例子:

absolute="/foo/bar"
current="/foo/baz/foo"

# Magic

relative="../../bar"

我如何创造魔法(希望不是太复杂的代码…)?


当前回答

我将只使用Perl来完成这个不那么简单的任务:

absolute="/foo/bar"
current="/foo/baz/foo"

# Perl is magic
relative=$(perl -MFile::Spec -e 'print File::Spec->abs2rel("'$absolute'","'$current'")')

其他回答

#!/bin/sh

# Return relative path from canonical absolute dir path $1 to canonical
# absolute dir path $2 ($1 and/or $2 may end with one or no "/").
# Does only need POSIX shell builtins (no external command)
relPath () {
    local common path up
    common=${1%/} path=${2%/}/
    while test "${path#"$common"/}" = "$path"; do
        common=${common%/*} up=../$up
    done
    path=$up${path#"$common"/}; path=${path%/}; printf %s "${path:-.}"
}

# Return relative path from dir $1 to dir $2 (Does not impose any
# restrictions on $1 and $2 but requires GNU Core Utility "readlink"
# HINT: busybox's "readlink" does not support option '-m', only '-f'
#       which requires that all but the last path component must exist)
relpath () { relPath "$(readlink -m "$1")" "$(readlink -m "$2")"; }

上面的shell脚本是受pini的启发(谢谢!)它会触发一个错误 在Stack Overflow的语法高亮显示模块中(至少在我的预览中是这样) 帧)。因此,如果高亮显示不正确,请忽略。

一些注意事项:

Removed errors and improved code without significantly increasing code length and complexity Put functionality into functions for easiness of use Kept functions POSIX compatible so that they (should) work with all POSIX shells (tested with dash, bash, and zsh in Ubuntu Linux 12.04) Used local variables only to avoid clobbering global variables and polluting the global name space Both directory paths DO NOT need to exist (requirement for my application) Pathnames may contain spaces, special characters, control characters, backslashes, tabs, ', ", ?, *, [, ], etc. Core function "relPath" uses POSIX shell builtins only but requires canonical absolute directory paths as parameters Extended function "relpath" can handle arbitrary directory paths (also relative, non-canonical) but requires external GNU core utility "readlink" Avoided builtin "echo" and used builtin "printf" instead for two reasons: Due to conflicting historical implementations of builtin "echo" it behaves differently in different shells -> POSIX recommends that printf is preferred over echo. Builtin "echo" of some POSIX shells will interpret some backslash sequences and thus corrupt pathnames containing such sequences To avoid unnecessary conversions, pathnames are used as they are returned and expected by shell and OS utilities (e.g. cd, ln, ls, find, mkdir; unlike python's "os.path.relpath" which will interpret some backslash sequences) Except for the mentioned backslash sequences the last line of function "relPath" outputs pathnames compatible to python: path=$up${path#"$common"/}; path=${path%/}; printf %s "${path:-.}" Last line can be replaced (and simplified) by line printf %s "$up${path#"$common"/}" I prefer the latter because Filenames can be directly appended to dir paths obtained by relPath, e.g.: ln -s "$(relpath "<fromDir>" "<toDir>")<file>" "<fromDir>" Symbolic links in the same dir created with this method do not have the ugly "./" prepended to the filename. If you find an error please contact linuxball (at) gmail.com and I'll try to fix it. Added regression test suite (also POSIX shell compatible)

回归测试的代码清单(只需将其附加到shell脚本):

############################################################################
# If called with 2 arguments assume they are dir paths and print rel. path #
############################################################################

test "$#" = 2 && {
    printf '%s\n' "Rel. path from '$1' to '$2' is '$(relpath "$1" "$2")'."
    exit 0
}

