Thread: [Matplotlib-users] log(0) behaviour

After "from pylab import *", log(0.0) gives

Warning: Encountered divide by zero(s)  in log

Where is this behaviour documented?   What does it mean?   Why should =
log(0.0) divide by zero?   What does the plural "zeros" refer to?   Why =
break the IEEE standard that log(0.0) should silently return -Inf?

Keith

PS: at http://matplotlib.sourceforge.net/pylab_commands.html, "it's =
size" should be "its size."

kei...@bt... wrote:
> After "from pylab import *", log(0.0) gives
> 
> Warning: Encountered divide by zero(s)  in log
> 
> Where is this behaviour documented?   What does it mean?   Why should log(0.0) divide by zero?   What does the plural "zeros" refer to?   Why break the IEEE standard that log(0.0) should silently return -Inf?

http://www.opengroup.org/onlinepubs/000095399/functions/log.html

Look at the section "ERRORS", specifically "Pole Error". By and large the
implementation of log() comes from the underlying libm. The plural is because
log() is a ufunc and can take arrays, possibly with multiple zeros.

-- 
Robert Kern
rob...@gm...

"I have come to believe that the whole world is an enigma, a harmless enigma
 that is made terrible by our own mad attempt to interpret it as though it had
 an underlying truth."
  -- Umberto Eco

On Apr 25, 2006, at 6:59 AM, <kei...@bt...> wrote:

> After "from pylab import *", log(0.0) gives
>
> Warning: Encountered divide by zero(s)  in log
>
> Where is this behaviour documented?   What does it mean?   Why  
> should log(0.0) divide by zero?   What does the plural "zeros"  
> refer to?   Why break the IEEE standard that log(0.0) should  
> silently return -Inf?

Adding to Robert's reply, I don't believe that the standard mandates  
that programs should silently return Inf in the presence of divide by  
zero errors. It does allow for  programs to determine how they want  
to respond to such errors, and numpy allows you to  customize that.  
You can ignore them, print a warning (the default I believe), or  
raise an exception. Believe it or not, not everyone wants the same  
behavior for all cases.

Perry