I was trying to install docker in Ubuntu, and following the instructions I have come across a very strange sed command which I do not understand (this seems to be used to set-up bash autocompletion for docker):
sudo sed -i '$acomplete -F _docker docker' /etc/bash_completion.d/docker.io
What is that command doing? The -i command means in-place editing, but what does the $acomplete -F _docker docker mean? The $ is matching the last line, but what is it doing? I do not even recognize any sed command there! For example, a substitution command would look like:
$s/in/out/
Could somebody explain that expression for me?
It appends complete -F _docker docker to the file.
From sed's manual:
a \
text Append text, which has each embedded newline preceded by a backslash.
Related
#!/bin/bash
jobname="job_201312161447_0003"
jobname_pre=${jobname:0:16}
jobname_post=${jobname:17}
This bash script gives me Bad substitution error on ubuntu. Any help will be highly appreciated.
The default shell (/bin/sh) under Ubuntu points to dash, not bash.
me#pc:~$ readlink -f $(which sh)
/bin/dash
So if you chmod +x your_script_file.sh and then run it with ./your_script_file.sh, or if you run it with bash your_script_file.sh, it should work fine.
Running it with sh your_script_file.sh will not work because the hashbang line will be ignored and the script will be interpreted by dash, which does not support that string substitution syntax.
I had the same problem. Make sure your script didnt have
#!/bin/sh
at the top of your script. Instead, you should add
#!/bin/bash
For others that arrive here, this exact message will also appear when using the env variable syntax for commands, for example ${which sh} instead of the correct $(which sh)
Your script syntax is valid bash and good.
Possible causes for the failure:
Your bash is not really bash but ksh or some other shell which doesn't understand bash's parameter substitution. Because your script looks fine and works with bash.
Do ls -l /bin/bash and check it's really bash and not sym-linked to some other shell.
If you do have bash on your system, then you may be executing your script the wrong way like: ksh script.sh or sh script.sh (and your default shell is not bash). Since you have proper shebang, if you have bash ./script.sh or bash ./script.sh should be fine.
Try running the script explicitly using bash command rather than just executing it as executable.
Also, make sure you don't have an empty string for the first line of your script.
i.e. make sure #!/bin/bash is the very first line of your script.
Not relevant to your example, but you can also get the Bad substitution error in Bash for any substitution syntax that Bash does not recognize. This could be:
Stray whitespace. E.g. bash -c '${x }'
A typo. E.g. bash -c '${x;-}'
A feature that was added in a later Bash version. E.g. bash -c '${x#Q}' before Bash 4.4.
If you have multiple substitutions in the same expression, Bash may not be very helpful in pinpointing the problematic expression. E.g.:
$ bash -c '"${x } multiline string
$y"'
bash: line 1: ${x } multiline string
$y: bad substitution
Both - bash or dash - work, but the syntax needs to be:
FILENAME=/my/complex/path/name.ext
NEWNAME=${FILENAME%ext}new
I was adding a dollar sign twice in an expression with curly braces in bash:
cp -r $PROJECT_NAME ${$PROJECT_NAME}2
instead of
cp -r $PROJECT_NAME ${PROJECT_NAME}2
I have found that this issue is either caused by the marked answer or you have a line or space before the bash declaration
Looks like "+x" causes problems:
root#raspi1:~# cat > /tmp/btest
#!/bin/bash
jobname="job_201312161447_0003"
jobname_pre=${jobname:0:16}
jobname_post=${jobname:17}
root#raspi1:~# chmod +x /tmp/btest
root#raspi1:~# /tmp/btest
root#raspi1:~# sh -x /tmp/btest
+ jobname=job_201312161447_0003
/tmp/btest: 4: /tmp/btest: Bad substitution
in my case (under ubuntu 18.04), I have mixed $( ${} ) that works fine:
BACKUPED_NB=$(ls ${HOST_BACKUP_DIR}*${CONTAINER_NAME}.backup.sql.gz | wc --lines)
full example here.
