Suddenly, a NaN error in Netscape

[Adam]

Hi!

I've modified a javascript order form calculation script that tallies
up the subtotal, shipping/handling, and total automatically.

I did some modifications and it still worked fine, even in Netscape,
but at some point the Shipping/Handling field has ceased to display
correctly in that browser, showing a Not A Number error. (In Safari it
looks fine, and it calculates correctly in all browsers I've tested it
with.)

http://wiseowlmultimedia.com/learnin...s_orderUS.html

I'm no java genius, just a copy-paste-and-tinker boy, so I've
basically got no clue what I messed up. My fine-tooth comb has turned
up no reason I can see why it broke. Can anyone help?

Thanks!
Adam

PS - It does act a bit peculiar in IE - I found I needed to click on
all the subtotal fields to make them show anything - but I suppose
that's just a java quirk of Explorer...

[Richard Cornford]

Michael Winter wrote:
<snip>

It seems that Netscape is correct; there are invalid characters
in the file. When I examined it with a binary editor, the
indentation is made with character code 160 (0xA0). I don't know
why, but either you inadvertantly inserted them, your editor did,
or your news reader did when you copied my post. Check the file
format your editor uses. Make sure it uses plain or ANSI text.

<snip>

Not a newsreader but Developersdex's web interface, and probably many
other web interfaces. Rather than doing the obvious thing and displaying
news messages in PRE tags to preserve the formatting inherent in the
original content, numerous web interfaces attempt to re-format posts as
HTML, padding indentation with &nbsp: and inserting <br> at the and of
lines. Obviously copying &nbsp; (0xA0) will transfer it literally into
some text editors (not unreasonably).

Developersdex used to be quite a reasonable web interface to Usenet but
they have gone down hill very fast recently. The writing was probably on
the wall the moment they decided to start spamming anyone who started a
new thread in a group they covered. But they also seem to have done
something to reduce the overall quality of their users. In the past
questions originating from Developersdex seemed, in the most part, to be
from developers with at least some understanding of what they were
doing. These days the average post originating on Developersdex is on a
par with a Forum4designers post.

Richard.

[Michael Winter]

On Thu, 12 Aug 2004 16:08:56 +0100, Richard Cornford
<Ri*****@litote s.demon.co.uk> wrote:

[snip]

Not a newsreader but Developersdex's web interface, [...]
Obviously wasn't paying attention when I deleted the Developersdex
signature. :|

[snip]
Obviously copying &nbsp; (0xA0) will transfer it literally into
some text editors (not unreasonably).

I didn't realise that 0xA0 was &nbsp;. That would certainly explain it.

[snip]

Thanks,
Mike

--
Michael Winter
Replace ".invalid" with ".uk" to reply by e-mail

[Richard Cornford]

Michael Winter wrote:

Richard Cornford wrote:

<snip>

Obviously copying &nbsp; (0xA0) will transfer it
literally into some text editors (not unreasonably).

I didn't realise that 0xA0 was &nbsp;. That would
certainly explain it.

Well, I have seen it before. Opera <= 6 won't process scripts that
contains 0xA0 (as whitespace) either. It is one of those problems where
positing code to Usenet just gets a lot of "it works in browser X for
me" responses and you can't see why it wouldn't work, or tell what is
going on until a URL is available.

Richard.

[Adam Abrams]

Invisible characters were indeed the source of the trouble!

Just as a test, I ran my faulty page through BBEdit's "Zap Gremlins"
filter
- and then, lo and behold, it worked. Brilliant!

(I had pasted the code from the website where I was viewing your reply
straight into GoLive, then uploaded. I won't do it that way again,
that's
for certain!)

I did need to modify the form to set all the quantity fields to "0", to
avoid
generating an error for each empty Quantity field. That being done, it
works like a charm.

Again, many many thanks for taking the time to work this whole thing
out for me. Besides enabling me to create the kind of page I envisioned,
the code will act as a studying point in improving my comprehension of
Javascript.

All the best!

Adam
http://www.adamabrams.com/

*** Sent via Developersdex http://www.developersdex.com ***
Don't just participate in USENET...get rewarded for it!

[Dr John Stockton]

JRS: In article <opsclaiqbkx13k vk@atlantis>, dated Wed, 11 Aug 2004
22:37:48, seen in news:comp.lang. javascript, Michael Winter <M.Winter@bl
ueyonder.co.inv alid> posted :

In addition to the modified checks, I removed the rounding when
calculating the total. As 'q' will always be an integer, the value will
never go beyond two decimal places, making the rounding redundant.
I've not checked in your code; but you might consider whether the
observation that 10% of the lines given by
for (j=1 ; j<=100 ; j++) document.writel n(j*0.01, ' ', j, '<br>')
give more than two digits after the decimal point is pertinent.

ISTM that only if the factor, 0.01 above, is a multiple of 0.25 can one
rely on getting only two displayed digits. However, it seems that most
cases do give just two digits; therefore, a single test is likely to
pass.

Indeed, 3*0.1 and 0.1*3 both give me 0.3000000000000 0004 .


In article <opsck9qouox13k vk@atlantis>, dated Wed, 11 Aug 2004 22:21:04,
seen in news:comp.lang. javascript, Michael Winter <M.******@bluey onder.c
o.invalid> posted :
It should be:

/^\d+(\.\d{2})?$/

Fixed, thanks.

