I need some help with a shell code. Now I have this code:
find $dirname -type f -exec md5sum '{}' ';' | sort | uniq --all-repeated=separate -w 33 | cut -c 35-
This code finds duplicated files (with same content) in a given directory. What I need to do is to update it - find out latest (by date) modified file (from duplicated files list), print that file name and also give opportunity to delete that file in terminal.
Doing this in pure bash is a tad awkward, it would be a lot easier to write
this in perl or python.
Also, if you were looking to do this with a bash one-liner, it might be feasible,
but I really don't know how.
Anyhoo, if you really want a pure bash solution below is an attempt at doing
what you describe.
Please note that:
I am not actually calling rm, just echoing it - don't want to destroy your files
There's a "read -u 1" in there that I'm not entirely happy with.
Here's the code:
#!/bin/bash
buffer=''
function process {
if test -n "$buffer"
then
nbFiles=$(printf "%s" "$buffer" | wc -l)
echo "================================================================================="
echo "The following $nbFiles files are byte identical and sorted from oldest to newest:"
ls -lt -c -r $buffer
lastFile=$(ls -lt -c -r $buffer | tail -1)
echo
while true
do
read -u 1 -p "Do you wish to delete the last file $lastFile (y/n/q)? " answer
case $answer in
[Yy]* ) echo rm $lastFile; break;;
[Nn]* ) echo skipping; break;;
[Qq]* ) exit;;
* ) echo "please answer yes, no or quit";;
esac
done
echo
fi
}
find . -type f -exec md5sum '{}' ';' |
sort |
uniq --all-repeated=separate -w 33 |
cut -c 35- |
while read -r line
do
if test -z "$line"
then
process
buffer=''
else
buffer=$(printf "%s\n%s" "$buffer" "$line")
fi
done
process
echo "done"
Here's a "naive" solution implemented in bash (except for two external commands: md5sum, of course, and stat used only for user's comfort, it's not part of the algorithm). The thing implements a 100% Bash quicksort (that I'm kind of proud of):
#!/bin/bash
# Finds similar (based on md5sum) files (recursively) in given
# directory. If several files with same md5sum are found, sort
# them by modified (most recent first) and prompt user for deletion
# of the oldest
die() {
printf >&2 '%s\n' "$#"
exit 1
}
quicksort_files_by_mod_date() {
if ((!$#)); then
qs_ret=()
return
fi
# the return array is qs_ret
local first=$1
shift
local newers=()
local olders=()
qs_ret=()
for i in "$#"; do
if [[ $i -nt $first ]]; then
newers+=( "$i" )
else
olders+=( "$i" )
fi
done
quicksort_files_by_mod_date "${newers[#]}"
newers=( "${qs_ret[#]}" )
quicksort_files_by_mod_date "${olders[#]}"
olders=( "${qs_ret[#]}" )
qs_ret=( "${newers[#]}" "$first" "${olders[#]}" )
}
[[ -n $1 ]] || die "Must give an argument"
[[ -d $1 ]] || die "Argument must be a directory"
dirname=$1
shopt -s nullglob
shopt -s globstar
declare -A files
declare -A hashes
for file in "$dirname"/**; do
[[ -f $file ]] || continue
read md5sum _ < <(md5sum -- "$file")
files[$file]=$md5sum
((hashes[$md5sum]+=1))
done
has_found=0
for hash in "${!hashes[#]}"; do
((hashes[$hash]>1)) || continue
files_with_same_md5sum=()
for file in "${!files[#]}"; do
[[ ${files[$file]} = $hash ]] || continue
files_with_same_md5sum+=( "$file" )
done
has_found=1
echo "Found ${hashes[$hash]} files with md5sum=$hash, sorted by modified (most recent first):"
# sort them by modified date (using quicksort :p)
quicksort_files_by_mod_date "${files_with_same_md5sum[#]}"
for file in "${qs_ret[#]}"; do
printf " %s %s\n" "$(stat --printf '%y' -- "$file")" "$file"
done
read -p "Do you want to remove the oldest? [yn] " answer
if [[ ${answer,,} = y ]]; then
echo rm -fv -- "${qs_ret[#]:1}"
fi
done
if((!has_found)); then
echo "Didn't find any similar files in directory \`$dirname'. Yay."
fi
I guess the script is self-explanatory (you can read it like a story). It uses the best practices I know of, and is 100% safe regarding any silly characters in file names (e.g., spaces, newlines, file names starting with hyphens, file names ending with a newline, etc.).
