How to split a string in bash script

Question

Trying to extract some parts of a string into variables. Initial string looks like this: /dev/sdc2: LABEL="PAKAGE" UUID="9501DCBF768CEBB9" TYPE="ntfs" I getting correct PORT and LABEL values, but cant get the rest: UUID and TYPE. Instead of them I getting just a piece of it.

#!/bin/bash

INDEX1=""
INDEX2=""
LENGTH=""
PORT=""
LABEL=""
UUID=""
TYPE=""

line=`echo /dev/sdc2: LABEL="PAKAGE" UUID="9501DCBF768CEBB9" TYPE="ntfs" | sed s/\"//g`
echo $line

INDEX1=`expr index "$line" /dev/`
PORT=${line:$INDEX1-1:9}

INDEX1=`expr index "$line" LABEL=`
INDEX2=`expr index "$line" UUID`
INDEX1=`expr $INDEX1 + 5`
INDEX2=`expr $INDEX2 - 2`
LABEL=${line:$INDEX1:$INDEX2-$INDEX1}

INDEX1=`expr index "$line" UUID=`
INDEX2=`expr index "$line" TYPE`
INDEX1=`expr $INDEX1 + 4`
INDEX2=`expr $INDEX2 - 2`
UUID=${line:$INDEX1:$INDEX2-$INDEX1}

INDEX1=`expr index "$line" TYPE=`
INDEX1=`expr $INDEX1 + 4`
TYPE=${line:$INDEX1}

echo $PORT
echo $LABEL
echo $UUID
echo $TYPE


Here the output of this script:
/dev/sdc2: LABEL=PAKAGE UUID=9501DCBF768CEBB9 TYPE=ntfs
/dev/sdc2
PAKAGE
GE UUID=9501DCBF768CEBB9 T
KAGE UUID=9501DCBF768CEBB9 TYPE=ntfs

Where am I wrong?

It would be better/safer to not do this in bash at all: The Python shlex module (in the standard library) is a better lexer for shell syntax than anything bash has available built-in. Otherwise, you end up with approaches that either are dangerous (as with eval) or fragile (as with trying to roll your own parser). — Charles Duffy
– Charles Duffy, Commented Sep 27, 2013 at 20:15

biomancer · Accepted Answer · 2013-09-27 20:25:45Z

3

I think it's better done with regexp.

#!/bin/bash
src='/dev/sdc2: LABEL="PAKAGE" UUID="9501DCBF768CEBB9" TYPE="ntfs"'

regex='([^:]+): LABEL="([^"]+)" UUID="([^"]+)" TYPE="([^"]+)"'

if [[ $src =~ $regex ]]; then
    echo "matches"
    i=1
    n=${#BASH_REMATCH[*]}
    while [[ $i -lt $n ]]
    do
        echo "  capture[$i]: ${BASH_REMATCH[$i]}"
        let i++
    done
else
    echo "does not match"
fi

Output:

 matches
  capture[1]: /dev/sdc2
  capture[2]: PAKAGE
  capture[3]: 9501DCBF768CEBB9
  capture[4]: ntfs

Based on this.

edited Sep 27, 2013 at 20:25

answered Sep 27, 2013 at 19:39

biomancer

1,3041 gold badge9 silver badges20 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

glenn jackman Over a year ago

A more generic regex would be regex='([^:]+): LABEL="([^"]+)" UUID="([^"]+)" TYPE="([^"]+)"' -- that does not assume anything about the values of the variables

Chord Over a year ago

I do not think its possible with regexp due to different delimiters. But it really works just fine. Many thanks!

glenn jackman · Accepted Answer · 2013-09-27 20:11:44Z

1

This uses eval and is therefore risky. Ensure you know where your data is coming from.

$ line='/dev/sdc2: LABEL="this is the label" UUID="this is the uuid" TYPE="ntfs"'
$ IFS=: read port variables <<< "$line"
$ echo "$port"
/dev/sdc2
$ echo "$variables"
 LABEL="this is the label" UUID="this is the uuid" TYPE="ntfs"
$ eval "$variables"
$ printf "%s='%s'\n" label "$LABEL" uuid "$UUID" type "$TYPE"
label='this is the label'
uuid='this is the uuid'
type='ntfs'

answered Sep 27, 2013 at 20:11

glenn jackman

249k42 gold badges233 silver badges362 bronze badges

Comments

Rubens · Accepted Answer · 2013-09-27 20:47:05Z

1

Defining patterns over values, instead of keys, isn't the best way to deal with parsing. You might be better off doing it this way:

$ line='/dev/sdc2: LABEL="PAKAGE" UUID="9501DCBF768CEBB9" TYPE="ntfs"'
$ entry=( ${line//\"/} )

$ PORT=${entry[0]} PORT=${PORT//:/}     # defining variable PORT
$ declare ${entry[1]}                   # defining variable LABEL
$ declare ${entry[2]}                   # defining variable UUID
$ declare ${entry[3]}                   # defining variable TYPE    

$ echo -e "${PORT}\n${LABEL}\n${UUID}\n${TYPE}"
sdc2:
PAKAGE
9501DCBF768CEBB9
ntfs

edited Sep 27, 2013 at 20:47

answered Sep 27, 2013 at 20:28

Rubens

14.8k11 gold badges68 silver badges94 bronze badges

2 Comments

Rubens Over a year ago

@1_CR Mis-reverse-engineered the OP's example (: Thanks for pointing it out!

glenn jackman Over a year ago

You're assuming there's no whitespace in the quoted values.

iruvar · Accepted Answer · 2013-09-27 19:57:28Z

0

You could resort to simply clearing the string of all the undesirable portions and using read to populate your variables

read -r PORT LABEL UUID TYPE <<< "$(echo '/dev/sdc2: LABEL="PAKAGE" UUID="9501DCBF768CEBB9" TYPE="ntfs"' |\
sed 's/\(.*\): \+LABEL="\([^"]\+\)" \+UUID="\([^"]\+\)" \+TYPE="\([^"]\+\)"/\1 \2 \3 \4/')"
echo $PORT
/dev/sdc2
echo $LABEL
PAKAGE
echo $UUID
9501DCBF768CEBB9
echo $TYPE
ntfs

answered Sep 27, 2013 at 19:57

iruvar

23.5k7 gold badges58 silver badges83 bronze badges

2 Comments

glenn jackman Over a year ago

You are assuming there is no whitespace in any of the variable values.

iruvar Over a year ago

@glennjackman, you're right. I actually like your eval solution better

Gilles Quénot · Accepted Answer · 2013-09-27 21:17:28Z

0

Without eval, but same logic using bash :

$ line=\"/dev/sdc2: LABEL="PAKAGE" UUID="9501DCBF768CEBB9" TYPE="ntfs"\"
$ echo "${line%% *}"
/dev/sdc2
$ echo "$LABEL"
PAKAGE
$ echo "$UUID"
9501DCBF768CEBB9
$ echo "$TYPE"
ntfs

edited Sep 27, 2013 at 21:17

answered Sep 27, 2013 at 20:50

Gilles Quénot

188k43 gold badges232 silver badges229 bronze badges

2 Comments

glenn jackman Over a year ago

I don't get that output: I see "/dev/sdc2: and ntfs"

Gilles Quénot Over a year ago

Added bash requirement.

Collectives™ on Stack Overflow

How to split a string in bash script

5 Answers 5

2 Comments

Comments

2 Comments

2 Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

5 Answers 5

2 Comments

Comments

2 Comments

2 Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related