这是对@pini目前评分最高的解决方案(遗憾的是，它只处理少数情况)的更正，全功能改进

提醒:'-z'测试如果字符串是零长度(=空)，'-n'测试如果字符串不是空。

# both $1 and $2 are absolute paths beginning with /
# returns relative path to $2/$target from $1/$source
source=$1
target=$2

common_part=$source # for now
result="" # for now

while [[ "${target#$common_part}" == "${target}" ]]; do
    # no match, means that candidate common part is not correct
    # go up one level (reduce common part)
    common_part="$(dirname $common_part)"
    # and record that we went back, with correct / handling
    if [[ -z $result ]]; then
        result=".."
    else
        result="../$result"
    fi
done

if [[ $common_part == "/" ]]; then
    # special case for root (no common path)
    result="$result/"
fi

# since we now have identified the common part,
# compute the non-common part
forward_part="${target#$common_part}"

# and now stick all parts together
if [[ -n $result ]] && [[ -n $forward_part ]]; then
    result="$result$forward_part"
elif [[ -n $forward_part ]]; then
    # extra slash removal
    result="${forward_part:1}"
fi

echo $result

测试用例:

compute_relative.sh "/A/B/C" "/A"           -->  "../.."
compute_relative.sh "/A/B/C" "/A/B"         -->  ".."
compute_relative.sh "/A/B/C" "/A/B/C"       -->  ""
compute_relative.sh "/A/B/C" "/A/B/C/D"     -->  "D"
compute_relative.sh "/A/B/C" "/A/B/C/D/E"   -->  "D/E"
compute_relative.sh "/A/B/C" "/A/B/D"       -->  "../D"
compute_relative.sh "/A/B/C" "/A/B/D/E"     -->  "../D/E"
compute_relative.sh "/A/B/C" "/A/D"         -->  "../../D"
compute_relative.sh "/A/B/C" "/A/D/E"       -->  "../../D/E"
compute_relative.sh "/A/B/C" "/D/E/F"       -->  "../../../D/E/F"

假设您已经安装了:bash、pwd、dirname、echo;relpath是

#!/bin/bash
s=$(cd ${1%%/};pwd); d=$(cd $2;pwd); b=; while [ "${d#$s/}" == "${d}" ]
do s=$(dirname $s);b="../${b}"; done; echo ${b}${d#$s/}

我从pini和其他一些想法中得到了答案

注意:这要求两个路径都是现有文件夹。文件将无法工作。

另一个解决方案，纯bash + GNU readlink，在以下上下文中易于使用:

ln -s "$(relpath "$A" "$B")" "$B"

编辑:确保“$B”是不存在或没有软链接在这种情况下，否则relpath遵循这个链接，这不是你想要的!

这几乎适用于当前所有的Linux。如果readlink -m在您这边不起作用，请尝试readlink -f。请参见https://gist.github.com/hilbix/1ec361d00a8178ae8ea0查看可能的更新:

: relpath A B
# Calculate relative path from A to B, returns true on success
# Example: ln -s "$(relpath "$A" "$B")" "$B"
relpath()
{
local X Y A
# We can create dangling softlinks
X="$(readlink -m -- "$1")" || return
Y="$(readlink -m -- "$2")" || return
X="${X%/}/"
A=""
while   Y="${Y%/*}"
        [ ".${X#"$Y"/}" = ".$X" ]
do
        A="../$A"
done
X="$A${X#"$Y"/}"
X="${X%/}"
echo "${X:-.}"
}

注:

Care was taken that it is safe against unwanted shell meta character expansion, in case filenames contain * or ?. The output is meant to be usable as the first argument to ln -s: relpath / / gives . and not the empty string relpath a a gives a, even if a happens to be a directory Most common cases were tested to give reasonable results, too. This solution uses string prefix matching, hence readlink is required to canonicalize paths. Thanks to readlink -m it works for not yet existing paths, too.

