How to extract numbers from a string?

Question

I have string contains a path

string="toto.titi.12.tata.2.abc.def"

I want to extract only the numbers from this string.

To extract the first number:

tmp="${string#toto.titi.*.}"
num1="${tmp%.tata*}"

To extract the second number:

tmp="${string#toto.titi.*.tata.*.}"
num2="${tmp%.abc.def}"

So to extract a parameter I have to do it in 2 steps. How to extract a number with one step?

Answer 1

You can use tr to delete all of the non-digit characters, like so:

echo toto.titi.12.tata.2.abc.def | tr -d -c 0-9

Answer 2

To extract all the individual numbers and print one number word per line pipe through -

tr '\n' ' ' | sed -e 's/[^0-9]/ /g' -e 's/^ *//g' -e 's/ *$//g' | tr -s ' ' | sed 's/ /\n/g'

Breakdown:

• Replaces all line breaks with spaces: tr '\n' ' '
• Replaces all non numbers with spaces: sed -e 's/[^0-9]/ /g'
• Remove leading white space: -e 's/^ *//g'
• Remove trailing white space: -e 's/ *$//g'
• Squeeze spaces in sequence to 1 space: tr -s ' '
• Replace remaining space separators with line break: sed 's/ /\n/g'

Example:

echo -e " this 20 is 2sen\nten324ce 2 sort of" | tr '\n' ' ' | sed -e 's/[^0-9]/ /g' -e 's/^ *//g' -e 's/ *$//g' | tr -s ' ' | sed 's/ /\n/g'

Will print out

20
2
324
2

Answer 3

Here is a short one:

string="toto.titi.12.tata.2.abc.def"
id=$(echo "$string" | grep -o -E '[0-9]+')

echo $id // => output: 12 2

with space between the numbers. Hope it helps...

Answer 4

Parameter expansion would seem to be the order of the day.

$ string="toto.titi.12.tata.2.abc.def"
$ read num1 num2 <<<${string//[^0-9]/ }
$ echo "$num1 / $num2"
12 / 2

This of course depends on the format of $string. But at least for the example you've provided, it seems to work.

This may be superior to anubhava's awk solution which requires a subshell. I also like chepner's solution, but regular expressions are "heavier" than parameter expansion (though obviously way more precise). (Note that in the expression above, [^0-9] may look like a regex atom, but it is not.)

You can read about this form or Parameter Expansion in the bash man page. Note that ${string//this/that} (as well as the <<<) is a bashism, and is not compatible with traditional Bourne or posix shells.

Answer 5

Convert your string to an array like this:

$ str="toto.titi.12.tata.2.abc.def"
$ arr=( ${str//[!0-9]/ } )
$ echo "${arr[@]}"
12 2

Answer 6

This would be easier to answer if you provided exactly the output you're looking to get. If you mean you want to get just the digits out of the string, and remove everything else, you can do this:

d@AirBox:~$ string="toto.titi.12.tata.2.abc.def"
d@AirBox:~$ echo "${string//[a-z,.]/}"
122

If you clarify a bit I may be able to help more.

Answer 7

You can also use sed:

echo "toto.titi.12.tata.2.abc.def" | sed 's/[0-9]*//g'

Here, sed replaces

• any digits (class [0-9])
• repeated any number of times (*)
• with nothing (nothing between the second and third /),
• and g stands for globally.

Output will be:

toto.titi..tata..abc.def

Answer 8

Use regular expression matching:

string="toto.titi.12.tata.2.abc.def"
[[ $string =~ toto\.titi\.([0-9]+)\.tata\.([0-9]+)\. ]]
# BASH_REMATCH[0] would be "toto.titi.12.tata.2.", the entire match
# Successive elements of the array correspond to the parenthesized
# subexpressions, in left-to-right order. (If there are nested parentheses,
# they are numbered in depth-first order.)
first_number=${BASH_REMATCH[1]}
second_number=${BASH_REMATCH[2]}

Answer 9

Using awk:

arr=( $(echo $string | awk -F "." '{print $3, $5}') )
num1=${arr[0]}
num2=${arr[1]}

Answer 10

Hi adding yet another way to do this using 'cut',

echo $string | cut -d'.' -f3,5 | tr '.' ' '

This gives you the following output: 12 2

Answer 11

Fixing newline issue (for mac terminal):

cat temp.txt | tr '\n' ' ' | sed -e 's/[^0-9]/ /g' -e 's/^ *//g' -e 's/ *$//g' | tr -s ' ' | sed $'s/ /\\\n/g'

Answer 12

Assumptions:

• there is no embedded white space
• the string of text always has 7 period-delimited strings
• the string always contains numbers in the 3rd and 5th period-delimited positions

One bash idea that does not require spawning any subprocesses:

$ string="toto.titi.12.tata.2.abc.def"

$ IFS=. read -r x1 x2 num1 x3 num2 rest <<< "${string}"
$ typeset -p num1 num2
declare -- num1="12"
declare -- num2="2"

In a comment OP has stated they wish to extract only one number at a time; the same approach can still be used, eg:

$ string="toto.titi.12.tata.2.abc.def"

$ IFS=. read -r x1 x2 num1 rest <<< "${string}"
$ typeset -p num1
declare -- num1="12"

$ IFS=. read -r x1 x2 x3 x4 num2 rest <<< "${string}"
$ typeset -p num2
declare -- num2="2"

A variation on anubhava's answer that uses parameter expansion instead of a subprocess call to awk, and still working with the same set of initial assumptions:

$ arr=( ${string//./ } )
$ num1=${arr[2]}
$ num2=${arr[4]}
$ typeset -p num1 num2
declare -- num1="12"
declare -- num2="2"