The Naked Scientists

The Naked Scientists Forum

Author Topic: Borland C++ builder log() bug?  (Read 4654 times)

Offline Boogie

  • Full Member
  • ***
  • Posts: 63
    • View Profile
Borland C++ builder log() bug?
« on: 21/08/2012 23:47:04 »
It appears that Borland C++ compiler (version 6) has a bug with the log() function.

For example, a compiled application using log(15) returns 2.708. My calculator and OpenOffice agree that log(15) = 1.706

Anyone else encounter this? I fought this problem almost all day thinking I was doing something wrong!!!


 

Offline RD

  • Neilep Level Member
  • ******
  • Posts: 8128
  • Thanked: 53 times
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #1 on: 22/08/2012 00:27:39 »
Logs can have different bases ...

Quote
The logarithm to base b = 10 is called the common logarithm and has many applications in science and engineering. The natural logarithm has the constant e (≈ 2.718) as its base; its use is widespread in pure mathematics, especially calculus. The binary logarithm uses base b = 2 and is prominent in computer science.
https://en.wikipedia.org/wiki/Logarithm

For example, a compiled application using log(15) returns 2.708.

That looks like log to base e 

« Last Edit: 22/08/2012 00:40:25 by RD »
 

Offline CliffordK

  • Neilep Level Member
  • ******
  • Posts: 6321
  • Thanked: 3 times
  • Site Moderator
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #2 on: 22/08/2012 00:46:44 »
Also see the change of base formula in the Wikipedia article above.

https://en.wikipedia.org/wiki/Logarithm#Change_of_base

It looks complicated, but it turns out to be  relatively simple calculation because "b" is a constant, so you just calculate the constant divisor.  If Borland c++ doesn't already have the function, you could write a quick function to change the base to the desired base (10?)
 

Offline Boogie

  • Full Member
  • ***
  • Posts: 63
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #3 on: 22/08/2012 02:07:47 »
Thanks Clifford, I agree. I think Borland got them backwards. The base 10 command : log10(x) produces 1.176.

This drove me nuts. The log function was only a small part of the algorithm I was working on and it took me forever to finally suspect it. I did not expect this sort of library bug with a version six release.

I just noticed that I really "fat fingered" the return value in my initial post. :-[
 

Offline CliffordK

  • Neilep Level Member
  • ******
  • Posts: 6321
  • Thanked: 3 times
  • Site Moderator
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #4 on: 22/08/2012 06:16:18 »
I don't think it is a Borland thing, but rather a C++ thing...  and perhaps Standard C.

http://www.cplusplus.com/reference/clibrary/cmath/log/

Probably somebody 30 years ago decided to use Log(x) as the natural log, rather than using ln(x).
Then, to make each version of the program more or less backward compatible, the old syntax was pushed forward.

In Unix, ln means to "link" files.
 

Offline Boogie

  • Full Member
  • ***
  • Posts: 63
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #5 on: 22/08/2012 16:33:54 »
Yeah, I think you're right. Unless I'm just getting confused, Dynamic C appears to be doing the same thing.

One solution :

#define LOG         log10   //lib is wrong
#define LOG10     log       //lib is wrong

Now when I use LOG(x) it replaces it with log10(x). LOG10(x) is replaced by log(x)

I was a little suprised that this worked. It's a little anal, but it might save me some head scratching if I need to go back and trouble shoot the algorithm sometime in the future.
 

Offline CliffordK

  • Neilep Level Member
  • ******
  • Posts: 6321
  • Thanked: 3 times
  • Site Moderator
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #6 on: 22/08/2012 19:05:16 »
Yeah, I think you're right. Unless I'm just getting confused, Dynamic C appears to be doing the same thing.

One solution :

#define LOG         log10   //lib is wrong
#define LOG10     log       //lib is wrong

Now when I use LOG(x) it replaces it with log10(x). LOG10(x) is replaced by log(x)

I was a little suprised that this worked. It's a little anal, but it might save me some head scratching if I need to go back and trouble shoot the algorithm sometime in the future.

I would be careful with such a definition.
If someone else is likely to work on your program in the future, it is a sure method to spell disaster.

It would be extremely confusing to define the natural log, Loge(x) to be written as Log10(x).

If ln(x) is not used elsewhere, you might choose to define it as your natural log function.

Then you would merely define log(x) and log10(x) as meaning the same thing.  Or...  you could define log10(x) to throw an error so you could remember what you had done.
« Last Edit: 22/08/2012 19:53:56 by CliffordK »
 

Offline Boogie

  • Full Member
  • ***
  • Posts: 63
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #7 on: 22/08/2012 20:15:38 »
If ln(x) is not used elsewhere, you might choose to define it as your natural log function.

Yes! That is a much better solution. Good point, thanks!

I knew confusion would get to me sooner or later.
 

Offline CliffordK

  • Neilep Level Member
  • ******
  • Posts: 6321
  • Thanked: 3 times
  • Site Moderator
    • View Profile
Re: Borland C++ builder log() bug?
« Reply #8 on: 22/08/2012 20:40:55 »
Ahhh
Are you using caps in your #define functions?

It may be a good mnemonic, but could also lead to a mess with using both upper and lower case log functions.

I think I'd just do a find & replace, and be done with it.  Perhaps remember to double check your work for a bit until you get used to the way natural logs are used in C++.

I must admit, however, that I was one of those people that would often define "dir" and "ls" to mean the same thing.
 

The Naked Scientists Forum

Re: Borland C++ builder log() bug?
« Reply #8 on: 22/08/2012 20:40:55 »

 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums