Smalltalk › Squeak › Squeak - Dev

[BUG] Complex equality problem

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

5 messages Options

Nicolas Cellier-3

[BUG] Complex equality problem

Testing equality
aNumber = (aNumber + 0 i)
will answer false...

Testing equality
(aNumber + 0 i) = aNumber
will answer true...

(1 + 0 i) = 1.
1 = (1 + 0 i).

This is due to premature test
isNumber ifFalse [^false].

Obviously, Complex cannot answer true to isNumber because most senders of this
message will trust faithfully isKindOf: Number, and Complex is not in this
hierarchy.

The test should be some sort of
isArithmeticValue ifFalse: [^false]
and Complex Quaternion and other arithmetic classes extending numbers should
answer true to isArithmeticValue.

This is another clue that VW ArithmeticValue superclass of Number make sense.
And another clue that Complex (ComplexNumber please) are not much used.

Jerome Peace

Re: [BUG] Complex equality problem

Hi Nicolas,

I am very happy to see your attention to language
(isComplex) and detail(this issue.) on the list.

+1 on both issues.

It would be good if you could familiarize yourself
with Mantis.

http://bugs.impara.de/my_view_page.php

is a good place to start. Ask for a new account.
When you have it bring each of these issues up
(seprately with links to the other).

This way the issue can be tracked; will have staying
power until its resolved; and not be lost in the mail
stream.

Great bug finding.

Yours in service, -- Jerome Peace

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

stéphane ducasse-2

Re: [BUG] Complex equality problem

I suggest that you write SUnit tests to document your intention and
that we can follow
what you are doing.

> Hi Nicolas,
>
> I am very happy to see your attention to language
> (isComplex) and detail(this issue.) on the list.
>
> +1 on both issues.
>
> It would be good if you could familiarize yourself
> with Mantis.
>
> http://bugs.impara.de/my_view_page.php
>
> is a good place to start. Ask for a new account.
> When you have it bring each of these issues up
> (seprately with links to the other).
>
> This way the issue can be tracked; will have staying
> power until its resolved; and not be lost in the mail
> stream.
>
> Great bug finding.
>
> Yours in service, -- Jerome Peace
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>

Nicolas Cellier-3

[BUG] Complex equality problem

In reply to this post by Nicolas Cellier-3

One more evidence of Complex equality problem.
(Am i sole user of Complex in the world ?).

try:
1 i = nil.

I am joining a SUnit test (already existed, it is a enhanced version).

I am joining what i think a good patch.
I use #isNumberExtension testing implemented in Object Number and Complex.
(this would extend naturally to quaternions, octonions, ...)

This will replace the isNumber test in Number subclass >> =
(needed against the 1 = (1 + 0 i) bug)

Since the bug is already published on mantis how do i let them know about this
fix ?

ComplexEqualPatch.1.cs (2K) Download Attachment

ComplexEqualTest.3.cs (892 bytes) Download Attachment

Ken Causey-3

Re: [BUG] Complex equality problem ([closed] already on Mantis ID 2688)

In reply to this post by Nicolas Cellier-3

http://bugs.impara.de/view.php?id=2688

On Wed, 2006-02-08 at 00:19 +0100, nicolas cellier wrote:

> Testing equality
> aNumber = (aNumber + 0 i)
> will answer false...
>
> Testing equality
> (aNumber + 0 i) = aNumber
> will answer true...
>
> (1 + 0 i) = 1.
> 1 = (1 + 0 i).
>
> This is due to premature test
> isNumber ifFalse [^false].
>
> Obviously, Complex cannot answer true to isNumber because most senders of this
> message will trust faithfully isKindOf: Number, and Complex is not in this
> hierarchy.
>
> The test should be some sort of
> isArithmeticValue ifFalse: [^false]
> and Complex Quaternion and other arithmetic classes extending numbers should
> answer true to isArithmeticValue.
>
> This is another clue that VW ArithmeticValue superclass of Number make sense.
> And another clue that Complex (ComplexNumber please) are not much used.
>
>
>

signature.asc (196 bytes) Download Attachment