots of data, the problem is almost certainly one of data
design, and speeding up the javascript won't solve that underlying
issue.
[color=blue][color=green]
>>[Lists can be up to 1000 items...][/color]
>
>One thousand items is not that bad,[/color]

Yes it is, a select list with 1000 items is unusable (which is why
presumably he's needing to provide alternative navigation mechanisms)
the solution is obvious, change the navigation system so a SELECT box
never needs to have 1000 items in it!

Jim.
--
comp.lang.javas cript FAQ - http://jibbering.com/faq/

DU · Jul 20 '05, 02:36 PM

Re: optimizing code...

Ian Richardson wrote:
[color=blue]
> I'm writing a large Javascript application (uncompressed source around
> 400K) which is doing almost all the initialisation it needs to in a
> just-in-time manner. However, I have included an option for almost all
> of this to be done when the application first starts.
>
> Under this circumstance, and since the last few nightly builds of
> Mozilla, I've been getting:
>
>
> Script warning:
>
> A script on this page is causing mozilla to run slowly. If it continues
> to run, your computer may become unresponsive.
>
> Do you want to abort the script?
>
>
> I think this is being displayed because of the time it's taking to start
> up rather than the amount of code it's wading through[/color]

Is there a meaningful difference here? Is that really useful to know?
Anyway, isn't the time to start proportional to the amount of loading to
do and amount of execution to go through?

, but it's as good[color=blue]
> an excuse as any to look at optimizing code. [I've just realised and
> fixed one processor-intensive chunk of code which was being processed
> six times during startup; fine except the first four (possibly five,
> depending on circumstances) were completely unnecessary!]
>
> What I need is:
>
> 1. A means of being able to forcibly skip this message
>[/color]

Go see a car mechanic and ask him if he could disable the leds warning
about low oil pressure, door still opened, low battery voltage, etc..
things like that. Do you believe that removing these warning signs and
bells will help your car drive to remain a quiet, serene and enjoyable
experience?

Write this into a single javascript function:
while(true);
Now, wouldn't you appreciate your browser to warn you somehow about a
very demanding function which may be abusing, hogging your own system
resources and propose you of a way to abort such script?
[color=blue]
> 2. A faster form of loop than for(x in list)
>[/color]

If x is an extremely big number, then a javascript application is not
recommendable for your particular webpage. A java applet might be the
answer.
[color=blue]
> 3. A faster sort than a quicksort for a SELECT list
>
> 4. A faster form of search - I'm currently using binary search
>
> 5. A fast way of getting one OPTION at any point in a SELECT list moved
> to the top of the list. I'm currently using a method akin to a bubble sort.
>
>[Lists can be up to 1000 items...]
>
> I know I might be asking for the moon on a stick, but any help would be
> greatly appreciated.
>
> Thanks,
>
> Ian[/color]

You're asking for help about an unidentified webpage. There is little we
can suggest without being able to actually examine the code. There are
many ways and techniques to make javascript functions run faster and
there are many ways to make a webpage markup code be downloaded, parsed
and rendered faster. There are tutorials, books, articles on all this.

Are you familiar with memory footprint softwares and algorithmic
theoretical works?

DU

**Richard Cornford** · Jul 20 '05, 02:37 PM

Re: optimizing code...

"Jim Ley" <jim@jibbering. com> wrote in message
news:40310cd5.1 272629@news.ind ividual.net...
<snip>[color=blue][color=green][color=darkred]
>>>2. A faster form of loop than for(x in list)[/color]
>>
>>Of all of the - for - loop options - for(var prop in obj)
>>- is certainly slowest but nobody can tell if any alternatives
>>would be better (even applicable) without knowing the context
>>of its use.[/color]
>
>It is however dangerous and best avoided.[/color]

I wouldn't argue with that.

