Only allow floating points regex

Question

How can I check if a specific string is a floating points? This are possible floating points:

12.245
+.0009
3.11e33
43.1E11
2e-14

This is what I tried:

grep "^[+\-\.0-9]" 
grep "^[+-]*[0-9]"
grep "^[+\-\.0-9]" 

And other lots of related things, but none filtered anything at all. Almost every string got through. How would I tackle this problem?

Answer 1

grep -xE '[-+]?[0123456789]*\.?[0123456789]+([eE][-+]?[0123456789]+)?'

With -x, we're anchoring the regexp at the start and and of the line so the lines have to match that pattern as a whole as opposed to the pattern being found anywhere in the line.

If you wanted to match on all the ones supported by POSIX/C strtod() as recognised by many implementations of the printf utility for instance:

r=[$(locale decimal_point)]
d=[0123456789]
h=[0123456789abcdefABCDEF]
grep -xE "[[:space:]]*[-+]?($d*$r?$d+([eE][-+]?$d+)?|\
0[xX]$h*$r?$h*([pP][-+]?$d+)?|\
[iI][nN][fF]([iI][nN][iI][tT][yY])?|\
[nN][aA][nN]|\
NAN\([abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789_]+\))"

So also including things like 0x3f, 0xFP-4, -Infinity, NAN(whatever).

$ printf '%g\n' 0x3f 0xFp-4 -Infinity 'NAN(whatever)'
63
0.9375
-inf
nan

Answer 2

Alternative Python solution (for less sophisticated input items):

Sample input.txt file:

11
12.245
+.0009
---0
3.11e33
43.1E11
2e-14
t12
aaa
10.001

---

check_float.py script:

import sys

with open(sys.argv[1], 'r') as inp:
    f = 'No'
    for l in inp.read().splitlines():
        try:
            if float(l) and '.' in l: f = 'Yes'
        except ValueError:
            f = 'No'
        finally:
            print '{0} - {1}'.format(l, f)

---

Usage:

python check_float.py input.txt

The output:

11 - No
12.245 - Yes
+.0009 - Yes
---0 - No
3.11e33 - Yes
43.1E11 - Yes
2e-14 - Yes
t12 - No
aaa - No
10.001 - Yes

Answer 3

Disclaimer: This is an imperfect solution. Perl's Scalar::Util::looks_like_number() function may not be the best choice of routine for doing this. See StéphaneChazelas comments below.

I'm leaving it here if anyone wants to look at it and pick the coproc bits out of it.

---

Rather than trying to craft your own regular expression to match the many possible floating point number format, use a library that already has implemented it:

perl -MScalar::Util -ne 'exit !Scalar::Util::looks_like_number($_)'

As a bash shell function:

is_number () {
    perl -MScalar::Util -ne 'exit !Scalar::Util::looks_like_number($_)' <<<"$1"
}

is_number hello && echo 'hello is a number'
is_number 1.234 && echo '1.234 is a number'

As a bash co-process (to avoid starting up a Perl process for every time you want to test a number):

coproc PERLIO=:raw perl -MScalar::Util -ne \
    'print Scalar::Util::looks_like_number($_) ? "Yes" : "No", "\n"'

while IFS= read -r -p 'Number please: ' possnum; do
    printf '%s\n' "$possnum" >&${COPROC[1]}
    read -u ${COPROC[0]}

    case "$REPLY" in
        Yes)    printf '%s is a number\n' "$possnum"       ;;
        No)     printf '%s is _not_ a number\n' "$possnum" ;;
    esac
done

kill "$COPROC_PID"

Or combining the two:

coproc PERLIO=:raw perl -MScalar::Util -ne \
    'print Scalar::Util::looks_like_number($_) ? "Yes" : "No", "\n"'

is_number () {
    printf '%s\n' "$1" >&${COPROC[1]}

    local REPLY
    read -u ${COPROC[0]}

    [ "$REPLY" = 'Yes' ] && return 0

    return 1
}

while IFS= read -r -p 'Number please: ' possnum; do
    if is_number "$possnum"; then
        printf '%s is a number\n' "$possnum"
    else
        printf '%s is a _not_ a number\n' "$possnum"
    fi
done

kill "$COPROC_PID"