sed replacement variable has a double backslash, but will not write properly-CodePudding

In a bash script I am working on to manipulate some configuration files, I have several fields working properly but one field that can take a double backslash config_item = "configfoo\\configbar" as part of the configuration string.

I am reading in from the user with read -rp 'Configuration Item: ' config_item

Where it echos properly configfoo\\configbar

I've been unable to successfully land that into the sed command that updates the file.

sed -i -e '/config_item=/ s/=.*/='\"$config_item\"'/' ${dir}/.env

I get an error

sed: -e expression #1, char 19: unknown command: `\'

Which I assume is related to the 1st slash escaping. I've tried four slashes also trying to figure this out.

How do I tell sed to replace with the variable exactly as the user types it, or is there an alternative string preperation step I need to do so that my configuration file line reads

config_item = "configfoo\\configbar"

There are hundreds of posts on the topic, but I am not trying to substitute the slashes. I just need the string to flow into the substitution.

CodePudding user response：

If you absolutely don't want to escape your input string then you cannot use sed; awk is able to do it though:

awk '
    BEGIN {
        FS = OFS = "="
        value = ARGV[2]
        delete ARGV[2]
    }
    $1 == "config_item" { $0 = $1 OFS "\"" value "\"" }
    1
' "$dir"/.env "$config_item" > "$dir"/.env.tmp &&
mv "$dir"/.env.tmp "$dir"/.env

CodePudding user response：

This might work for you (GNU sed and bash):

read -rp 'Configuration Item: ' config_item
export config_item
sed -Ei '/config_item\s*=/s/(.*=\s*).*/printf '\''\1"%s"'\'' "$config_item"/e' file

Read input into config_item variable.

Export config_item_variable.

Replace the entire line containing config_item = and replace by config_item = and the bash variable config_item.