Approximate comparison of two lists of floats

[Peter Otten]

Christian Doll wrote:

> i have e little performance problem with my code...
> 
> i have to compare many lists of very much floats. at moment i have
> nested for-loops
> 
> for a in range( len(lists) ):
>     for b in range( a+1 , len(lists) ):
>         for valuea in lists[a]:
>             equal=False
>             for valueb in lists[b]:
>                 if inTolerance( valuea , valueb , 1.0): # inTolerance
> is an own function, which checks if the difference of valuea and
> valueb is not more then 1.0%
>                     equal=True
>                     break
>     if equal:
>         print a , "and" , b , "are equal"

My crystal ball says that if you profile the above you'll find that the 
above spends most of the time in the inTolerance() function that you don't 
provide.
 
> i found a version with set which is faster, but i cannot assign an
> tolerance (%)
> for a in range( len(lists) ):
>     for b in range( a+1 , len(lists) ):
>         if len( lists[a] ) ==
> len( set( lists[a] ).intersection( set( lists[b] ) ) ):
>             print a , "and" , b , "are equal"

I can't think of a problem that can be solved with that ;)
 
> have you an idea how i can change my code, that i can compare many
> lists of floats with a tolerance in percentage very fast?

You can usually speed up number-crunching tasks with numpy, but it looks 
like you don't have a clear notion what vectors should be regarded as equal.

Perhaps you can provide a bit of background information about the problem 
you are trying to solve with the code, preferably in plain (if bad) english.