在旧系统上，readlink -m不可用，如果文件不存在，readlink -f将失败。所以你可能需要一些像这样的解决方法(未经测试!):

readlink_missing()
{
readlink -m -- "$1" && return
readlink -f -- "$1" && return
[ -e . ] && echo "$(readlink_missing "$(dirname "$1")")/$(basename "$1")"
}

这在$1包含的情况下是不正确的。或. .对于不存在的路径(如/doesnotexist/./a)，但它应该涵盖大多数情况。

(用readlink_missing替换上面的readlink -m——)

编辑，因为下面是反对票

下面是一个测试，这个函数确实是正确的:

check()
{
res="$(relpath "$2" "$1")"
[ ".$res" = ".$3" ] && return
printf ':WRONG: %-10q %-10q gives %q\nCORRECT %-10q %-10q gives %q\n' "$1" "$2" "$res" "$@"
}

#     TARGET   SOURCE         RESULT
check "/A/B/C" "/A"           ".."
check "/A/B/C" "/A.x"         "../../A.x"
check "/A/B/C" "/A/B"         "."
check "/A/B/C" "/A/B/C"       "C"
check "/A/B/C" "/A/B/C/D"     "C/D"
check "/A/B/C" "/A/B/C/D/E"   "C/D/E"
check "/A/B/C" "/A/B/D"       "D"
check "/A/B/C" "/A/B/D/E"     "D/E"
check "/A/B/C" "/A/D"         "../D"
check "/A/B/C" "/A/D/E"       "../D/E"
check "/A/B/C" "/D/E/F"       "../../D/E/F"

check "/foo/baz/moo" "/foo/bar" "../bar"

困惑吗?好吧，这是正确的结果!即使你认为它不符合问题，以下是正确的证明:

check "http://example.com/foo/baz/moo" "http://example.com/foo/bar" "../bar"

毫无疑问，……/bar是从页面moo中看到的页面栏的准确且唯一正确的相对路径。其他一切都是完全错误的。

采用问题的输出很简单，显然假设current是一个目录:

absolute="/foo/bar"
current="/foo/baz/foo"
relative="../$(relpath "$absolute" "$current")"

这将返回所请求的内容。

在你感到惊讶之前，这里有一个稍微复杂一点的relpath变体(注意细微的区别)，它也应该适用于url语法(因此，由于一些bash魔法，末尾/幸存下来):

# Calculate relative PATH to the given DEST from the given BASE
# In the URL case, both URLs must be absolute and have the same Scheme.
# The `SCHEME:` must not be present in the FS either.
# This way this routine works for file paths an
: relpathurl DEST BASE
relpathurl()
{
local X Y A
# We can create dangling softlinks
X="$(readlink -m -- "$1")" || return
Y="$(readlink -m -- "$2")" || return
X="${X%/}/${1#"${1%/}"}"
Y="${Y%/}${2#"${2%/}"}"
A=""
while   Y="${Y%/*}"
        [ ".${X#"$Y"/}" = ".$X" ]
do
        A="../$A"
done
X="$A${X#"$Y"/}"
X="${X%/}"
echo "${X:-.}"
}

这里有一些检查，只是为了弄清楚:它确实像所说的那样工作。

check()
{
res="$(relpathurl "$2" "$1")"
[ ".$res" = ".$3" ] && return
printf ':WRONG: %-10q %-10q gives %q\nCORRECT %-10q %-10q gives %q\n' "$1" "$2" "$res" "$@"
}

#     TARGET   SOURCE         RESULT
check "/A/B/C" "/A"           ".."
check "/A/B/C" "/A.x"         "../../A.x"
check "/A/B/C" "/A/B"         "."
check "/A/B/C" "/A/B/C"       "C"
check "/A/B/C" "/A/B/C/D"     "C/D"
check "/A/B/C" "/A/B/C/D/E"   "C/D/E"
check "/A/B/C" "/A/B/D"       "D"
check "/A/B/C" "/A/B/D/E"     "D/E"
check "/A/B/C" "/A/D"         "../D"
check "/A/B/C" "/A/D/E"       "../D/E"
check "/A/B/C" "/D/E/F"       "../../D/E/F"

check "/foo/baz/moo" "/foo/bar" "../bar"
check "http://example.com/foo/baz/moo" "http://example.com/foo/bar" "../bar"

check "http://example.com/foo/baz/moo/" "http://example.com/foo/bar" "../../bar"
check "http://example.com/foo/baz/moo"  "http://example.com/foo/bar/" "../bar/"
check "http://example.com/foo/baz/moo/"  "http://example.com/foo/bar/" "../../bar/"