--
© John Stockton, Surrey, UK. ?@merlyn.demon. co.uk Turnpike v4.00 IE 4 ©
<URL:http://jibbering.com/faq/> JL / RC : FAQ for news:comp.lang. javascript
<URL:http://www.merlyn.demo n.co.uk/js-index.htm> jscr maths, dates, sources.
<URL:http://www.merlyn.demo n.co.uk/> TP/BP/Delphi/jscr/&c, FAQ items, links.

[Michael Winter]

On Thu, 12 Aug 2004 13:30:45 +0100, Dr John Stockton
<sp**@merlyn.de mon.co.uk> wrote:

[Example of inexactitude of floats in JavaScript]

I sometimes forget (when it matters, it seems) that floating-point
representations , in every form I should imagine, are approximates.

Although the errant bits probably won't interfere in even the extreme
cases, the rounding should remain, but in a better form. That includes
changing toCurrency() to take pennies:

function totalItem(i, e) {
/* Get references to the form controls. */
var oP = e['Item' + i + 'Price'],
oQ = e['Item' + i + 'Quantity'],
oT = e['Item' + i + 'Total'];
/* Get the values (convert them later). */
var p = oP.value, q = oQ.value, t;

/* Check that price is a well-formatted
* decimal. If not, quit. */
if(!/^(([1-9]\d*)|0)\.\d{2}$/.test(p)) {return;}
/* Convert price to number and multiply to work in
* pennies, reducing chance of rounding errors. */
p = Math.round(+p * 100);

/* Check that quantity is a positive
* integer. If not, quit. */
if(!/^([1-9]\d*)|0$/.test(q)) {return;}

/* Calculate total (in pennies). */
t = Math.round(p * q);

/* Convert the total to a formatted
* string and update the form. */
oT.value = toCurrency(t, '$', ',');
return t;
}

function totalOrder() {
var e = document.forms['orderform'].elements,
n = 40,
t = 0,
s;

/* Loop through items 1..n, computing the item
* total and adding it to the order total. */
for(var i = 1; i <= n; ++i) {
var x = totalItem(i, e);

/* Check for an error. */
if('undefined' == typeof x) {
alert('Item ' + i + ' contains an invalid value.\n' +
'Please correct it and try again.');
e['Item' + i + 'Quantity'].focus();
return;
}
t += x;
}
/* Less than $100, add 15%; $100 or more, add 10%.
* Note: totalItem() returns pennies. */
s = Math.round(t * ((t < 10000) ? 0.15 : 0.1));
/* Update the final totals. */
e['OrderTotal'].value = toCurrency(t, '$', ',');
e['SH'].value = toCurrency(s, '$', ',');
e['GrandTotal'].value = toCurrency(t + s, '$', ',');
}

/* n - Number to format (in pennies).
* c - Currency symbol to use.
* g - Grouping symbol.
*
* Outputs a number of the form cngnnngnnn.nn
*
* For example, toCurrency(1426 35.7, '£', ',') produces
* £1,426.36
*/
function toCurrency(n, c, g) {
var s = (0 > n) ? '-' : ''; n = Math.abs(n);
var m = String(Math.rou nd(n));
var j, i = '', f;

while(m.length < 3) {m = '0' + m;}
f = m.substring((j = m.length - 2));
while(j > 3) {
i = g + m.substring(j - 3, j) + i;
j -= 3;
}
i = m.substring(0, j) + i;
return s + c + i + '.' + f;
}

Hopefully, that's the end of the matter.

[snip]

Many thanks for these continued corrections,
Mike

--
Michael Winter
Replace ".invalid" with ".uk" to reply by e-mail

[Michael Winter]

On 12 Aug 2004 16:04:15 GMT, Adam Abrams <ad********@sha w.ca> wrote:

[snip]

I did need to modify the form to set all the quantity fields to"0", to
avoid generating an error for each empty Quantity field.That being done,
it works like a charm.
Hmm, yes. I overlooked that as well. I apologise.
Again, many many thanks for taking the time to work this wholething out
for me. Besides enabling me to create the kind of page Ienvisioned, the
code will act as a studying point in improving mycomprehension of
Javascript.

It's no problem.

If you want to improve your JavaScript, I don't think there are any places
better than this newsgroup. There are many knowledgeable people here and
they can teach you a lot.

Mike

--
Michael Winter
Replace ".invalid" with ".uk" to reply by e-mail

[Dr John Stockton]

JRS: In article <opscmtyocgx13k vk@atlantis>, dated Thu, 12 Aug 2004
18:34:39, seen in news:comp.lang. javascript, Michael Winter <M.Winter@bl
ueyonder.co.inv alid> posted :

I sometimes forget (when it matters, it seems) that floating-point
representation s, in every form I should imagine, are approximates.

Add to your imaginer a decimal float notation.

The mantissa of an IEEE Double is generally considered as binary, but
could equally well be considered as Hex.

Now instead of those Hex digits use decimal ones; there are (about?) 13
of them.

Any sum of money up to 10^13 pence can now be rendered exactly.

The signed exponent is 11 bits, which would support a range of +-400;
more than enough to keep Mr Brown happy, at least for a while.
ISTM that if CPUs had hardware instructions for true decimal types, both
integer and float (which nowadays would need only a little more
silicon), and software writers implemented such types, much time could
be saved.

--
© John Stockton, Surrey, UK. ?@merlyn.demon. co.uk Turnpike v4.00 IE 4 ©
<URL:http://jibbering.com/faq/> JL / RC : FAQ for news:comp.lang. javascript
<URL:http://www.merlyn.demo n.co.uk/js-index.htm> jscr maths, dates, sources.
<URL:http://www.merlyn.demo n.co.uk/> TP/BP/Delphi/jscr/&c, FAQ items, links.