Formula for beginners
Doctor: Peter, I finally did it!
Peter: What is it this time? What kind of pointless invention did you come up with now?
Doctor: I came up with a revolutionary method for processing formulas with a computer. Take a look at this table.
Normal notationDoctor's "revolutionary" notation
1 + 21 2 +
3 * 4 + 73 4 * 7 +
10 / ( 2 - 12 )	10 2 12 - /
( 3 - 4 ) * ( 7 + 2 * 3 )3 4 - 7 2 3 * + *
Peter: Huh?
Doctor: Hahaha. I guess it's still too complicated for you, a beginner. This is where it gets interesting.
Peter: Uh, what...?
Doctor: You know that computers have a data structure called a stack, right? It's the one where you put in things in first and take them out last.
Peter: Yes, I know, but...
Doctor: This revolutionary notation uses that stack. For example, in 10 2 12 - /, the processing goes like this:
Object10212-/
 ↓↓↓↓2-12↓10/-10
Stack
.
.
10
.
2
10
12
2
10
.
-10
10
.
.
-1
Doctor: See? You don't need to worry about parentheses or operator precedence. The word order is "divide 10 by the result of subtracting 2 from 12." It sounds a bit like Japanese, doesn't it? With this revolutionary invention, our research lab is safe and sound. Hahaha.
Peter: Uh, Doctor? This is something we learned in the basic course at Aizu University when I was in Japan. They called it "Reverse Polish Notation," and everyone in the class could program it easily.
Doctor: ...
And so, Peter had to teach the program to the Doctor. Create a program that takes a formula written in Reverse Polish Notation as input and outputs the result of the calculation.
Input
Multiple datasets will be given. In each dataset, one line will be given, consisting of a formula written in Reverse Polish Notation (a string of up to 80 characters consisting of integers and operation symbols separated by a single space (half-width) character). No formulas that divide a value by 0 or a value close to 0 will be given.
The number of datasets is no more than 50.
Output
For each dataset, output the result of the calculation (a real number) on one line. The calculation result may include an error of up to 0.00001.
Sample Input
10 2 12 - /
3 4 - 7 2 3 * + *
-1 -2 3 + +
Output for the Sample Input
-1.000000
-13.000000
0.000000
