Reg Ex for specific number in string

Question

I'd like to match numbers (int and real) in a string, but not if they are part of an identifier; eg, i'd like to match 5.5 or 42, but not x5. Strings are roughly of the form "x5*1.1+42*y=40". So far, I came up with

([0-9]*[.])?[0-9]+[^.*+=<>]

This correctly ignores x0, but also 0 or 0.5 (12.45, however, works). Changing the + to * leads to wrong matchings.

It would be very nice if someone could point out my error.

Thanks!

Answer 1

This is actually not simple. Float literals are more complex than you assumed, being able to contain an e or E for exponential format. Also, you can have prefixed signs ( + or - ) for the number and/or the exponent. All in all it can be done like this:

re.findall(r'(?:(?<![a-zA-Z_0-9])|[+-]\s*)[\d.]+(?:[eE][+-]?\d+)?',
           'x5*1.1+42*y=40+a123-3.14e-2')

This returns:

['1.1', '+42', '40', '-3.14e-2']

You should consider though whether a thing like 4+3 should lead to ['4', '3'] or ['4', '-3'] . If the input was 4+-3 the '-3' would clearly be preferable. But to distinguish these isn't easy and you should consider using a proper formula parser for these.

Maybe the standard module ast can help you. The expression must be a valid Python expression in this case, so a thing like a+b=40 isn't allowed because left of the equal sign is no proper lvalue . But for valid Python objects you could use ast like this:

import ast

def find_all_numbers(e):
  if isinstance(e, ast.BinOp):
    for r in find_all_numbers(e.left):
      yield r
    for r in find_all_numbers(e.right):
      yield r
  elif isinstance(e, ast.Num):
    yield e.n

list(find_all_numbers(ast.parse('x5*1.1+42*y-40').body[0].value))

Returns:

[1.1, 42, 40]

Answer 2

You could do it with something like

\b\d*(\.\d+)?\b

It matches any number of digits ( \\d* ) followed by an optional decimal part ( (\\.\\d+)? ). The \\b matches word boundaries , ie the location between a word character and a non word character. And since both digits and (english) letters are word characters , it won't match the 5 in a sequence like x5 .

See this regex101 example .

The main reason your try fails is that it ends with [^.*+=<>] which requires the number (or rather match ) to end with a character other than . , * , = , + , < or > . And when ending with a single digit, like 0 and 0.5 , the digit gets eaten by the [0-9]+ , and there's nothin to match the [^.*+=<>] left, and thus it fails. In the case with 12.45 it first matches 12.4 and then the [^.*+=<>] matches the 5 .

Answer 3

Do something like ((?<![a-zA-Z_])\\d+(\\.\\d+)?)

It is using negative lookbehind in order not to select anything having [a-zA-Z_] prior to it. Check it out here in Regex101 .

About your regex ([0-9]*[.])?[0-9]+[^.*+=<>] use [0-9]+ instead of [0-9]* as it will not allow .05 to be captured, only 0.5. Another thing is [^.*+=<>] this part, you could add ? to the end of it in order to allow it not to have characters as well. Example 1.1 wont be captured as ([0-9]*[.])?[0-9]+ is satisfied but not [^.*+=<>] that comes after it as well.