How to detect a floating point number using a regular expression

user_135078 0
Viewed 47298

What is a good regular expression for handling a floating point number (i.e. like Java's Float)

The answer must match against the following targets:

 1) 1.  
 2) .2   
 3) 3.14  
 4) 5e6  
 5) 5e-6  
 6) 5E+6  
 7) 7.e8  
 8) 9.0E-10  
 9) .11e12

In summary, it should

  • ignore preceding signs
  • require the first character to the left of the decimal point to be non-zero
  • allow 0 or more digits on either side of the decimal point
  • permit a number without a decimal point
  • allow scientific notation
  • allow capital or lowercase 'e'
  • allow positive or negative exponents

For those who are wondering, yes this is a homework problem. We received this as an assignment in my graduate CS class on compilers. I've already turned in my answer for the class and will post it as an answer to this question.

[Epilogue] My solution didn't get full credit because it didn't handle more than 1 digit to the left of the decimal. The assignment did mention handling Java floats even though none of the examples had more than 1 digit to the left of the decimal. I'll post the accepted answer in it's own post.

floating-point regex
7 Answers

Just make both the decimal dot and the E-then-exponent part optional:

[1-9][0-9]*\.?[0-9]*([Ee][+-]?[0-9]+)?

I don't see why you don't want a leading [+-]? to capture a possible sign too, but, whatever!-)

Edit: there might in fact be no digits left of the decimal point (in which case I imagine there must be the decimal point and 1+ digits after it!), so a vertical-bar (alternative) is clearly needed:

(([1-9][0-9]*\.?[0-9]*)|(\.[0-9]+))([Ee][+-]?[0-9]+)?
user_958100
Admin1

Here is what I turned in.

(([1-9]+\.[0-9]*)|([1-9]*\.[0-9]+)|([1-9]+))([eE][-+]?[0-9]+)?

To make it easier to discuss, I'll label the sections

( ([1-9]+ \. [0-9]* ) | ( [1-9]* \. [0-9]+ ) | ([1-9]+))  ( [eE] [-+]? [0-9]+ )?     
--------------------------------------------------------  ----------------------    
                           A                                       B

A: matches everything up to the 'e/E'
B: matches the scientific notation

Breaking down A we get three parts

 ( ([1-9]+ \. [0-9]* ) | ( [1-9]* \. [0-9]+ ) | ([1-9]+) )
   ----------1----------   ---------2----------   ---3----

Part 1: Allows 1 or more digits from 1-9, decimal, 0 or more digits after the decimal (target 1)
Part 2: Allows 0 or more digits from 1-9, decimal, 1 or more digits after the decimal (target 2)
Part 3: Allows 1 or more digits from 1-9 with no decimal (see #4 in target list)

Breaking down B we get 4 basic parts

 ( [eE] [-+]? [0-9]+  )?   
   ..--1- --2-- --3--- -4- ..

Part 1: requires either upper or lowercase 'e' for scientific notation (e.g. targets 8 & 9)
Part 2: allows an optional positive or negative sign for the exponent (e.g. targets 4, 5, & 6)
Part 3: allows 1 or more digits for the exponent (target 8)
Part 4: allows the scientific notation to be optional as a group (target 3)

user_1350780

@Kelly S. French, this regular expression matches all your test cases.

^[+-]?(\d+\.\d+|\d+\.|\.\d+|\d+)([eE][+-]?\d+)?$

Source: perldoc perlretut

user_50576730 
'([-+])?\d*(\.)?\d+(([eE]([-+])?)?\d+)?'

That's the regular expression I have arrived at when trying to solve this kind of task in Matlab. Actually, it won't correctly detect numbers like (1.) but some additional changes may solve the problem... well, maybe the following would fix that:

'([-+])?(\d+(\.)?\d*|\d*(\.)?\d+)(([eE]([-+])?)?\d+)?'
user_30422350
Related