<snip>[color=blue]
>The questions all seem to not be about javascript, but about
>how to deal with lots of data, the problem is almost certainly
>one of data design, and speeding up the javascript won't solve
>that underlying issue.[/color]

Probably, but we do see plenty of scripts where each and every form
control property access is using an absolute property accessor and that
alone is adding up to cripple the script.
[color=blue][color=green][color=darkred]
>>>[Lists can be up to 1000 items...][/color]
>>
>>One thousand items is not that bad,[/color]
>
>Yes it is, a select list with 1000 items is unusable (which is
>why presumably he's needing to provide alternative navigation
>mechanisms)[/color]

Well, the browser can add those 1000 options in under half a second (on
a 500MHz PC) but yes the resulting select element is unusable. I was
hoping the 1000 items was an exception that had to be handled rather
than the norm, but thanking about it all of the searching and sorting
does imply an attempt at providing alternative navigation for the select
list and therefor an expectation that it will never be usable in the
normal way.
[color=blue]
>the solution is obvious, change the navigation system so a
>SELECT box never needs to have 1000 items in it![/color]

Removing the cause of a problem is a valid approach, preferably before
writing 400Kb of code.

Richard.

**Ian Richardson** · Jul 20 '05, 02:38 PM

Re: optimizing code...

Richard Cornford wrote:[color=blue]
> "Ian Richardson" <zathras@chaos. org.uk> wrote in message
> news:c0pvpv$19j l07$1@ID-99375.news.uni-berlin.de...
> <snip>
>[color=green]
>>What I need is:
>>1. A means of being able to forcibly skip this message[/color]
>
>
> The are no mechanisms for disabling the message from javascript. In so
> far as it can be prevented it is usually done by splitting the code up
> into chunks and executing each in sequence with setTimeout so the
> browser gets a chance to catch up with what it needs to do, and reset
> the counters it uses to measure the time a script has been running.[/color]

I'm only using this sort of technique as part of some code for loading
additional .js files - and that's only required for Internet Explorer :-(

I believe I've found where my code is falling down...
[color=blue][color=green]
>>2. A faster form of loop than for(x in list)[/color][/color]

I've got an associative array of several hundred items used for language
tokenisation, and I've got a for(x in list) type loop to find items in
it. This loop is potentially called over 600 times during the startup
sequence!!

The code works but was one of those bits I originally implemented as a
quick fix. Time to change this hideous system to something better,
possibly an array of Options.
[color=blue][color=green]
>>3. A faster sort than a quicksort for a SELECT list
>>4. A faster form of search - I'm currently using binary search[/color][/color]

I don't think I'm going to easily improve on these.
[color=blue][color=green]
>>5. A fast way of getting one OPTION at any point in a SELECT list
>> moved to the top of the list. I'm currently using a method akin
>> to a bubble sort.[/color][/color]
[color=blue]
> That is a bit strange as moving an item to the top of a list only
> requires that two items are considered/moved (the one that is to go to
> the top of the list and the one already there). That would imply that
> there are other criteria in the specification for the task.[/color]

I'm fully populating the lists, then quicksorting the list, then
shuffling a specified item (which could be anywhere in the list) to the
top, preserving the sorted order of the remainder. I could change this to:

1. put item I know I want at the top of the list first

2. populate the list, checking that I don't put the first item in again.

3. quicksort the list from the second item to the end.
[color=blue][color=green]
>>[Lists can be up to 1000 items...][/color]
>
>
> One thousand items is not that bad, you can't expect instantaneous
> results form the browser/javascript combination but there is often room
> for improvement.[/color]

I panicked. Some lists are up to 60 items, some 100, some around 300,
and two are 1000. Some of these lists are numbers only, but I'm doing
this to cut down on the amount of input validation and subsequent
reformatting required.

Mind you, if it's quicker (in both IE and Mozilla) to build the list
using DOM code then replacing the list instead of using lots of
list[list.length] = new Option(blah, blah) I'm all in favour of it. Any
suggestions?
[color=blue]
> With javascript performance is often a trade off, usually exchanging
> increased memory use for performance. Generally you should aim to avoid
> doing the same thing more than once wherever practical. Apart form that
> there are lots of little things that done one way can inhibit
> performance when accumulated in a big script, but are difficult to
> generalise about and best addressed by discussing actual code. There is
> also a lot of interest on this group in optimum implementations .[/color]

I'm afraid I can't post the code now but I anticipate it'll be public in
the next few weeks (finally!). I've been doing lots of work to cut down
on the startup time, restructuring the code to do as much just-in-time
initialisation as possible.

