I'm using
sed (GNU sed) 4.4
GNU bash, version 4.4.12(1)-release (x86_64-pc-linux-gnu)
I've a complicated set of sed arguments in a bash script, about 20 different -e expression. Here is a simple example as a one-liner. It converts aa bb cc to aaBBcc:
sed -e 's# ##g' -e 's#b#B#g' <<< "aa bb cc"
or
k=('-e' 's# ##g' '-e' 's#b#B#g'); sed "${k[@]}" <<< "aa bb cc"
However, there are 20-ish -e expression and most are complicated. The script is only for me, so it doesn't have to follow convention or policy. To make the arguments readable / editable (to me), I assign them to a variable with extra whitespace, columnated, indented .... Here is a simplified version of what I mean:
#!/bin/bash
k="-e s# # #g \
-e s# b # B #g \
"
That simplified example doesn't show how useful that approach is to me. Anyway, here is the "working" script:
#!/bin/bash
k="-e s# # #g \
-e s# b # B #g \
"
k=$(sed -e 's# ##g' <<< "$k") #1 remove all spaces
k=$(sed -e 's|###|# ##|g' <<< "$k") #2 put needed space back in
k=$(sed -e 's#-e#|-e #g' <<< "$k") #3 delimit the args with "|"
k=$(sed -e 's#|##' <<< "$k") #4 remove the leading "|"
z=$IFS; IFS="|"; k=($k); IFS=$z #5 convert variable to array
sed "${k[@]}" <<< "aa bb cc" #6 process the string
Output is:
aaBBcc
It works and it is readable for me. But it is really complicated, and took me quite awhile to figure out how to massage k into a form that sed would take.
It fails to work if I quote the expressions, as in -e 's#b#B#g'
Is there a less complicated way, and/or a way to quote the expressions? Must work with k whitespaced as above, sed 4.4, bash 4.4.12(1).
#######################################################
added 2022-09-26 14:58 PST:
Here is a real world script for converting a URL before bookmarking. The caveat is that I wrote for my usage. I don't have to figure out what the code is trying to do because I already know the paradigm, I invented it, or reinvented it.
https://www.ebay.com/sch/i.html?_from=R40&_trksid=123456&_nkw=(vintage,vtg) (polartec,fleece) (full,zip,zips,zipper,zippered,zipping) -(hilfiger,"old navy",hooded,camo,camouflage,vest,small,medium,xl,xxl,half,quarter,"1/4","1/2", lined,winnie,toddler,kids,ladies,womens,women)&_sacat=11450&LH_TitleDesc=0&_odkw=(vintage,vtg) fleece (full,zip,zips,zipper,zippered,zipping) -(hilfiger,"old navy",hooded,camo,camouflage,vest,small,medium,xl,xxl,half,quarter,"1/4","1/2", lined,winnie,toddler,kids,ladies,womens,women)&_osacat=11450&_sop=10&LH_PrefLoc=3&_ipg=240&_udhi=99
into
https://www.ebay.com/sch/i.html?&_nkw=(vintage,vtg) (polartec,fleece) (full,zip,zips,zipper,zippered,zipping) -(hilfiger,old navy,hooded,camo,camouflage,vest,small,medium,xl,xxl,half,quarter,1/4,1/2, lined,winnie,toddler,kids,ladies,womens,women)&_sacat=1145011450&_sop=10&LH_PrefLoc=3&_udhi=99&_ipg=240
#!/bin/bash
echo
k="-e s# [&]*_from=R40 # # \
-e s# [&]*_trk[^&]* # # \
-e s# [&]*_odkw[^&]* # # \
-e s# [&]*_osacat[^&] # # \
-e s# [&]*_sacat=0 # # \
-e s# [&]*LH_TitleDesc=0 # # \
-e s# # #g \
-e s# / # / #g \
-e s# ( # ( #g \
-e s# ) # ) #g \
-e s# , # , #g \
-e s# " # #g \
-e s# &_ipg=[0-9]* # # \
-e s# $ # \&_ipg=240 # \
"
k=$(sed -e 's# ##g' \
-e 's|###|# ##|g' \
-e 's#-e#|-e #g' \
-e 's#|##' \
<<< "$k" \
)
z=$IFS; IFS="|"; k=($k); IFS=$z
sed "${k[@]}" <<< "$1"
CodePudding user response：
See I'm trying to put a command in a variable, but the complex cases always fail!
You want
k=(
-e 's/ //g'
-e 's/b/B/g'
)
sed "${k@]}" ...
Even if this code is only for you, you'll never be able to maintain what you have after ignoring it for a while. Readability and good practice is good for you too.
