Test for a number in bash

Question

Is this the correct way to test for a number, with double [[]] enclosing :digit: and single quotes surrounding the regex ?

if [[ "$var" =~ '^[[:digit:]]+$' ]]; then

1.234 and -66 are both numbers. Do you want to test for non-negative integers? — Chris Davies, Jan 29 '23 at 08:12

score 1 · Accepted Answer · answered Jan 29 '23 at 04:14

It is not, as a simple test will show us:

#!/bin/bash
for var in 1 2 3 a b c; do
  if [[ "$var" =~ '^[[:digit:]]+$' ]]; then
    echo "$var is a number"
  else
    echo "$var is NOT a number"
  fi
done

Running the above produces:

1 is NOT a number
2 is NOT a number
3 is NOT a number
a is NOT a number
b is NOT a number
c is NOT a number

If you have an editor that runs shellcheck automatically, or if you paste your code into https://shellcheck.net/, you'll see the following error:

Line 4:
        if [[ "$var" =~ '^[[:digit:]]+$' ]]; then
                        ^-- SC2076 (warning): Remove quotes from right-hand side of =~ to match as a regex rather than literally.

And that tells us we should instead write:

#!/bin/bash
for var in 1 2 3 a b c; do
    if [[ "$var" =~ ^[[:digit:]]+$ ]]; then
        echo "$var is a number"
    else
        echo "$var is NOT a number"
    fi
done

Which produces:

1 is a number
2 is a number
3 is a number
a is NOT a number
b is NOT a number
c is NOT a number

This is documented in the bash man page:

An additional binary operator, =~, is available, with the same precedence as == and !=. When it is used, the string to the right of the operator is considered a POSIX extended regular expression and matched accordingly... If any part of the pattern is quoted, the quoted portion is matched literally. This means every character in the quoted portion matches itself, instead of having any special pattern matching meaning.

"44.7 is NOT a number" and "-73 is NOT a number" - your test only considers non-negative integers — Chris Davies, Jan 29 '23 at 08:14
Depends on the version of bash and whether the compat31 option is enabled. — Stéphane Chazelas, Jan 29 '23 at 09:09
@roaima It's true, this only checks for integers, which appears to be what the OP was trying to do. — larsks, Jan 29 '23 at 13:36
I want only to test for integers as I want the values to refer to the context number in grep. — Vera, Jan 29 '23 at 20:31

Stéphane Chazelas · Answer 2 · 2023-01-29T09:36:19.737

Depends what you mean by number. 0x12, inf, 1.2, -12, ¹²³, ⑱, 1e6, 1,000,000, Ⅷ, 0x123p-5, 6,23, 0b100110 are all numbers by some definitions of number, and 08, 019, 9999999999999999999999999999999999 are not valid numbers by some others.

To test whether a string is made of one or more ASCII decimal digits, you can use:

Standardly:

valid() {
  case $1 in
    ("" | *[!0123456789]*) false;;
    (*) true;;
  esac
}

Or in ksh, bash (bash -O extglob with older versions) or zsh -o kshglob:

valid() [[ $1 = +([0123456789]) ]]

Or in bash / ksh93 / zsh / yash:

valid() [[ $1 =~ ^[0123456789]+$ ]]

valid() [[ $1 =~ '^[0123456789]+$' ]] would also work in zsh¹ and bash 3.1 or bash -O compat31. In zsh, if the rematchpcre option was enabled, you'd need to change it to [[ $1 =~ '^[0123456789]\z' ]] as $ in PCRE matches at the end of the subject but also before a newline character at the end of the subject so would match on $'123\n' for instance. You could also use [[ $1 =~ '^\d+\z' ]] then.

In zsh:

valid() [[ $1 = <-> ]]

In any case, avoid [[:digit:]] which on some systems matches decimal digits other than the ASCII arabic ones. And except in zsh, and especially in bash, do not use [0-9] for input validation as that often match random characters that happen to sort between 0 and 9.

^{¹ Note however that in zsh, variable values can contain the NUL byte and the standard regexp API (as opposed to PCRE's) doesn't support matching strings with NULs in them. When the rematchpcre option is not enabled, [[ $1 =~ '^[0123456789]+$' ]] would match on $'123\0xyz' for instance.}

Test for a number in bash

2 Answers2