maximum value in a column of file

Fredrik Lundh fredrik at pythonware.com
Wed Jul 23 11:44:50 CEST 2008


maurizio wrote:

> i tryed to use the module max of numpy,
> the problem is that i don't know how to put the column of the file in an 
> array.
> (i'm new in phyton).
> anyway if you think there is a better way.....

What kind of file is it?  Did you pick numpy because you want to do 
matrix operations (beyond just finding a maximum value), or was it just 
the first thing you stumbled upon when researching the problem?

A simple pattern for finding the maximum value in a file, using only 
plain Python code, is:

     max_value = ... some very small value ...
     for line in file:
          value = ... extract value from line ...
          if value > max_value:
              max_value = value

If it's not obvious what "some very small value" is, given the range of 
data you're working with, you can do

     max_value = None
     for line in file:
          value = ... extract value from line ...
          if max_value is None or value > max_value:
              max_value = value

instead (this leaves max_value set to None if the file is empty)

A more experienced Python hacker might write

     def get_all_values(file):
         for line in file:
             value = ... extract value from line ...
             yield value

     ...

     max_value = max(get_all_values(file))

instead.  But this still leaves us with the problem of extracting the 
value.  The best way to do that depends on the kind of files you're 
working with; for fixed-format text files, you could use string slicing 
and int/float for conversion (e.g. "value = float(line[10:20])", see the 
tutorial for details); for other text formats, you could use 
split/partition or regular expressions, or maybe an existing module 
(such as "csv"); for binary formats, there's a large
number of existing tools.

And if you really want to use numpy for other reasons than just getting 
a maximum value from a column, there's plenty of stuff in NumPy and 
SciPy (http://www.scipy.org/) that might be useful.

So, in other words, I guess we still need more info.

</F>




More information about the Python-list mailing list