Find all text between $...$ delimiters using bash script - regex

Question

I have a text file, and I'm trying to get an array of strings containing between $..$ delimiters (LaTeX formulas) using bash script. My current code doesn't work, result is empty:
#!/bin/bash
array=($(grep -o '\$([^\$]*)\$' test.txt))
echo ${array[#]}
I tested this regex here, it finds the matches. I use the following test string:
b5f1e7$bfc2439c621353$d1ce0$629f$b8b5
Expected result is
bfc2439c621353 629f
But echo returns empty. Although if I use '[0-9]\+' it works:
5 1 7 2439 621353 1 0 629 8 5
What do I do wrong?

Answer 1

How about:
grep -o '\$[^$]*\$' test.txt | tr -d '$'
This is basically performing your original grep (but without the brackets, which were causing it to not match), then removing the first/last characters from each match.

Answer 2

You may use awk with input field separator as $:
s='b5f1e7$bfc2439c621353$d1ce0$629f$b8b5'
awk -F '$' '{for (i=2; i<=NF; i+=2) print $i}' <<< "$s"
Note that this awk command doesn't validate input. If you want awk to allow for only valid inputs then you may use this gnu awk command with FPAT:
awk -v FPAT='\\$[^$]*\\$' '{for (i=1; i<=NF; i++) {gsub(/\$/, "", $i); print $i}}' <<< "$s"
bfc2439c621353
629f

Answer 3

What about this?
grep -Eo '\$[^$]+\$' a.txt | sed 's/\$//g'
I'm using sed to replace the $.

Answer 4

Try escaping your braces:
tst> grep -o '\$\([^\$]*\)\$' test.txt
$bfc2439c621353$
$629f$
of course, you then have to strip out the $ signs (-o prints the entire match). You can try sed instead:
tst> sed 's/[^\$]*\$\([^\$]*\)\$[^\$]*/\1\n/g' test.txt
bfc2439c621353
629f

Answer 5

Why is your expected output given b5f1e7$bfc2439c621353$d1ce0$629f$b8b5 the two elements bfc2439c621353 629f rather than the three elements bfc2439c621353 d1ce0 629f?
Here's a single grep command to extract those:
$ grep -Po '\$\K[^\$]*(?=\$)' <<<'b5f1e7$bfc2439c621353$d1ce0$629f$b8b5'
bfc2439c621353
d1ce0
629f
(This requires GNU grep as compiled with libpcre for -P)
This uses \$\K (equivalent to (?<=\$)to look behind at the first $ and (?=\$) to look ahead to the next $. Since these are lookarounds, they are not absorbed by grep in the process and therefore d1ce0 is available to be found.
Here's a single POSIX sed command to extract those:
$ sed 's/^[^$]*\$//; s/\$[^$]*$//; s/\$/\n/g' \
<<<'b5f1e7$bfc2439c621353$d1ce0$629f$b8b5'
bfc2439c621353
d1ce0
629f
This does not use any GNU notation and should work on any POSIX-compatible system (such as OS X). It removes the leading and trailing portions that aren't wanted, then replaces each $ with a newline.

Answer 6

Using bash regex:
var="b5f1e7\$bfc2439c621353\$d1ce0\$629f\$b8b5" # string to var
while [[ $var =~ ([^$]*\$)([^$]*)\$(.*) ]] # matching
do
echo -n "${BASH_REMATCH[2]} " # 2nd element has the match
var="${BASH_REMATCH[3]}" # 3rd is the rest of the string
done
echo # trailing newline
bfc2439c621353 629f