#######################################################
# If NOT called with 2 arguments run regression tests #
#######################################################

format="\t%-19s %-22s %-27s %-8s %-8s %-8s\n"
printf \
"\n\n*** Testing own and python's function with canonical absolute dirs\n\n"
printf "$format\n" \
    "From Directory" "To Directory" "Rel. Path" "relPath" "relpath" "python"
IFS=
while read -r p; do
    eval set -- $p
    case $1 in '#'*|'') continue;; esac # Skip comments and empty lines
    # q stores quoting character, use " if ' is used in path name
    q="'"; case $1$2 in *"'"*) q='"';; esac
    rPOk=passed rP=$(relPath "$1" "$2"); test "$rP" = "$3" || rPOk=$rP
    rpOk=passed rp=$(relpath "$1" "$2"); test "$rp" = "$3" || rpOk=$rp
    RPOk=passed
    RP=$(python -c "import os.path; print os.path.relpath($q$2$q, $q$1$q)")
    test "$RP" = "$3" || RPOk=$RP
    printf \
    "$format" "$q$1$q" "$q$2$q" "$q$3$q" "$q$rPOk$q" "$q$rpOk$q" "$q$RPOk$q"
done <<-"EOF"
    # From directory    To directory           Expected relative path

    '/'                 '/'                    '.'
    '/usr'              '/'                    '..'
    '/usr/'             '/'                    '..'
    '/'                 '/usr'                 'usr'
    '/'                 '/usr/'                'usr'
    '/usr'              '/usr'                 '.'
    '/usr/'             '/usr'                 '.'
    '/usr'              '/usr/'                '.'
    '/usr/'             '/usr/'                '.'
    '/u'                '/usr'                 '../usr'
    '/usr'              '/u'                   '../u'
    "/u'/dir"           "/u'/dir"              "."
    "/u'"               "/u'/dir"              "dir"
    "/u'/dir"           "/u'"                  ".."
    "/"                 "/u'/dir"              "u'/dir"
    "/u'/dir"           "/"                    "../.."
    "/u'"               "/u'"                  "."
    "/"                 "/u'"                  "u'"
    "/u'"               "/"                    ".."
    '/u"/dir'           '/u"/dir'              '.'
    '/u"'               '/u"/dir'              'dir'
    '/u"/dir'           '/u"'                  '..'
    '/'                 '/u"/dir'              'u"/dir'
    '/u"/dir'           '/'                    '../..'
    '/u"'               '/u"'                  '.'
    '/'                 '/u"'                  'u"'
    '/u"'               '/'                    '..'
    '/u /dir'           '/u /dir'              '.'
    '/u '               '/u /dir'              'dir'
    '/u /dir'           '/u '                  '..'
    '/'                 '/u /dir'              'u /dir'
    '/u /dir'           '/'                    '../..'
    '/u '               '/u '                  '.'
    '/'                 '/u '                  'u '
    '/u '               '/'                    '..'
    '/u\n/dir'          '/u\n/dir'             '.'
    '/u\n'              '/u\n/dir'             'dir'
    '/u\n/dir'          '/u\n'                 '..'
    '/'                 '/u\n/dir'             'u\n/dir'
    '/u\n/dir'          '/'                    '../..'
    '/u\n'              '/u\n'                 '.'
    '/'                 '/u\n'                 'u\n'
    '/u\n'              '/'                    '..'

