[SOLVED] Help with bash regex

linuxuser371038 · 05-18-2024, 03:26 AM

Example output of mmcli -L command:

Code:

/org/freedesktop/ModemManager1/Modem/0 [QUALCOMM INCORPORATED] 0

In the script I got for checking sms from online someone had written a regex but it is not right for my new modem as it only searched 0-9 and for some reason when trying last night the modem number was into the 20s. It back down at 0 today. I think it was because I was pugging and unplugging quite a bit. So I would like to regex double digits.

Original regex line:

Code:

MODEMNO=$(mmcli -L | grep -o "Modem/[0-9]" | grep -o [0-9]$)

One I cobbled together however it produces 3 outputs:

Code:

MODEMNO=$(mmcli -L|grep -Eo '[[:digit:]]+')

The above produces several outputs, which would, if using the example output from above, be:

Code:

1
0
0

I only want the middle number. 0 in this case but factor for any any number into the double digits. It was late last night and ran out of steam. I realized I could probably simplify with using a regex which uses other parts of the string but only picking the number part. This would have been rudimentary a while ago but I have forgotten it!

I gather I could also use grouping with the existing one and choose the middle but, again, forgotten how to do groupings.

Also I read that grep is not required at all and the regex could be done right in a bash variable. I would be interested in knowing that too and just generally the simplest way but I my attention was waning last night and so stopped and posting where I got up to.

I haven't coded in about a year and it was embarrassing to notice how rusty I was. Then again bash and regex were only periphery to my main language of python.

Petri Kaukasoina · 05-18-2024, 03:51 AM

Code:

MODEMNO=$(mmcli -L|sed 's_.*Modem/__;s_ .*__')

MadeInGermany · 05-18-2024, 04:57 AM

The consequent fix on the original is:

Code:

MODEMNO=$( mmcli -L | grep -Eo "Modem/[0-9]+" | grep -Eo "[0-9]+$" )

With GNU grep -P it shrinks to

Code:

MODEMNO=$( mmcli -L | grep -Po "(?<=Modem/)[0-9]+" )

or

Code:

MODEMNO=$( mmcli -L | grep -Po "Modem/\K[0-9]+" )

With sed:

Code:

MODEMNO=$( mmcli -L | sed -nE "s#.*Modem/([0-9]+).*#\1#p" )

pan64 · 05-18-2024, 09:43 AM

Code:

answer=( $(mmcli -L) )
echo ${answer##*/}

no sed, no grep, no pipe, just pure bash.

Code:

b=( $(mmcli -L) )
answer="${b[0]}"                  # interesting, it works without it too.
echo ${answer##*/}

MadeInGermany · 05-18-2024, 11:27 AM

@pan64,
the unquoted $(mmcli -L) tries filename generation (disable with set -f unless you know that nmcli -L never produces a wildcard).

Another (and safe) solution with bash builtins:

Code:

[[ "$( mmcli -L )" =~ Modem/([0-9]+) ]] ; MODEMNO=${BASH_REMATCH[1]}

pan64 · 05-18-2024, 11:35 AM

Quote:

Originally Posted by MadeInGermany

@pan64, you didn't see the text after the modem/0 ?
And what if it would contain another / character?
Also the unquoted $(mmcli -L) tries filename generation (disable with set -f unless you know that nmcli -L never produces a wildcard).

Another (and safe) solution with bash builtins:

Code:

[[ "$( mmcli -L )" =~ Modem/([0-9]+) ]] ; MODEMNO=${BASH_REMATCH[1]}

You are right, I would not accept that too. mmcli -L will list modems, () will split the output, and actually the script will just cut everything before the last / (of the first word). Using the sample by OP it works, in general it is not really safe.
If I need to vote probably I would prefer sed. Bash is not really good at regex.

MadeInGermany · 05-18-2024, 12:00 PM

Lately seeing that the following text is separated by whitespace, I have edited my post.
Only my point with the filename generation is valid.

I think bash uses the Posix ERE engine, while sed -E uses the ERE engine with some GNU extensions (like \s).
Bash stores the BASH_REMATCH[1] after a match-only, while sed can only use the \1 immediately, within the s command.

grail · 05-19-2024, 02:03 AM

In second example of assigning b array to answer, you could of course do the expansion in one go:

Code:

b=( $(mmcli -L) )
echo "${b[0]##*/}"

grail · 05-19-2024, 02:04 AM

In second example of assigning b array to answer, you could of course do the expansion in one go:

Code:

b=( $(mmcli -L) )
echo "${b[0]##*/}"

linuxuser371038 · 05-19-2024, 05:48 AM

Quote:

Originally Posted by MadeInGermany

The consequent fix on the original is:

Code:

MODEMNO=$( mmcli -L | grep -Eo "Modem/[0-9]+" | grep -Eo "[0-9]+$" )

With GNU grep -P it shrinks to

Code:

MODEMNO=$( mmcli -L | grep -Po "(?<=Modem/)[0-9]+" )

or

Code:

MODEMNO=$( mmcli -L | grep -Po "Modem/\K[0-9]+" )

With sed:

Code:

MODEMNO=$( mmcli -L | sed -nE "s#.*Modem/([0-9]+).*#\1#p" )

You are missing the part though where I said I want between 0-99? Which is the whole impetus for wanting to change the original.

I have not tried each yet but aren't the rest also missing the part where I wanted 0-99? Since I see only single digit so far.

Apologies if some do factor this as I didn't test each one yet but have not seen it mentioned yet.

pan64 · 05-19-2024, 06:26 AM

that is the + after [0-9], which means: one ore more of it
see regex quantifiers

linuxuser371038 · 05-19-2024, 12:26 PM

Quote:

Originally Posted by pan64

Code:

answer=( $(mmcli -L) )
echo ${answer##*/}

no sed, no grep, no pipe, just pure bash.

Code:

b=( $(mmcli -L) )
answer="${b[0]}"                  # interesting, it works without it too.
echo ${answer##*/}

Thanks, I just tried this. Clean and succinct. Could I have an explainer?

Reminds me of how in python I would try and cram loads of nested list comprehensions just for fun even if it looked unwieldy.

pan64 · 05-20-2024, 01:56 AM

what is missing?

Code:

mmcli -L             # this is your command
$( )                 # will execute the command and returns the output of it
( content*)          # will split the content into an array (which is actually the output of your command)
answer=(..)          # assign the array to the variable "answer" (or b)
answer="${b[0]}"     # assign the first element of array b to the variable answer (which is now not an array)
${answer##*/}        # remove everything from the string until the last / in the variable answer