I used #!bin/bash as well tried all approaches like no line before or after #!bin/bash.
Then also tried using +x but still didn't work.
Finally i tried running the script ./script.sh it worked fine.
#!/bin/bash
jobname="job_201312161447_0003"
jobname_post=${jobname:17}
root#ip-10-2-250-36:/home/bitnami/python-module/workflow_scripts# sh jaru.sh
jaru.sh: 3: jaru.sh: Bad substitution
root#ip-10-2-250-36:/home/bitnami/python-module/workflow_scripts# ./jaru.sh
root#ip-10-2-250-36:/home/bitnami/python-module/workflow_scripts#
I have a textfile and would like to save all lines that are ending with .m2 or .M2. I tried several ways including this here
D:\filetype\core\sed.exe -n -e "s/^\(.*\)\.M2//" D:\filetype\listfile\test.txt > D:\filetype\listfile\test2.txt
But as result i only get a emtpy textfile, so i guess something is wrong with my code.
The other way was
D:\filetype\core\sed.exe -n -e "^/\(.*)\/.M2\/" D:\filetype\listfile\test.txt > D:\filetype\listfile\test2.txt
But in this case i wasn't able to locate the source of the error
unknown command: `^'
Thanks if someone can see my fault.
You can use below sed command:
sed -n -e '/\.[mM]2$/p' <file_name>
This will print all the lines which have .m2 or .M2 at the end
Now comming to issues with your commands. Your first command does:
sed -n -e "s/^\(.*\)\.M2//"
which is a search and replace command indiacated by s in the command. Syntax for this command is s/search_text/replace_text/. Now if you are look at your command carefully, you are searching for something and replacing it with nothing - as indicated by last // in your command - where replace_text is missing.
Your second command does
sed -n -e "^/\(.*)\/.M2\/"
which is incorrect syntax. General syntax of a sed command is :
sed -e 'line_range command'
where line range - which is optional - can be line numbers like1, 5 , 2-5, or a regular expression like /[gG]/, /[gG][iIuU]rl/.
If line_range is missing, the first letter in sed command should be a command. In your case: line_range is missing - which is fine syntax wise -, however the first letter is ^ - which is not a sed command - because of which you are getting syntax error.
The command that I suggested is
sed -n -e '/\.[mM]2$/p'
Here, line_range is the regular expression /\.[mM]2$/ - which says "any line which has .m2 or .M2 at the end", and command is p, which is the letter for print command.
Sed is mostly used to transform text. Why not use grep instead?
grep -i "\.m2$"
This will match case insensitively (-i) any line ending with .m2.
I am trying to add a line in a file using sed after first match. I looked my threads online, addressing the same issue and tried below versions of the command:
sudo sed -e '/cfg_file/cfg_file='$NEW_FILE -e '/cfg_file/cfg_file=/q' MAIN_CONF_FILE
sudo sed -i '0,/cfg_file/ a\cfg_file='$NEW_FILE $MAIN_CONF_FILE
sudo sed -i '1s/cfg_file/cfg_file='$NEW_FILE $MAIN_CONF_FILE
sudo sed -i '1/cfg_file/s/cfg_file='$NEW_FILE $MAIN_CONF_FILE
Unfortunately, nothing worked for me. Either they show error, in case of point 3, or show similar behavior of adding lines after each match.
SAMPLE FILE
cfg_file=some_line1
cfg_file=some_line2
Now I want to add a line after first match of cg_file, like below.
EXPECTED RESULT
cfg_file=some_line1
cfg_file=my_added_line_after_first_match_only.
cfg_file=some_line2
Help me in adding line after first match and correcting my command.
Since you're on Ubuntu, you are using GNU sed. GNU sed has some weird features and some useful ones; you should ignore the weird ones and use the useful ones.
In context, the useful one is ranges starting at line 0. Weird ones are the way it messes with a, i and c commands.
