I am trying to pipe output from git to sed, which will inturn replace a value in file.config that represents the last commit on a branch.

I am unsure how to do this using piping in bash. I thought maybe something like this? But it doesnt work (outputs nothing)

git rev-parse --short HEAD | sed -i 's/"commit".*:.*".*"/"commit": "${$_}"/g' file.config

Is this even possible, or is there a better approach I can take?

Contents of file.config

{
    "commit" : "684e8ba31"
}

Contents of file.config (after running command)

{
    "commit" : "${$_}"
}

Contents of file.config (expected output)

{
    "commit" : "441d6fc22"
}

Ouput from git rev-parse --short HEAD

441d6fc22

CodePudding user response：

sed -i '/"commit" *:/s/: *"[^"]*"/: "'$(git rev-parse HEAD)'"/' file.config

will do it. You're confusing command-line arguments with stdin input. They're both good and widely-used ways of feeding the invoked program data but the differences in capacity and timing matter, so spend more time studying how they work.

CodePudding user response：

Simulating output from git call:

$ cat git.out
441d6fc22

One idea where we capture the git output to a variable and then feed the variable to sed:

$ read -r sha < <(cat git.out)
$ sed -E "s/^(.*\"commit\"[^\"]*).*/\1\"${sha}\"/" file.config
{
    "commit" : "441d6fc22"
}

Where:

• -E - enable extended regex support
• (.*\"commit\"[^\"]*) - (1st capture group) everything from start of line up to and including "commit" and then everything that follows up to the first occurrence of a literal "
• .* rest of line
• \1\"${sha}"\" - replace with contents of 1st capture group followed by literal ", contents of sha variable and another literal "

Once satisified with the result the -i option can be added:

$ sed -Ei "s/^(.*\"commit\"[^\"]*).*/\1\"${sha}\"/" file.config
$ cat file.config
{
    "commit" : "441d6fc22"
}

Pulling original git call into the mix:

read -r sha < <(git rev-parse --short HEAD)
sed -Ei "s/^(.*\"commit\"[^\"]*).*/\1\"${sha}\"/" file.config