Given your updated real-life example, here's another thought:
declare -a regex str flag cmds
regex =( '[&]*_from=R40' ); str =( '' ); flag =( '' )
regex =( '[&]*_trk[^&]*' ); str =( '' ); flag =( '' )
regex =( '[&]*_odkw[^&]*' ); str =( '' ); flag =( '' )
regex =( '[&]*_osacat[^&]' ); str =( '' ); flag =( '' )
regex =( '[&]*_sacat=0' ); str =( '' ); flag =( '' )
regex =( '[&]*LH_TntleDesc=0' ); str =( '' ); flag =( '' )
regex =( ' ' ); str =( ' ' ); flag =( 'g' )
regex =( '/' ); str =( '\/' ); flag =( 'g' )
# note the backslash in the "str" value: ^^ -- that's the "s///" delimiter
regex =( '(' ); str =( '(' ); flag =( 'g' )
regex =( ')' ); str =( ')' ); flag =( 'g' )
regex =( ',' ); str =( ',' ); flag =( 'g' )
regex =( '"' ); str =( '' ); flag =( 'g' )
regex =( '&_npg=[0-9]*' ); str =( '' ); flag =( '' )
regex =( '$' ); str =( '\&_npg=240' ); flag =( '' )
n=${#regex[@]}
for ((i=0; i < n; i )); do
cmds =( -e "s/${regex[i]}/${str[i]}/${flag[i]}" )
done
url='https://www.ebay.com/sch/i.html?_from=R40&_trksid=123456&_nkw=(vintage,vtg) (polartec,fleece) (full,zip,zips,zipper,zippered,zipping) -(hilfiger,"old navy",hooded,camo,camouflage,vest,small,medium,xl,xxl,half,quarter,"1/4","1/2", lined,winnie,toddler,kids,ladies,womens,women)&_sacat=11450&LH_TitleDesc=0&_odkw=(vintage,vtg) fleece (full,zip,zips,zipper,zippered,zipping) -(hilfiger,"old navy",hooded,camo,camouflage,vest,small,medium,xl,xxl,half,quarter,"1/4","1/2", lined,winnie,toddler,kids,ladies,womens,women)&_osacat=11450&_sop=10&LH_PrefLoc=3&_ipg=240&_udhi=99'
sed "${cmds[@]}" <<< "$url"
https://www.ebay.com/sch/i.html?&_nkw=(vintage,vtg) (polartec,fleece) (full,zip,zips,zipper,zippered,zipping) -(hilfiger,old navy,hooded,camo,camouflage,vest,small,medium,xl,xxl,half,quarter,1/4,1/2, lined,winnie,toddler,kids,ladies,womens,women)&_sacat=11450&LH_TitleDesc=011450&_sop=10&LH_PrefLoc=3&_ipg=240&_udhi=99&_npg=240
CodePudding user response：
Combining @glenn_jackman answer with my idiosyncrasies:
#!/bin/bash
declare -a k
k=('-e s# [&]*_from=R40 # # '
'-e s# [&]*_trk[^&]* # # '
'-e s# [&]*_odkw[^&]* # # '
'-e s# [&]*_osacat[^&] # # '
'-e s# [&]*_sacat=0 # # '
'-e s# [&]*LH_TitleDesc=0 # # '
'-e s# # #g '
'-e s# / # / #g '
'-e s# ( # ( #g '
'-e s# ) # ) #g '
'-e s# , # , #g '
'-e s# " # #g '
'-e s# &_ipg=[0-9]* # # '
'-e s# $ # \&_ipg=240 # '
)
n=${#k[@]}
for ((i=0; i<n; i )); do
k[i]=$(sed -e 's% *# *%#%g' <<< "${k[i]}")
done
sed "${k[@]}" <<< "$1"
or
#!/bin/bash
declare -a k
k =('-e s# [&]*_from=R40 # # ')
k =('-e s# [&]*_trk[^&]* # # ')
k =('-e s# [&]*_odkw[^&]* # # ')
k =('-e s# [&]*_osacat[^&] # # ')
k =('-e s# [&]*_sacat=0 # # ')
k =('-e s# [&]*LH_TitleDesc=0 # # ')
k =('-e s# # #g ')
k =('-e s# / # / #g ')
k =('-e s# ( # ( #g ')
k =('-e s# ) # ) #g ')
k =('-e s# , # , #g ')
k =('-e s# " # #g ')
k =('-e s# &_ipg=[0-9]* # # ')
k =('-e s# $ # \&_ipg=240 # ')
n=${#k[@]}
for ((i=0; i<n; i )); do
k[i]=$(sed -e 's| *# *|#|g' <<< "${k[i]}")
done
sed "${k[@]}" <<< "$1"
exit
CodePudding user response：
Why not just use ; to join multiple set operations into a single parameter? Something like this:
k="s# # #g;
s# b # B #g;
"
k=$(sed 's# ##g; s|###|# ##|g' <<<"$k") # Clean up spaces in $k
sed "$k" <<< "aa bb cc"
Result: "aaBBcc". Your big pattern would look like this:
k="s# [&]*_from=R40 # # ;
s# [&]*_trk[^&]* # # ;
s# [&]*_odkw[^&]* # # ;
s# [&]*_osacat[^&] # # ;
s# [&]*_sacat=0 # # ;
s# [&]*LH_TitleDesc=0 # # ;
s# # #g ;
s# / # / #g ;
s# ( # ( #g ;
s# ) # ) #g ;
s# , # , #g ;
s# " # #g ;
s# &_ipg=[0-9]* # # ;
s# $ # \&_ipg=240 # ;
"
You could also do the whitespace-mangling as part of the sed command with a command substitution:
sed "$(sed 's# ##g; s|###|# ##|g' <<<"$k")" <<< "$1"