MAIN_CONF_FILE=/tmp/copy.of.main.config.file
NEWFILE="my_added_line_after_first_match_only."
sed -e '0,/^cfg_file=/ { /^cfg_file/ a\' \
-e "cfg_file=$NEWFILE" \
-e '}' \
"$MAIN_CONF_FILE"
In classic sed, the a command is followed by backslash-newline, and each subsequent line of the script is appended up to and including the first line without a backslash at the end (and the backslash is removed). Each -e argument functions as a line in the script. Distinguish between the shell lines escaped with backslash at the end and the sed script lines with backslash at the end.
Example at work
$ cat /tmp/copy.of.main.config.file | so
cfg_file=some_line1
cfg_file=some_line2
$ cat script.sh
MAIN_CONF_FILE=/tmp/copy.of.main.config.file
NEWFILE="my_added_line_after_first_match_only."
SED=/opt/gnu/bin/sed
${SED} \
-e '0,/^cfg_file=/ { /^cfg_file/ a\' \
-e "cfg_file=$NEWFILE" \
-e '}' \
"$MAIN_CONF_FILE"
$ bash script.sh
cfg_file=some_line1
cfg_file=my_added_line_after_first_match_only.
cfg_file=some_line2
$
This is based on your attempt 2, but avoids some of the weird stuff.
Basic sanity
As I noted, it is not sensible to experiment with sudo and the -i option to sed. You don't use those until you know that the script will do the job correctly. It is dangerous to do anything as root via sudo. It is doubly dangerous when you don't know whether what you're trying to use will work. Don't risk wrecking your system.
I'm having an issue trying to capture the output of a sed command in a makefile variable.
JS_SRC:=$(shell sed -n 's#.*src="\([^"]*\.js\).*#\1#p' index.html)
Which gives me
sed: -e expression #1, char 34: unknown option tos'
`
I've been trying to escape things and the like, but am always given that error.
All variations of escaping I have run, run fine from the terminal.
How does a makefile call the shell command?. /usr/bin/sh -c "cmd?" or something different?.
Somethings being interpolated but I have no idea what.
JS_SRC:=$(shell sed -n "s/.*src=\"\\([^\"]*\\.js\\).*/\\1/p" index.html)
Appears to work. I figured this out via running make -d and seeing the process it was creating.
What was baffling is that it did different things with ' vs " in the sed argument. " is run with /bin/sh -c "args" so I was able to tweak the escaping to get what I needed to appear there. Using ' seems to invoke sed directly.
There is a whole heap of escaping, that i imagine is unnecessary (I don't need to interpolate variables in the sed expression, but it sends it to a shell I understand. So it will have to do ! :)
I have to convert filenames having a forward slash to filenames with a back slash in a makefile using mingw32. I used the following sed command to store the result in a variable:
ORIGINAL=./a/b/main1.c ./a/b/main2.c ./a/b/main3.c
sed command:
RESULT=$(shell echo $(ORIGINAL) | sed 's/\//\\/g')
And the resulting output is:
.\a\b\main1.c .abmain2.c .abmain3.c
It works fine if I run it directly on bash. Can anyone tell me whats wrong?
Thanks!
I'm trying to duplicate your test in a makefile, but I don't have your environment.
But I would not use '/' as a command separator if I am searching for the same character. Use another character, like ':'
sed 's:/:\\:g'
RESULT=$(sed 's/\//\\/g' <<< "$ORIGINAL")
$ ORIGINAL='./a/b/main1.c ./a/b/main2.c ./a/b/main3.c'
$ echo "$ORIGINAL"
./a/b/main1.c ./a/b/main2.c ./a/b/main3.c
$ RESULT=$(sed 's/\//\\/g' <<< "$ORIGINAL")
$ echo $RESULT
.\a\b\main1.c .\a\b\main2.c .\a\b\main3.c
Most likely the shell that make is invoking is not bash. However, I find the behavior you're seeing very strange for any shell (it replaces in the first word but not the rest?!?!)
Have you considered using GNU make's $(subst ...) function, instead of the shell?