In the full version of the code it's possible to import data to preset
certain sections, either via a .js configuration file or the query
string, and this was forcing some of the just-in-time initialisation to
occur immediately. When nice'n'quick Mozilla (in comparison to IE on my
machine) started bringing up the slow script warnings I immediately
blamed the long SELECT lists or a function being performed on them. A
little more searching seems to have found a far more likely cause - the
associative array of blah["name1"]...blah["nameN"] language tokens and
the function to locate items in it.

I'll have a look, make the changes as soon as I can, and see how much
these two steps shave off the startup time.

Ian

**Richard Cornford** · Jul 20 '05, 02:39 PM

Re: optimizing code...

"Ian Richardson" <zathras@chaos. org.uk> wrote in message
news:c0v842$1c5 qug$1@ID-99375.news.uni-berlin.de...
<snip>[color=blue]
>I've got an associative array of several hundred items used
>for language tokenisation,[/color]

Wit javascript the term "associativ e array" is usually used to describe
normal javascript object behaviour. What I think about when you say
"language tokenisation" is unlikely to anything even similar to what you
mean when you use the term.
[color=blue]
>and I've got a for(x in list) type loop to find items in
>it. This loop is potentially called over 600 times during the
>startup sequence!![/color]

That sounds like the worst possible way of doing something that might be
achieved with one or two operations or, at worst, a call to a short
method, with the right object design.
[color=blue]
>The code works but was one of those bits I originally implemented
>as a quick fix. Time to change this hideous system to something
>better, possibly an array of Options.[/color]

If you mean an array of objects created with the - Option -
constructors, that will probably not be a good idea as they are fairly
heavyweight objects (particularly on DOM browsers) when the information
being contained would be, at most, two text strings and two boolean
values.

But you have decided to talk about code without detail, example or
specification so there is nothing specific that can be said on the
subject. The chances were good that if you had provided the specifics of
the date, its storage and retrieval requirements, and possibly the code
that you are currently using (with some test case date), someone would
have already posted a javascript implementation optimised for the task
(and there would now be a discussion going on about if, and/or how, it
could be better).

<snip>[color=blue]
>I'm fully populating the lists, then quicksorting the list,
>then shuffling a specified item (which could be anywhere in
>the list) to the top, preserving the sorted order of the remainder.
>I could change this to:[/color]
[color=blue]
> 1. put item I know I want at the top of the list first
>
> 2. populate the list, checking that I don't put the first item
> in again.
>
> 3. quicksort the list from the second item to the end.[/color]

Either of those approaches would be bad. You should not be attempting to
sort the options list in a select element at all. You should keep the
addition, removal and relocation of DOM nodes to the absolute minimum
necessary. The data that is used to populate the options list should be
stored in a form optimised to the type of operations you intend to
perform on it and "sorted" in, or from, that storage.

<snip>[color=blue]
>Mind you, if it's quicker (in both IE and Mozilla) to build
>the list using DOM code then replacing the list instead of
>using lots of list[list.length] = new Option(blah, blah) I'm
>all in favour of it. Any suggestions?[/color]

optionsCollecti onReference[optionsCollecti onReference.len gth] = ...

Is the most reliable cross-browser code for appending an option element
to the options collection. Reliability concerns should override
consideration of the speed of the operation at this stage. One of the
reasons for doing this only when necessary.

<snip>[color=blue]
>I'm afraid I can't post the code now but ...[/color]
<snip>

Your choice. Good luck.

Richard.

**Ian Richardson** · Jul 20 '05, 02:40 PM

Re: optimizing code...

