sed with backslash in replacement text.

[Janus Ng]

I have just installed FreeNAS for a week. There were 2 updates after my initial installation. I appreciate it quick development very much. Thanks developers!

As I have many files with names in many languages, I prefer to set the locale in bash to utf-8. This feature is useful to all users; hence, I change the /etc/login.conf, add a line ":lang=en_US.UTF-8:\" to default setting such as:

default:\

:passwd_format=sha512:\
...​

becomes

default:\

:lang=en_US.UTF-8:\
:passwd_format=sha512:\
...​

However, I found that the file is rest every time the OS is updated. That's why I wrote a shell script to accomplish the task. I can run it just after an update next time. The script is listed as following.

#!/bin/bash
d=$(date +%Y-%m-%d)
f=/etc/login.conf
i=0
n=$f.$d.$i
while [ -a "$n" ]; do
i=$(($i + 1))
n=$f.$d.$i
done

sed -i .$d.$i -E 's/^default:\\$/&\'$'\n'$'\t:lang=en_US.UTF-8:\\/g' $f
echo
echo $f backed up to $n
echo $f is updated with en_US.UTF-8 locale.

cap_mkdb /etc/login.conf
echo
echo Please relogin to use new settings.​

However, it reports the following error.

sed: 2: "s/^default:\\$/&\
:lan ...": unterminated substitute in regular expression​

sed complaint about the backslashes in the substitute text.

Please help me out! Thanks a ton!

 

[jgreco]

\\/g strikes me as wrong. echo your expression in the shell to find out what is being passed to sed.

 

[Janus Ng]

Thank you very much, jgreco!
You make my date!

 

[jgreco]

That was it, then, huh?

With all the quoting and stuff like that involved in shell scripting, it sometimes gets hard to see through multiple layers, and I find that seeing what is actually being passed to the underlying utilities (grep, sed, whatever) is sometimes extremely helpful in debugging quoting issues.

 

[Janus Ng]

Yeah, your advice woke me up where the problem is, jgreco.

I have taken the advice on stackoverflow, just write it literally and made it platform independent as

sed -i .$d.$i -E 's/^default:\\$/\\\

:Lang=en_US.UTF-8:\\/g' $f​

Thank you!

 

[anodos]

Yeah, your advice woke me up where the problem is, jgreco.

I have taken the advice on stackoverflow, just write it literally and made it platform independent as

sed -i .$d.$i -E 's/^default:\\$/\\\

:Lang=en_US.UTF-8:\\/g' $f​

Thank you!

You know, you can use other characters as delimiters to avoid 'leaning toothpick syndrome'

 

[jgreco]

You know, you can use other characters as delimiters to avoid 'leaning toothpick syndrome'

@anodos endorses '!' with csh.

 

[Janus Ng]

You know, you can use other characters as delimiters to avoid 'leaning toothpick syndrome'

Can you enlighten me on how to use other characters as delimiters? I didn't know it was possible.
By the way, is it possible to prevent it from adding duplicated lines on consecutive runs? Thanks!

 

[jgreco]

Can you enlighten me on how to use other characters as delimiters? I didn't know it was possible.
By the way, is it possible to prevent it from adding duplicated lines on consecutive runs? Thanks!

The character after the "s" is the delimiter. So "s@one@two@", "s/one/two/", "s,one,two," are all equivalent. Special care need be taken that you pick a delimiter that isn't part of your expression or even more special fun can result.

 

[Janus Ng]

jgreco,

Thank you!