It uses bash's globs, so it might be a bit slow if you have a bloated directory tree.
There are a few error checkings, but many are missing, so don't use as-is in production! (it's a trivial but rather tedious taks to add these).
The algorithm is as follows: scan each file in the given directory tree; for each file, will compute its md5sum and store in associative arrays:
files with keys the file names and values the md5sums.
hashes with keys the hashes and values the number of files the md5sum of which is the key.
After this is done, we'll scan through all the found md5sum, select only the ones that correspond to more than one file, then select all files with this md5sum, then quicksort them by modified date, and prompt the user.
A sweet effect when no dups are found: the script nicely informs the user about it.
I would not say it's the most efficient way of doing things (might be better in, e.g., Perl), but it's really a lot of fun, surprisingly easy to read and follow, and you can potentially learn a lot by studying it!
It uses a few bashisms and features that only are in bash version ≥ 4
Hope this helps!
Remark. If on your system date has the -r switch, you can replace the stat command by:
date -r "$file"
Remark. I left the echo in front of rm. Remove it if you're happy with how the script behaves. Then you'll have a script that uses 3 external commands :).
I was using a code to rename files and a very interesting problem emerged: It worked in a computer running Mac OSX Lion 10.7.5, but it failed on Mac OSX 10.6.8 Snow Leopard.
The code is:
for i in *; do mv $i `echo $i | sed 's/..//'`; done
and the error I got is:
usage: mv [-f | -i | -n] [-v] source target
mv [-f | -i | -n] [-v] source ... directory
It's very weird because mv is working normally if not in a for loop...
Anyone know what should I do to get it working?
You should use bash string substitution:
for i in *; do mv $i ${i/??/}; done
or
for i in *; do mv $i ${i#??}; done
If you have files with spaces (or certain other shell metachaarcters) it's critical that you put all variable references in double-quotes. Otherwise, if you have e.g. i="File Name.txt", you'll be running a command like mv File Name.txt le Name.txt and it'll think you're specifying 4 filenames not just 2. The other standard mistake (which you're not making) is using for file in $(ls), which gets confused by spaces before the names even make it into the variable.
Also, I'm a bit concerned about short filenames and name conflicts. Even if you think you don't have any short filenames, if you have the dotglob shell option set you'll wind up trying to rename the pseudo-files "." and "..", which won't go well at all. Also, if you have e.g. files named "abcdefg" and "cdefg", the script will rename the first over the second (silently erasing the second), then rename that to "efg".
So, here's my proposed rewrite (also using #TrueY's suggestion for shortening the filename):
for i in *; do
if [ ${#i} -le 2 ]; then
echo "$i: not renamed (too short)" >&2
elif [ -e "${i:2}" ]; then
echo "$i: not renamed (${i:2} already exists)" >&2
else
mv "$i" "${i:2}"
fi
done
I have a directory with lots of files. I want to keep only the 6 newest. I guess I can look at their creation date and run rm on all those that are too old, but is the a better way for doing this? Maybe some linux command I could use?
Thanks!
:)
rm -v $(ls -t mysvc-*.log | tail -n +7)
ls -t, list sorted by time
tail -n +7, +7 here means length-7, so all but first 7 lines
$() makes a list of strings from the enclosed command output
rm to remove the files, of course
Beware files with space in their names, $() splits on any white-space!
Here's my take on it, as a script. It does handle spaces in file names even if it is a bit of a hack.
#!/bin/bash
eval set -- $(ls -t1 | sed -e 's/.*/"&"/')
if [[ $# -gt 6 ]] ; then
shift 6
while [[ $# -gt 0 ]] ; do
echo "remove this file: $1" # rm "$1"
shift
done
fi
The second option to ls up there is a "one" for one file name per line. Doesn't actually seem to matter, though, since that appears to be the default when ls isn't feeding a tty.
Hi guys i've a problem with grep . I don't know if there is another search code in shell script.
I'm trying to backup a folder AhmetsFiles which is stored in my Flash Disk , but at the same time I've to group them by their extensions and save them into [extensionName] Folder.
