FAQ Topic - How do I modify the current browser window?

[Dr J R Stockton]

In comp.lang.javascript message <[email protected]>,

Lines: 127

VK said the following on 3/20/2007 1:19 PM:

Are you really that anal?
...

The previous FAQ maintainer wasted everybody's time in foolish argument,
instead of incorporating useful suggestions into a published FAQ. Stop
heading in the same direction.

You really need a new hobby. BTW, where is that rounding function? :-X

<URL:http://www.merlyn.demon.co.uk/js-misc0.htm#CDC> now shows
conversion from an expression (using covert eval) to a Number and then
(in two Forms) either to 4 bytes or 4 words shown in binary respectively
representing the value as an IEEE Single or to an IEEE Double. Those
displays are editable and can be converted both back to Number (shown by
default toString) and on to *exact* representations of the sign,
exponent, mantissa, and complete value as decimal strings. AFAIK, it's
OK for all finite values up to or almost up to the limit of the type;
but it could do with more testing.

In Doubles :-
0.01 +0.01000000000000000020816681711721685132943093776702880859375
0.06 +0.059999999999999997779553950749686919152736663818359375
0.06+0.01 +0.06999999999999999278355033993648248724639415740966796875
0.07 +0.070000000000000006661338147750939242541790008544921875
Ruler ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^

The resolution of a Double (Number) is about 1 in 9e15.

Seeing that might help VK understand the situation.

It's a good idea to read the newsgroup c.l.j and its FAQ. See below.

 

[VK]

VK said the following on 3/20/2007 1:19 PM:



Are you really that anal?

No, I'm trying to show what is an algorithm / specification and what
is garbage. I'm not expecting an immediate result though maybe I was
too rushing the events in the previous post. Array vs Object took
about 3 year, window vs Global about 2 years. So in one year you might
learn what is "anal retentive" clarification for clarification itself
- and what is a necessary editing of a badly written algorithm.
Everything takes time, sometimes more that one would like.

And you got a very definitive answer to your question.

It doesn't matter - see above. If your extra answer was needed to read
the FAQ 5.1 then automatically FAQ 5.1 is garbage as a technical text.
However simple this concept may be, it may take good time to be fully
realized. As I said I'm not pushing anymore the events quicker than
they can go. As a really small compromise you might repost in this
thread your statement of irrelevance of case and other attributes of
<FAQ***RY> - only w/o flying mice, rear ends, anal problems and other
irrelevant though pictorial elements. This way it could be possible to
make an authoritative reference without scaring sh** out of people.

Hold on there Tonto. I am not going to re-word and entry that has
*never* had a problem from anyone other than you simply to add some
pedantic garbage about whether case matters or not and whether a closing
tag is included or not. When I do FAQ Updates I simply search for the
letters FAQENTRY using Google Groups

<snip>

That is the problem we are discussion. No one gives a "flying mouse
read end" of what are _you_ doing. The question is what _users_ are
instructed to do.

You really need a new hobby. BTW, where is that rounding function? :-X

The theory is almost there. As I said before the javascript rounding
question turned out to be a side outcome of my personal structural
linguistics researches - something I did not expect when I was
starting it. I guess the community may "starve" a bit more on the
default round() method, the sky will not fail on the earth because of
that.

If you have a spare time then
http://groups.google.com/group/comp.lang.javascript/msg/b1470812c5e4d3b5
is still waiting for your attention. This problem is _times_ much more
often and important than some 1.0000x rounding.
A rather extensive explanation of the screen update delay is already
given, but I left some more researches to do: like testing "soft
break" usability extends. It also may be useful to shorten the FAQ to
the minimum but to make an external link illustrating the UAs screen
update delay in all details. So there is plenty of important work for
a few days at the very least for yourself and the possible volunteers.

 

[VK]

<URL:http://www.merlyn.demon.co.uk/js-misc0.htm#CDC> now shows
conversion from an expression (using covert eval) to a Number and then
(in two Forms) either to 4 bytes or 4 words shown in binary respectively
representing the value as an IEEE Single or to an IEEE Double. Those
displays are editable and can be converted both back to Number (shown by
default toString) and on to *exact* representations of the sign,
exponent, mantissa, and complete value as decimal strings. AFAIK, it's
OK for all finite values up to or almost up to the limit of the type;
but it could do with more testing.

In Doubles :-
0.01 +0.01000000000000000020816681711721685132943093776702880859375
0.06 +0.059999999999999997779553950749686919152736663818359375
0.06+0.01 +0.06999999999999999278355033993648248724639415740966796875
0.07 +0.070000000000000006661338147750939242541790008544921875
Ruler ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^

The resolution of a Double (Number) is about 1 in 9e15.

Seeing that might help VK understand the situation.

That's hardly. I'm not sure at all what this - sorry if sounds too
rough - a la Big Math toying may be good for: but itself and
especially for the precision questions.

Actually I never asked you what Dr. are you and if you are Dr. at all
- and it's not my damn business anyway while on the Usenet. But from
your web-site and some of your comments I may conclude that you are -
or pretending to be - in some higher math related science. In this
case you are always welcome aboard - the math nature badly tolerate
"closed" researches.
Well, my semantical symmetriades may stay with me until published
if ever but they are not in relevance whatsoever to any javascript
matters; all precision-related stuff is open right now.

To get you on the stream:

1) Problem One:
What real numbers can be handled exactly without symbolic
substitutions and within a finite algorithm? Obviously the one having
(0) in period. What reals are these? The formula is done but the
strong proof is still in process:
http://groups.google.com/group/sci.math/msg/be0bf2131d0f0c97
http://groups.google.com/group/sci.math/msg/bead11fea343cf3a

2) Problem Two:
A real with (0) in period could be considered as full set of decadic
rationals - bearing in mind that the relevant lemma for Problem One is
still under investigation. But a decadic rational with (0) in period
easily becomes an infinite binary sequence with a period other than
(0) so it can be stored exactly in decimal form but lead to precision
loss in binary form. Just for a sample: 0.1
Only decadic rationals having equivalent representation as dyadic
rationals retain (0) in period so can be stored exactly. The exact set
formula is already done:
http://groups.google.com/group/sci.math/msg/d9cbdee772dfb835

Here comes where you may help if you want to:

3) Problem Three:
The above two problems were resolved for an ideal Turing machine, so a
finite algorithm was the only requirement, storage limitations
disregarded. Obviously it is not true for IEEE-754 FP-DP where we have
54 bits (including the implicit one) for mantissa storage space. This
way for the set resulted from two "filters" above: for this set we
have to find an algorithm to check that:
Real N being transformed from decimal to binary form will result in
bit sequence no longer than 54 bit from the first bit set to 1 and to
the last bit set to 1 with border bits included. Exponent part value
is irrelevant.
Algorithms to transform decimals to binary are well known, so if you
like a "combo" of math and javascript programming then here is a good
field.

If you feel like more of usability and best choice studies then you
may jump onto the "contextual rounding" problem. That's a nasty one
without (?) obvious solutions. For explanations and real life samples
read:
http://groups.google.com/group/comp.arch.arithmetic/msg/ce0907e3159cb1a6
http://groups.google.com/group/comp.arch.arithmetic/msg/d33556d6f79562a4

P.S. I linked only relevant individual messages. Reading whole threads
may be useful as well but up to you.

P.P.S. All of above is a free choice proposal of course.