Richard Cornford wrote:[color=blue]
> "Ian Richardson" <zathras@chaos. org.uk> wrote in message
> news:c0v842$1c5 qug$1@ID-99375.news.uni-berlin.de...
> <snip>
>[color=green]
>>I've got an associative array of several hundred items used
>>for language tokenisation,[/color]
>
>
> Wit javascript the term "associativ e array" is usually used to describe
> normal javascript object behaviour. What I think about when you say
> "language tokenisation" is unlikely to anything even similar to what you
> mean when you use the term.[/color]

I was referring to "associativ e array" based upon the terminology used
in pages 130-132 of JavaScript: The Definitive Guide, 4th Edition.

I was creating an array like:

token["key_phrase_sta rt"] = "This is a key phrase to use on screen";
token["key_phrase_end "] = ", and here is another one."

....and I intend to read the browser language so I can pull in files like
blah_messages_< browser_languag e>.js, but only after pulling in
blah_messages_e n.js, so I've always got a fully populated token[] array
and can easily spot where a translated token is missing. I could do some
of this server-side but, apart from being off-topic for this group, I
still have to prove to my colleagues that it can be done at all!

[I'm doing this because I fully expect that, once the application is
released, translations for other languages will become a priority
requirement.]
[color=blue][color=green]
>>and I've got a for(x in list) type loop to find items in
>>it. This loop is potentially called over 600 times during the
>>startup sequence!![/color]
>
>
> That sounds like the worst possible way of doing something that might be
> achieved with one or two operations or, at worst, a call to a short
> method, with the right object design.[/color]

You're not wrong. I was using for/in to traverse the array in one safe
(albeit very slow) way, when I could (and did) switch to using:

if(typeof token[nameX] != "undefined" ) etc...

Result: HUGE speed improvement!

I like typeof, and there's no problem for me using it given the target
range of browsers I'm aiming at.
[color=blue]
> If you mean an array of objects created with the - Option -
> constructors, that will probably not be a good idea as they are fairly
> heavyweight objects (particularly on DOM browsers) when the information
> being contained would be, at most, two text strings and two boolean
> values.[/color]

Indeed, I've created my own objects which just have the two text strings
instead of using Option, and this has also resulted in a slight speed
improvement. Note I also discovered that by using Option constructors it
was trimming whitespace in the .text which I definitely didn't want.
[color=blue]
> But you have decided to talk about code without detail, example or
> specification so there is nothing specific that can be said on the
> subject. The chances were good that if you had provided the specifics of
> the date, its storage and retrieval requirements, and possibly the code
> that you are currently using (with some test case date), someone would
> have already posted a javascript implementation optimised for the task
> (and there would now be a discussion going on about if, and/or how, it
> could be better).[/color]

I understand this but I'm currently caught between a rock and a hard
place. If I could post it I would. If I could post example code I would
but, being married, I don't have the luxury of spare time at home to put
together examples which reflect the code I'm using in the application
I'm developing for my employer.
[color=blue]
> <snip>
>[color=green]
>>I'm fully populating the lists, then quicksorting the list,
>>then shuffling a specified item (which could be anywhere in
>>the list) to the top, preserving the sorted order of the remainder.
>>I could change this to:[/color]
>
>[color=green]
>>1. put item I know I want at the top of the list first
>>
>>2. populate the list, checking that I don't put the first item
>> in again.
>>
>>3. quicksort the list from the second item to the end.[/color]
>
>
> Either of those approaches would be bad. You should not be attempting to
> sort the options list in a select element at all. You should keep the
> addition, removal and relocation of DOM nodes to the absolute minimum
> necessary. The data that is used to populate the options list should be
> stored in a form optimised to the type of operations you intend to
> perform on it and "sorted" in, or from, that storage.[/color]

I can't guarantee that, for different languages, the lists will be
initially specified in the language tokenisation files in an optimum
order. However, when initially building the SELECT list there's nothing
to stop me creating a local array of these tokens and sorting them
before populating the list. I'll certainly try out this one.

