"just a bug" !; Mathematica 4
- To: mathgroup at smc.vnet.net
- Subject: [mg27435] "just a bug" !; Mathematica 4
- From: pl10-mac <pl10 at st-and.ac.uk>
- Date: Sun, 25 Feb 2001 20:55:38 -0500 (EST)
- Sender: owner-wri-mathgroup at wolfram.com
Hi,
I've just noticed this issue : Integrate[Sqrt[(r - x)(r + x)], {x, -r, r}]
Okay - so version 3 works but version 4 doesn't.
I've read all the eloquent remarks, but the bottom line is surely this:
A maths package costing well over 1,000 pounds and written by some of the finest
mathematicians and programmers in the world - should be able to integrate
Integrate[Sqrt[(r - x)(r + x)], {x, -r, r}]
Pete Lindsay
> From: David Withoff <withoff at wolfram.com>
To: mathgroup at smc.vnet.net
> Date: Sun, 25 Feb 2001 00:53:20 -0500 (EST)
> To: mathgroup at smc.vnet.net
> Subject: [mg27435] [mg27396] [mg27396] Re: [mg27364] Re: A bug of Integrate[] in
> Mathematica 4.1 (and 4.0)
>
>>>> Thank you for the report. Integrate[] appears to be behaving
>>>> erroneously in all of the cases you present. I have notified
>>>> the the developers of the problems. Unfortunately, there does
>>>> not appear to be a workaround.
>>>
>>>> Unfortunately, there are no bug lists available, and I will
>>>> not be informed of the status of the bug until it is fixed.
>>>> Most likely, a patch will not be made available.
>>
>> I think this behaviour of Wolfram Research should be made public
>> somewhere on the web! Why do these guys not care about BUGS? A software
>> developer should fix his bugs and make this bug fixes available to the
>> users of his product.
>
> The unfortunate misstatements quoted above from technical support are
> at best misleading, and basically are just plain wrong. For that, we
> certainly apologize. Unfortunately, now that this misinformation is on
> the web, it has a life of its own.
>
> It is of course abundantly false that there are no bug lists, no bug
> workarounds, that people are never notified of bug fixes, and so forth.
> Many if not most of the items in the support.wolfram.com web site
> (the bug list) describe bugs or other behaviors that people find
> troublesome, and many of those items include fixes and workarounds.
> This and other information is also frequently distributed in other ways,
> both to technical support people and to users.
>
> In fact, if one strips away the generalizations and takes a close
> look at the specific examples that have been raised here, it turns
> out that in all of these cases solutions have already been provided
> and/or the problems are already discussed in the bug list and/or
> the reported behaviors are not really bugs at all.
>
> Regarding the following particular examle:
>
>> Another annoying bug in Integrate is its handling of branch cuts in real
>> integrals. A "nice" example is the area of a half circle. You may try if
>> there has something changed in version 4.1 compared to 3.0 or 4.0.
>> Reporting this bug they did not admit that it is one at all because
>> Integrate would do "complex integrals". After my reply that then there
>> should be a possibility to define the path of integration in the complex
>> plane I never got a convincing answer ;-(. So here is my example:
>>
>> Correct is the following:
>>
>> In[1]:== Integrate[Sqrt[r^2-x^2],{x,-r,r},Assumptions->{r>0}]
>>
>> 2
>> Pi r
>> Out[1]== ----- (correct area of a half circle!)
>> 2
>>
>> Now doing the same integral with a mathematically identical integrand:
>>
>> In[1]:== Integrate[Sqrt[(r-x)(r+x)],{x,-r,r},Assumptions->{r>0}]
>>
>> 2
>> Pi r Sqrt[r ]
>> Out[1]== -------------
>> 4
>>
>> In[2]:== Simplify[%,r>0]
>>
>> 2
>> Pi r
>> Out[2]== ----- (WRONG factor 1/2)
>> 4
>>
>> The reason is that Mathematica simply puts the boundaries which are on
>> the branch points of the integrand in the indefinite integrals which are
>> along branchcuts without proper I epsilon-descriptions (well known in
>> quantum theory)
>
> that class of bugs is described in the bug list
> (http://support.wolfram.com/Kernel/Symbols/System/Integrate.html).
>
> If someone didn't want to admit that this was a bug it was probably
> because they weren't sure, and didn't want to admit to something
> without knowing if it was true. If someone said specifically that
> this wasn't a bug, or tried to offer some explanation involving
> complex integrals (or quantum theory) then they were mistaken.
> It's just a bug. Probably it will be fixed in the next major release.
> Development of algorithms to do this sort of thing is a very difficult
> problem in mathematics, and yes it can be annoying that this problem
> has not yet been solved. As soon as all problems have been solved
> then we can all go on vacation.
>
> Dave Withoff
> Wolfram Research
>
>