下面是如何用它从问题中得到想要的结果:

absolute="/foo/bar"
current="/foo/baz/foo"
relative="$(relpathurl "$absolute" "$current/")"
echo "$relative"

如果你发现什么东西不起作用，请在下面的评论中告诉我。谢谢。

PS:

为什么relpath的论点与这里的所有其他答案相反?

如果你改变

Y="$(readlink -m -- "$2")" || return

to

Y="$(readlink -m -- "${2:-"$PWD"}")" || return

然后你可以去掉第二个参数，这样BASE就是当前目录/URL/任何东西。这只是Unix的原则。

#!/bin/bash
# both $1 and $2 are absolute paths
# returns $2 relative to $1

source=$1
target=$2

common_part=$source
back=
while [ "${target#$common_part}" = "${target}" ]; do
  common_part=$(dirname $common_part)
  back="../${back}"
done

echo ${back}${target#$common_part/}

这个脚本只对路径名有效。它不需要任何文件存在。如果传递的路径不是绝对的，那么行为就有点不寻常，但是如果两条路径都是相对的，那么应该能正常工作。

我只在OS X上测试过，所以可能不太便携。

#!/bin/bash
set -e
declare SCRIPT_NAME="$(basename $0)"
function usage {
    echo "Usage: $SCRIPT_NAME <base path> <target file>"
    echo "       Outputs <target file> relative to <base path>"
    exit 1
}

if [ $# -lt 2 ]; then usage; fi

declare base=$1
declare target=$2
declare -a base_part=()
declare -a target_part=()

#Split path elements & canonicalize
OFS="$IFS"; IFS='/'
bpl=0;
for bp in $base; do
    case "$bp" in
        ".");;
        "..") let "bpl=$bpl-1" ;;
        *) base_part[${bpl}]="$bp" ; let "bpl=$bpl+1";;
    esac
done
tpl=0;
for tp in $target; do
    case "$tp" in
        ".");;
        "..") let "tpl=$tpl-1" ;;
        *) target_part[${tpl}]="$tp" ; let "tpl=$tpl+1";;
    esac
done
IFS="$OFS"

#Count common prefix
common=0
for (( i=0 ; i<$bpl ; i++ )); do
    if [ "${base_part[$i]}" = "${target_part[$common]}" ] ; then
        let "common=$common+1"
    else
        break
    fi
done

#Compute number of directories up
let "updir=$bpl-$common" || updir=0 #if the expression is zero, 'let' fails

#trivial case (after canonical decomposition)
if [ $updir -eq 0 ]; then
    echo .
    exit
fi

#Print updirs
for (( i=0 ; i<$updir ; i++ )); do
    echo -n ../
done

#Print remaining path
for (( i=$common ; i<$tpl ; i++ )); do
    if [ $i -ne $common ]; then
        echo -n "/"
    fi
    if [ "" != "${target_part[$i]}" ] ; then
        echo -n "${target_part[$i]}"
    fi
done
#One last newline
echo

test.sh:

#!/bin/bash                                                                 

cd /home/ubuntu
touch blah
TEST=/home/ubuntu/.//blah
echo TEST=$TEST
TMP=$(readlink -e "$TEST")
echo TMP=$TMP
REL=${TMP#$(pwd)/}
echo REL=$REL

测试:

$ ./test.sh 
TEST=/home/ubuntu/.//blah
TMP=/home/ubuntu/blah
REL=blah