Thanks,

Ian

**Berislav Lopac** · Jul 20 '05, 02:40 PM

Re: optimizing code...

Ian Richardson wrote:[color=blue]
> I'm writing a large Javascript application (uncompressed source around
> 400K) which is doing almost all the initialisation it needs to in a
> just-in-time manner. However, I have included an option for almost all
> of this to be done when the application first starts.
>
> Under this circumstance, and since the last few nightly builds of
> Mozilla, I've been getting:
>
>
> Script warning:
>
> A script on this page is causing mozilla to run slowly. If it
> continues
> to run, your computer may become unresponsive.
>
> Do you want to abort the script?[/color]

I my experience, thie above message usually appears when the program gets
stuck in an endless loops. Check your code for that first.

Berislav

--
If the Internet is a Marx Brothers movie, and Web, e-mail, and IRC are
Groucho, Chico, and Harpo, then Usenet is Zeppo.

**Richard Cornford** · Jul 20 '05, 02:43 PM

Re: optimizing code...

"Ian Richardson" <zathras@chaos. org.uk> wrote in message
news:c11s4p$1ah 0rv$1@ID-99375.news.uni-berlin.de...
<snip>[color=blue][color=green]
>>... the term "associativ e array" is usually used to
>>describe normal javascript object behaviour. ...[/color][/color]
<snip>[color=blue]
>I was referring to "associativ e array" based upon the
>terminology used in pages 130-132 of JavaScript: The
>Definitive Guide, 4th Edition.
>
>I was creating an array like:
>
>token["key_phrase_sta rt"] = "This is a key phrase to use on screen";
>token["key_phrase_end "] = ", and here is another one."[/color]

That is what I thought. "associativ e array" is used to describe an
application of javascript objects as a flexible means of storing
key/value pairs. It is not javascript terminology as such, and not
really different form borrowing "HashTable" from Java to describe the
concept. It suffers as terminology because it encourages people to use
Array objects for its implementation (which works because of the
objectness of Arrays) when the behaviour of Object objects is enough for
the task.

In many cases the initial definition of such an object would be better
achieved with an object literal:-

var token = {
key_phrase_star t:"This is a key phrase to use on screen",
key_phrase_end: ", and here is another one."
};

<snip>[color=blue]
>[I'm doing this because I fully expect that, once the
>application is released, translations for other languages
>will become a priority requirement.][/color]

Language negotiation is within the capabilities of the server and a task
that probably should be done there.

<snip>[color=blue]
>if(typeof token[nameX] != "undefined" ) etc...
>
>Result: HUGE speed improvement!
>
>I like typeof, and there's no problem for me using it
>given the target range of browsers I'm aiming at.[/color]

If the values in - token - are always non-empty strings then you will
probably achieve another performance improvement changing the - typeof -
test, with its string comparison operation, replacing it with a
type-converting test:-

if(token[nameX]){
...
}

- The - if - test must resolve its expression to a boolean value, which
it will do by internally type-converting the value of - token[nameX] -
to boolean. Non-empty strings type-convert to boolean true while
undefined values type-convert to boolean false.

<snip>[color=blue]
>I understand this but I'm currently caught between a rock
>and a hard place. If I could post it I would. If I could
>post example code I would but, being married, I don't have
>the luxury of spare time at home to put together examples
>which reflect the code I'm using in the application I'm
>developing for my employer.[/color]
<snip>

(copyright issues aside) time spent solving a problem associated with a
project should be a valid part of a software authoring process (and so
funded) even if that time is spent trying to make it easy for someone
else to solve the problem. I can appreciate that the management in
charge of the budget may not see it that way.

Richard.

**Ian Richardson** · Jul 20 '05, 02:43 PM

Re: optimizing code...