    '/    a   b/å/⮀*/!' '/    a   b/å/⮀/xäå/?' '../../⮀/xäå/?'
    '/'                 '/A'                   'A'
    '/A'                '/'                    '..'
    '/  & /  !/*/\\/E'  '/'                    '../../../../..'
    '/'                 '/  & /  !/*/\\/E'     '  & /  !/*/\\/E'
    '/  & /  !/*/\\/E'  '/  & /  !/?/\\/E/F'   '../../../?/\\/E/F'
    '/X/Y'              '/  & /  !/C/\\/E/F'   '../../  & /  !/C/\\/E/F'
    '/  & /  !/C'       '/A'                   '../../../A'
    '/A /  !/C'         '/A /B'                '../../B'
    '/Â/  !/C'          '/Â/  !/C'             '.'
    '/  & /B / C'       '/  & /B / C/D'        'D'
    '/  & /  !/C'       '/  & /  !/C/\\/Ê'     '\\/Ê'
    '/Å/  !/C'          '/Å/  !/D'             '../D'
    '/.A /*B/C'         '/.A /*B/\\/E'         '../\\/E'
    '/  & /  !/C'       '/  & /D'              '../../D'
    '/  & /  !/C'       '/  & /\\/E'           '../../\\/E'
    '/  & /  !/C'       '/\\/E/F'              '../../../\\/E/F'
    '/home/p1/p2'       '/home/p1/p3'          '../p3'
    '/home/p1/p2'       '/home/p4/p5'          '../../p4/p5'
    '/home/p1/p2'       '/work/p6/p7'          '../../../work/p6/p7'
    '/home/p1'          '/work/p1/p2/p3/p4'    '../../work/p1/p2/p3/p4'
    '/home'             '/work/p2/p3'          '../work/p2/p3'
    '/'                 '/work/p2/p3/p4'       'work/p2/p3/p4'
    '/home/p1/p2'       '/home/p1/p2/p3/p4'    'p3/p4'
    '/home/p1/p2'       '/home/p1/p2/p3'       'p3'
    '/home/p1/p2'       '/home/p1/p2'          '.'
    '/home/p1/p2'       '/home/p1'             '..'
    '/home/p1/p2'       '/home'                '../..'
    '/home/p1/p2'       '/'                    '../../..'
    '/home/p1/p2'       '/work'                '../../../work'
    '/home/p1/p2'       '/work/p1'             '../../../work/p1'
    '/home/p1/p2'       '/work/p1/p2'          '../../../work/p1/p2'
    '/home/p1/p2'       '/work/p1/p2/p3'       '../../../work/p1/p2/p3'
    '/home/p1/p2'       '/work/p1/p2/p3/p4'    '../../../work/p1/p2/p3/p4'

    '/-'                '/-'                   '.'
    '/?'                '/?'                   '.'
    '/??'               '/??'                  '.'
    '/???'              '/???'                 '.'
    '/?*'               '/?*'                  '.'
    '/*'                '/*'                   '.'
    '/*'                '/**'                  '../**'
    '/*'                '/***'                 '../***'
    '/*.*'              '/*.**'                '../*.**'
    '/*.???'            '/*.??'                '../*.??'
    '/[]'               '/[]'                  '.'
    '/[a-z]*'           '/[0-9]*'              '../[0-9]*'
EOF


format="\t%-19s %-22s %-27s %-8s %-8s\n"
printf "\n\n*** Testing own and python's function with arbitrary dirs\n\n"
printf "$format\n" \
    "From Directory" "To Directory" "Rel. Path" "relpath" "python"
IFS=
while read -r p; do
    eval set -- $p
    case $1 in '#'*|'') continue;; esac # Skip comments and empty lines
    # q stores quoting character, use " if ' is used in path name
    q="'"; case $1$2 in *"'"*) q='"';; esac
    rpOk=passed rp=$(relpath "$1" "$2"); test "$rp" = "$3" || rpOk=$rp
    RPOk=passed
    RP=$(python -c "import os.path; print os.path.relpath($q$2$q, $q$1$q)")
    test "$RP" = "$3" || RPOk=$RP
    printf "$format" "$q$1$q" "$q$2$q" "$q$3$q" "$q$rpOk$q" "$q$RPOk$q"
done <<-"EOF"
    # From directory    To directory           Expected relative path

    'usr/p1/..//./p4'   'p3/../p1/p6/.././/p2' '../../p1/p2'
    './home/../../work' '..//././../dir///'    '../../dir'

