Reputation: 31726
I have a common use case that I'd like to write a function for: I often want to cd to some directory relative to some file.
$ gem which rspec/core | xargs echo -n | pbcopy
$ cd *paste and delete end until direcory looks right*
note: gem which rspec/core
prints something like "/Users/joshcheek/.rvm/gems/ruby-1.9.3-p125/gems/rspec-core-2.10.0/lib/rspec/core.rb"
$ gem which rspec/core | 2dir 3
Which will cd me into "/Users/joshcheek/.rvm/gems/ruby-1.9.3-p125/gems/rspec-core-2.10.0" (passing the argument "3" tells it to remove "lib/rspec/core.rb" from the end)
2dir() {
read dir
for i in $(seq 1 $1)
do
dir="${dir%/*}"
done
cd "$dir"
}
But the cd changes the function's directory, not mine. I've tried swapping it with an alias, but can't figure out how to make anonymous functions or pass the argument.
Upvotes: 4
Views: 9943
Reputation: 360035
Based on Jonathan Leffler's answer, but without a loop:
2dir () {
local levels name=${2:?"Usage: $FUNCNAME count path"};
printf -v levels '%*s' "$1" '';
cd "/${name%${levels// //*}}"
}
One annoyance is that it produces perfectly valid directories with leading double slashes (e.g. echo "$PWD"
outputs "//foo/bar/baz" after using the function).
Another is that it's "too clever by half."
Edit:
Fixed the double slash issue:
2dir () {
local levels name=${2:?"Usage: $FUNCNAME count path"};
printf -v levels '%*s' $1 '';
name=/${name%${levels// //*}};
cd "${name/\/\///}"
}
Upvotes: 0
Reputation: 753675
I'd use:
2dir()
{
name=${2:?'Usage: 2dir count path'}
count=$1
while [[ $count -gt 0 ]]; do name=$(dirname "$name"); ((count--)); done
cd "$name"
}
and use it as:
2dir 3 $(gem which rspec/core)
This works where your pipeline can't. The cd
in the pipe process affects that (sub-)shell, but cannot affect the current directory of the parent process. This function can be made to work.
And you can use your dir="${dir%/*}"
in place of my dirname
if you prefer, except that you'll end up in your home directory instead of the current directory (or root directory, depending on whether you gave a relative or absolute path name) if you specify 10 when there are only 5 components.
Upvotes: 9
Reputation: 125788
Here's a variant of @Jonathan Leffler's suggestion to streamline usage a little -- it makes the count argument optional, and avoids the need for $( )
around the command:
2dir() {
# If first arg is a number, use it as a trim count; otherwise assume 2
if [[ "$1" =~ ^[0-9]+$ ]]; then
count="$1"
shift
else
count=2
fi
if [[ $# -lt 1 ]]; then # Make sure a command was specified
echo "Usage: 2dir [count] command [commandargs ...]" >&2
return 1
fi
name="$("$@")" # Execute the remaining args as a command to get the target directory
while [[ $count -gt 0 ]]; do name=$(dirname "$name"); ((count--)); done
cd "$name"
}
Example uses:
2dir 3 gem which rspec/core
2dir gem which rspec/core
Upvotes: 2
Reputation: 3733
The command
gem which rspec/core | 2dir 3
is known as a "pipeline" in shell parlance. Each command in the pipeline is executed as a separate process. If one of the commands in the pipeline is a shell function, it may be executed by the current (interactive) shell process. But it is not guaranteed, and in your case this is not happening.
To fix your problem you just need to make sure that the function is evaluated in the interactive shell. You just need to fix the function and then use it differently. Here is the updated function:
2dir() {
declare -ir snip="$1"
declare dir="$2"
for i in $(seq 1 "$snip"); do
dir="${dir%/*}"
done
cd "$dir"
}
You use it like this:
$ 2dir 3 "$(gem which rspec/core)"
Upvotes: 1
Reputation: 100023
A shell script can't change the working directory of the interactive shell. Only an alias can do that, since it runs in the shell whose directory you are trying to change.
In other words:
There is a Linux process running the shell and accept commands from you. It has a working directory. When you tell it to execute a shell script, it creates a brand new process with an independent working directory, disconnected from the first.
Upvotes: 0