AhmetsFiles
An example : /media/FlashDisk/AhmetsFiles/lecture.pdf must be stored in /home/$(whoami)/Desktop/backups/pdf
Problem is i cant copy a file which name contains spaces.(lecture 2.pptx)
After this introduction here my code.
filename="/media/FlashDisk/extensions"
count=0
exec 3<&0
exec 0< $filename
mkdir "/home/$(whoami)/Desktop/backups"
while read extension
do
cd "/home/$(whoami)/Desktop/backups"
rm -rf "$extension"
mkdir "$extension"
cd "/media/FlashDisk/AhmetsFiles"
files=( `ls | grep -i "$extension"` )
fCount=( `ls | grep -c -i "$extension"` )
for (( i=0 ; $i<$fCount ; i++ ))
do
cp -f "/media/FlashDisk/AhmetsFiles/${files[$i]}" "/home/$(whoami)/Desktop/backups/$extension"
done
let count++
done
exec 0<&3
exit 0
Your looping is way more complicated than it needs to be, no need for either ls or grep or the files and fCount variables:
for file in *.$extension
do
cp -f "/media/FlashDisk/AhmetsFiles/$file" "$HOME/Desktop/backups/$extension"
done
This works correctly with spaces.
I'm assuming that you actually wanted to interpret $extension as a file extension, not some random string in the middle of the filename like your original code does.
Why don't you
grep -i "$extension" | while IFS=: read x ; do
cp ..
done
instead?
Also, I believe you may prefer something like grep -i ".$extension$" instead (anchor it to the end of line).
On the other hand, the most optimal way is probably
cp -f /media/FlashDisk/AhmetsFiles/*.$extension "$HOME/Desktop/backups/$extension/"
I have a problem to manage long paths. How can I get quickly to paths like
/Users/User/.../.../.../.../.../Dev/C/card.c
I tried an alias
alias cd C='cd /Users/User/.../.../.../.../.../Dev/C'
but I am unable to do aliases for two separate words. I have long lists of Bash aliases and paths in CDPATH, so I am hesitating to make them more. How can manage long paths?
[Ideas for Replies]
The user litb's reply revealed some of my problems in the management. Things, such as "CTRL+R", "!-3:1:2:4:x" and "incremental search", are hard for me. They probably help in navigating long directories and, in the sense, management.
Using symlinks is probably the best idea; but you can do it even easier than dumping them all into your home directory.
As you mentioned, BASH has a feature called CDPATH which comes in really handy here.
Just make a hidden folder in your homedir (so it doesn't clutter your homedir too much):
$ mkdir ~/.paths
$ cd ~/.paths
$ ln -s /my/very/long/path/name/to/my/project project
$ ln -s /some/other/very/long/path/to/my/backups backups
$ echo 'CDPATH=~/.paths' >> ~/.bashrc
$ source ~/.bashrc
This creates a directory in your homedir called ".paths" which contains symlinks to all your long directory locations which you regularly use, then sets the CDPATH bash variable to that directory (in your .bashrc) and re-reads the .bashrc file.
Now, you can go to any of those paths from anywhere:
$ cd project
$ cd backups
Leaving you with a short CDPATH, no cluttering aliasses, and more importantly: A really easy way to navigate to those long paths from other applications, such as UI applications, by just going into ~/.paths or adding that directory into your UI application's sidebar or so.
Probably the easiest all-round solution you can have.
Consider using symbolic links. I have a ~/work/ directory where I place symlinks to all my current projects.
You may also use shell variables:
c='/Users/User/.../.../.../.../.../Dev/C'
Then:
cd "$c"
Create symlinks in your home directory (or somewhere else of your choosing)
ln -s longDirectoryPath ~/MySymLinkName
See man ln for more details.
Probably the easiest solution is to use:
alias cdc='cd /Users/User/.../.../.../.../.../Dev/C'
alias cdbin='cd /Users/User/.../.../.../.../.../Dev/bin'
alias cdtst='cd /Users/User/.../.../.../.../.../Dev/tst'
if you're only really working on one project at a time. If you work on multiple projects, you could have another alias which changed the directories within those aliases above.