Richard Cornford wrote:

<snip>
[color=blue][color=green]
>>I was referring to "associativ e array" based upon the
>>terminology used in pages 130-132 of JavaScript: The
>>Definitive Guide, 4th Edition.
>>
>>I was creating an array like:
>>
>>token["key_phrase_sta rt"] = "This is a key phrase to use on screen";
>>token["key_phrase_end "] = ", and here is another one."[/color]
>
>
> That is what I thought. "associativ e array" is used to describe an
> application of javascript objects as a flexible means of storing
> key/value pairs. It is not javascript terminology as such, and not
> really different form borrowing "HashTable" from Java to describe the
> concept. It suffers as terminology because it encourages people to use
> Array objects for its implementation (which works because of the
> objectness of Arrays) when the behaviour of Object objects is enough for
> the task.
>
> In many cases the initial definition of such an object would be better
> achieved with an object literal:-
>
> var token = {
> key_phrase_star t:"This is a key phrase to use on screen",
> key_phrase_end: ", and here is another one."
> };[/color]

....which is almost identical to the way I'm now doing it.
[color=blue][color=green]
>>[I'm doing this because I fully expect that, once the
>>application is released, translations for other languages
>>will become a priority requirement.][/color]
>
> Language negotiation is within the capabilities of the server and a task
> that probably should be done there.[/color]

Indeed, but I'm doing it client-side as a proof of concept. Once the
application I'm developing has a real home at work (instead of just my
work PC and my home PC), I'll be able to start doing more work server-side.
[color=blue]
> If the values in - token - are always non-empty strings then you will
> probably achieve another performance improvement changing the - typeof -
> test, with its string comparison operation, replacing it with a
> type-converting test:-
>
> if(token[nameX]){
> ...
> }[/color]

If token[nameX] doesn't exist (which should never happen), then it'll
work but will show up as a Javascript warning in Mozilla; unacceptable
as far as I'm concerned. Mind you, if I can guarantee that token[nameX]
is always present there's no need to test for it in the first place ;-)

Ian

**Richard Cornford** · Jul 20 '05, 02:43 PM

Re: optimizing code...

"Ian Richardson" <zathras@chaos. org.uk> wrote in message
news:c17ll2$1eh 9k9$1@ID-99375.news.uni-berlin.de...
<snip>[color=blue][color=green]
>> if(token[nameX]){
>> ...
>> }[/color]
>
>If token[nameX] doesn't exist (which should never happen), then
>it'll work but will show up as a Javascript warning in Mozilla;
>unacceptable as far as I'm concerned. Mind you, if I can guarantee
>that token[nameX] is always present there's no need to test for it
>in the first place ;-)[/color]

If Mozilla have decided to generate a warning on type-converting tests
of undefined properties then whoever made that decision is an idiot. It
is completely acceptable and legal by ECMA 262 and a necessary (even
fundamental) practice in all cross-browser scripting.

Richard.

**Thomas 'PointedEars' Lahn** · Jul 20 '05, 02:44 PM

Re: optimizing code...

Richard Cornford wrote:
[color=blue]
> "Ian Richardson" <zathras@chaos. org.uk> wrote [...][color=green]
>>if(typeof token[nameX] != "undefined" ) etc...
>>
>>Result: HUGE speed improvement!
>>
>>I like typeof, and there's no problem for me using it
>>given the target range of browsers I'm aiming at.[/color]
>
> If the values in - token - are always non-empty strings then you will
> probably achieve another performance improvement changing the - typeof -
> test, with its string comparison operation, replacing it with a
> type-converting test:-
>
> if(token[nameX]){
> ...
> }
>
> - The - if - test must resolve its expression to a boolean value, which
> it will do by internally type-converting the value of - token[nameX] -
> to boolean. Non-empty strings type-convert to boolean true while
> undefined values type-convert to boolean false.[/color]

It is to be noted that it needs to be distinguished between a
property of type "undefined" with its sole value `undefined',
and an undefined property. Using the above test will trigger
a warning in the latter case, even an error if "token" was not
declared, while using the "typeof" operator will not in either
case (and misses backwards compatibility only before NN3).

So think twice before you use short and simple "if" statements
in favor of those longer ones using the "typeof" operator.

PointedEars

optimizing code...

optimizing code...

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment