What's the best way to get a number as input from python while maintaining type

I'm trying to get a number as CLI input from python. Valid input should be either an int or a float and I need t开发者_高级运维o maintain type. So validating an int and returning a float wouldn't work.

This is the best thing I've been able to come up with and it's not all that good.

def is_valid(n):
    try:
        if '.' in n: return float(n)
        return int(n)
    except ValueError:
        print "try again"

def num_input(s):
    n = raw_input(s)
    while is_valid(n) is None:
        n = raw_input(s)
    return is_valid(n)

valid_num = num_input("Enter a valid number: ")

Clearly this isn't the best way.

use a for loop to try all the conversions, I added complex type for demonstration:

def is_valid(n):
    for t in (int, float, complex):
        try:
            return t(n)
        except ValueError:
            pass
    raise ValueError("invalid number %s" % n)

print is_valid("10")
print is_valid("10.0")
print is_valid("1+3.0j")

def num_input(prompt, error):
    while True:
        result = raw_input(prompt)
        for candidate in (int, float):
            try: return candidate(result)
            except ValueError: pass
        print error

Try using the decimal module which will allow you to maintain exactly the precision of the entered number eg:

import decimal

def num_input(s):
    while True:
        try:
            return decimal.Decimal(raw_input(s))
        except decimal.InvalidOperation, e:
            print e.message


valid_num = num_input("Enter a decimal number: ")

print 'ANSWER: ', valid_num

See: http://docs.python.org/library/decimal.html

After considering the early answers and thinking about it a bit more the solution I came up with is:

def num_input(prompt, error):
    s = raw_input(prompt)
    for t in (int, float, complex):
        try: return t(s)
        except ValueError: pass
    print error
    return num_input(prompt, error) #better get it in the first 1k tries

I really want an input function not just a validation function; however, I think HYRY's suggestion to loop over int, float, complex is a good one. I took win's suggestion to use recursion instead of looping, understanding that a really confused and persistent user could exceed the max recursion depth. Although I don't need it now, I think Karl Knechtel is correct in making the error an arg instead of hard coded.