So, you'd use something like:
proj game17
cdc
make
proj roman_numerals
cdbin
rm -f *
proj game17 ; cdc
Since this is a useful thing to have, I decided to put together a series of scripts that can be used. They're all based aroung a configuration file that you place in your home directory, along with aliases to source scripts. The file "~/.cdx_data" is of the form:
scrabble:top=~/dev/scrabble
scrabble:src=~/dev/scrabble/src
scrabble:bin=~/dev/scrabble/bin
sudoku:top=~/dev/scrabble
sudoku:src=~/dev/scrabble/src
sudoku:bin=~/dev/scrabble/bin
sudoku:data=~/dev/scrabble/data
and lists all the relevant projects (scrabble and sodoku in this case) and their directories (which may be different for each project, but have top, bin, src and data in this example).
The first action is to initialize stuff, so put:
. ~/.cdx_init
at the end of your .bash_profile and create the "~/.cdx_init" file as:
alias cdxl='. ~/.cdx_list'
alias projl='. ~/.cdx_projlist'
alias cdx='. ~/.cdx_goto'
alias proj='. ~/.cdx_proj'
This sets up the four aliases to source the files which I'll include below. Usage is:
cdxl - List all directories in current project.
projl - List all projects.
proj - Show current project.
proj <p> - Set current project to <p> (if allowed).
cdx - Show current project/directory and expected/actual real
directory, since they can get out of sync if you mix cd and cdx.
cdx . - Set actual real directory to expected directory (in other words,
get them back into sync).
cdx <d> - Set directory to <d> (if allowed).
The actual script follow. First, ".cdx_list" which just lists the allowed directories in the current project (pipelines are broken into multiple lines for readability but they should all be on one line).
echo "Possible directories are:"
cat ~/.cdx_data
| grep "^${CDX_PROJ}:"
| sed -e 's/^.*://' -e 's/=.*$//'
| sort -u
| sed 's/^/ /'
Similarly, ".cdx_projlist" shows all the possible projects:
echo "Possible projects are:"
cat ~/.cdx_data
| grep ':'
| sed 's/:.*$//'
| sort -u
| sed 's/^/ /'
In the meaty scripts, ".cdx_proj" sets and/or shows the current project:
if [[ "$1" != "" ]] ; then
grep "^$1:" ~/.cdx_data >/dev/null 2>&1
if [[ $? != 0 ]] ; then
echo "No project name '$1'."
projl
else
export CDX_PROJ="$1"
fi
fi
echo "Current project is: [${CDX_PROJ}]"
and ".cdx_goto" is the same for directories within the project:
if [[ "$1" == "." ]] ; then
CDX_TMP="${CDX_DIR}"
else
CDX_TMP="$1"
fi
if [[ "${CDX_TMP}" != "" ]] ; then
grep "^${CDX_PROJ}:${CDX_TMP}=" ~/.cdx_data >/dev/null 2>&1
if [[ $? != 0 ]] ; then
echo "No directory name '${CDX_TMP}' for project '${CDX_PROJ}'."
cdxl
else
export CDX_DIR="${CDX_TMP}"
cd $(grep "^${CDX_PROJ}:${CDX_DIR}=" ~/.cdx_data
| sed 's/^.*=//'
| head -1
| sed "s:^~:$HOME:")
fi
fi
CDX_TMP=$(grep "^${CDX_PROJ}:${CDX_DIR}=" ~/.cdx_data
| sed 's/^.*=//'
| head -1
| sed "s:^~:$HOME:")
echo "Current project is: [${CDX_PROJ}]"
echo "Current directory is: [${CDX_DIR}]"
echo " [${CDX_TMP}]"
echo "Actual directory is: [${PWD}]"
unset CDX_TMP
It uses three environment variables which are reserved for its own use: "CDX_PROJ", "CDX_DIR" and "CDX_TMP". Other than those and the afore-mentioned files and aliases, there are no other resources used. It's the simplest, yet most adaptable solution I could come up with. Best of luck.
Revisiting. Today I received this link from a social bookmarking site, then I immediately remembered this question:
Navigation with bm
We keep a simple, plain text bookmarks
file and use a tool called bm to do
the look-ups. The tool can also be
used to edit the bookmark index
dynamically as shown below where we
add the directories from the previous
example to the index.
Once i cd'ed into such a long directory, i have that in the history. Then i just type Ctrl-R for the "(reverse-i-search)" prompt and type in a few characters, like Dev/C that appear somewhere in the path, and it shows me the command what i issued back then and i can easily jump to it again.
That works pretty well in practice. Because it won't find an entry if you haven't typed that path for quite some time, which would mean doing work to make things easier probably wouldn't be worth the time. But it definitely will find it if you used it recently. Which is exactly what i need.
In some way, it's a self-organizing cache for long commands & path-names :)
You might want to consider using a script like this in your .bashrc. I've used it on a daily basis ever since I read that post. Pretty bloody useful.
The user jhs suggested Pushd and Popd-commands. I share here some of my Bash-scripts that I found in Unix Power Tools -book. They are very cool when your directories get a way too long :)
#Moving fast between directories
alias pd=pushd
alias pd2='pushd +2'
alias pd3='pushd +3'
alias pd4='pushd +4'
The command 'pushd +n' "rotates" the stack. The reverse command 'popd +n' deletes the n entry of the stack. If your stack gets too long, use 'repeat n popd'. For examle, your stack is 12 directories long:
repeat 11 popd
When you want to see your stack, write 'pushd'. For further reading, I recommend the book on pages 625-626.
In your .bashrc find PS1='${debian_chroot:+($debian_chroot)}[\033[01;32m]\u#\h[\033[00m]:[\033[01;34m]
\W[\033[00m]\$ '
and replace the \w with \W.I already have it changed here. This will only give you the main directory where you are working. You can get the full directory by typing pwd
There are fundamental well-known ideas, like creating aliases:
alias cdfoo="cd /long/path/to/foo"
and also "dropping pebbles"
export foo=/long/path/to/foo
and also making the above "project-based". I use 'ticket based' directories.
topdir=ticket_12345
alias cdfoo="cd home/me/sandbox/$topdir/long/path/to/foo"
export foo="/home/me/sandbox/$topdir/long/path/to/foo"
but beyond all this, sometimes it's just handy to jump back and forth to where you've been recently, using command-line menus. (pushd and popd are cumbersome, IMHO).
I use acd_func.sh (listed below). Once defined, you can do
cd --
to see a list of recent directories, with a numerical menu
cd -2
to go to the second-most recent directory.
Very easy to use, very handy.
Here's the code:
# Insert into .profile, .bash_profile or wherever
# acd_func 1.0.5, 10-nov-2004
# petar marinov, http:/geocities.com/h2428, this is public domain
cd_func ()
{
local x2 the_new_dir adir index
local -i cnt
if [[ $1 == "--" ]]; then
dirs -v
return 0
fi
the_new_dir=$1
[[ -z $1 ]] && the_new_dir=$HOME
if [[ ${the_new_dir:0:1} == '-' ]]; then
#
# Extract dir N from dirs
index=${the_new_dir:1}
[[ -z $index ]] && index=1
adir=$(dirs +$index)
[[ -z $adir ]] && return 1
the_new_dir=$adir
fi
#
# '~' has to be substituted by ${HOME}
[[ ${the_new_dir:0:1} == '~' ]] && the_new_dir="${HOME}${the_new_dir:1}"
#
# Now change to the new dir and add to the top of the stack
pushd "${the_new_dir}" > /dev/null
[[ $? -ne 0 ]] && return 1
the_new_dir=$(pwd)
#
# Trim down everything beyond 11th entry
popd -n +11 2>/dev/null 1>/dev/null
#
# Remove any other occurence of this dir, skipping the top of the stack
for ((cnt=1; cnt <= 10; cnt++)); do
x2=$(dirs +${cnt} 2>/dev/null)
[[ $? -ne 0 ]] && return 0
[[ ${x2:0:1} == '~' ]] && x2="${HOME}${x2:1}"
if [[ "${x2}" == "${the_new_dir}" ]]; then
popd -n +$cnt 2>/dev/null 1>/dev/null
cnt=cnt-1
fi
done
return 0
}
alias cd=cd_func
if [[ $BASH_VERSION > "2.05a" ]]; then
# ctrl+w shows the menu
bind -x "\"\C-w\":cd_func -- ;"
fi
This might also be a useful function to put in your .bashrc; it moves up either a number of directories, or to a named directory, i.e. if you're in /a/b/c/d/ you can do up 3 or up a to end up in a.
I have no idea where I found this; if you know, please comment or add the attribution.
function up()
{
dir=""
if [ -z "$1" ]; then
dir=..
elif [[ $1 =~ ^[0-9]+$ ]]; then
x=0
while [ $x -lt ${1:-1} ]; do
dir=${dir}../
x=$(($x+1))
done
else
dir=${PWD%/$1/*}/$1
fi
cd "$dir";
}
If you want to switch to zsh, this is very easy-- just use "alias -g" (global alias, i.e. an alias that works anywhere in the command, not just the first word).
# alias -g c=/my/super/long/dir/name
# cd c
# pwd
/my/super/long/dir/name
In bash, I think the closest thing you'll get to 'aliasing' style is to write a function:
function ccd {
case "$1" in
c) cd /blah/blah/blah/long/path/number/one ;;
foo) cd /blah/blah/totally/different path ;;
"multiword phrase") cd /tmp ;;
esac
}
This means using something other than "cd" as the command when you want a shortcut, but other than that, it's flexible; you can also add an "ls" to the function so that it always reminds you what's in the directory after you cd, etc.
(Note that to use a multiword argument as above, you need to quote it on the command line, like this:
ccd "multiword phrase"
so it's not really all that convenient. But it'll work if you need to.)
Based on Andrew Medico's suggestion, check out J
Look into pushd, which allows you to maintain a stack of directories which you can push onto, pop off of, or rearrange.
Check out autojmp or dirmarks
Management requires both fast creation and removal of directories. Create many directiories:
mkdir -p user/new_dir/new/_dir/.../new_dir
Remove recursively many directories (be very careful when you are in lower directories!):
rm -r dir/.../new_dir/
For further reading, the cheat sheet may help you:
http://www.scribd.com/doc/2082838/Bash-Command-Line-History-Cheat-Sheet
It contains some nuggets, but I find it rather hard to read. I cannot get commands, like Meta+>, working. They probably help you in navigating long directories.
I realize the question is pretty old, but none of the scripts out there satisfied me, so I wrote a new one.
Here's the requirements I had in mind:
1) Use only bash commands -- I intend to use this on many different unices -- Linux, cygwin, HP-UX, AIX, and a couple others, so I couldn't depend on grep being consistent. Luckily I do have bash everywhere I work.
2) Short code -- I wanted to be able to bind this to a key in GNU screen, and just hit that key to paste the script into the current bash shell I'm using, so that I don't have to setup bash profiles on every system I use. Anything super long would be annoying and take too much time to paste.
3) No file usage -- Don't want to be littering shared logons with random files.
4) Act just like "cd" in the normal case. Don't want to have to think about which command to use before I start typing.
5) Provide "up" usage like this answer: How to manage Long Paths in Bash?
6) Keep a list of recently used directories, and switch to the most recent.
Here's the script:
#Jump History - Isaiah Damron
function jfind() {
lp=${JNHIST//==${PWD}==/==}
lp=${lp%%${lp#==*$1*==}}
lp=${lp##${lp%==*$1*==*}}
lp=${lp//==/}
[[ -d "$lp" ]] && echo $lp && return 0
return 1;
}
function jadd() {
[[ -z "$JNHIST" ]] && export JNHIST='=='
[[ 3000 -lt ${#JNHIST} ]] && export JNHIST=${JNHIST:0:3000} && export JNHIST="${JNHIST%==*}=="
export JNHIST="==$PWD${JNHIST//==${PWD}==/==}"
}
function j() {
{ cd $* 2> /dev/null && jadd; } \
|| { cd ${PWD/$1*/}$1 2> /dev/null && jadd; } \
|| { jfind $1 \
&& { cd $( jfind $1 ) 2> /dev/null && jadd; } ; } \
|| cd $*
}
function jh() {
[[ -z "$1" ]] && echo -e ${JNHIST//==/\\n}
[[ -n "$1" ]] && jfind $1 && cd $(jfind $1) && jadd
}
Usage:
jh [parameters]
If called on its own, without any parameters, it outputs the current history list. If it has a parameter, then it searches through the history for the most recently used directory that contains the string $1, and cd's to it.
j {parameters}
Does cd parameters. If that fails, it checks if any of the parent directories of $PWD match $1, and cd's to it. If that fails, then it calls jh $1. If that fails, then it outputs the result of cd parameters
Note: I used '==' as an internal separator. Hopefully you don't have any directories that contain a '==', but if you do you'll have to change around the script. Just :%s/==/whatever/g