    'home/p1/p2'        'home/p1/p3'           '../p3'
    'home/p1/p2'        'home/p4/p5'           '../../p4/p5'
    'home/p1/p2'        'work/p6/p7'           '../../../work/p6/p7'
    'home/p1'           'work/p1/p2/p3/p4'     '../../work/p1/p2/p3/p4'
    'home'              'work/p2/p3'           '../work/p2/p3'
    '.'                 'work/p2/p3'           'work/p2/p3'
    'home/p1/p2'        'home/p1/p2/p3/p4'     'p3/p4'
    'home/p1/p2'        'home/p1/p2/p3'        'p3'
    'home/p1/p2'        'home/p1/p2'           '.'
    'home/p1/p2'        'home/p1'              '..'
    'home/p1/p2'        'home'                 '../..'
    'home/p1/p2'        '.'                    '../../..'
    'home/p1/p2'        'work'                 '../../../work'
    'home/p1/p2'        'work/p1'              '../../../work/p1'
    'home/p1/p2'        'work/p1/p2'           '../../../work/p1/p2'
    'home/p1/p2'        'work/p1/p2/p3'        '../../../work/p1/p2/p3'
    'home/p1/p2'        'work/p1/p2/p3/p4'     '../../../work/p1/p2/p3/p4'
EOF
$ python -c "import os.path; print os.path.relpath('/foo/bar', '/foo/baz/foo')"

给:

../../bar

该脚本仅对绝对路径或没有绝对路径的相对路径的输入提供正确的结果。或者. .:

#!/bin/bash

# usage: relpath from to

if [[ "$1" == "$2" ]]
then
    echo "."
    exit
fi

IFS="/"

current=($1)
absolute=($2)

abssize=${#absolute[@]}
cursize=${#current[@]}

while [[ ${absolute[level]} == ${current[level]} ]]
do
    (( level++ ))
    if (( level > abssize || level > cursize ))
    then
        break
    fi
done

for ((i = level; i < cursize; i++))
do
    if ((i > level))
    then
        newpath=$newpath"/"
    fi
    newpath=$newpath".."
done

for ((i = level; i < abssize; i++))
do
    if [[ -n $newpath ]]
    then
        newpath=$newpath"/"
    fi
    newpath=$newpath${absolute[i]}
done

echo "$newpath"

这是我的版本。这是基于@Offirmo的回答。我使它与dash兼容,并修复了以下测试用例失败:

sh - compute-relative。“a / b / c - de - f / g " / a / b / c / def / g /" --> "../.. f - g - "

Now:

CT_FindRelativePath”a / b / c - de - f / g " / a / b / c / def / g /" --> "../../../ def - g”

查看代码:

# both $1 and $2 are absolute paths beginning with /
# returns relative path to $2/$target from $1/$source
CT_FindRelativePath()
{
    local insource=$1
    local intarget=$2

    # Ensure both source and target end with /
    # This simplifies the inner loop.
    #echo "insource : \"$insource\""
    #echo "intarget : \"$intarget\""
    case "$insource" in
        */) ;;
        *) source="$insource"/ ;;
    esac

    case "$intarget" in
        */) ;;
        *) target="$intarget"/ ;;
    esac

    #echo "source : \"$source\""
    #echo "target : \"$target\""

    local common_part=$source # for now

    local result=""

    #echo "common_part is now : \"$common_part\""
    #echo "result is now      : \"$result\""
    #echo "target#common_part : \"${target#$common_part}\""
    while [ "${target#$common_part}" = "${target}" -a "${common_part}" != "//" ]; 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
        if [ -z "${result}" ]; then
            result="../"
        else
            result="../$result"
        fi
        #echo "(w) common_part is now : \"$common_part\""
        #echo "(w) result is now      : \"$result\""
        #echo "(w) target#common_part : \"${target#$common_part}\""
    done

    #echo "(f) common_part is     : \"$common_part\""

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

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

    if [ -n "${result}" -a -n "${forward_part}" ]; then
        #echo "(simple concat)"
        result="$result$forward_part"
    elif [ -n "${forward_part}" ]; then
        result="$forward_part"
    fi
    #echo "result = \"$result\""

    # if a / was added to target and result ends in / then remove it now.
    if [ "$intarget" != "$target" ]; then
        case "$result" in
            */) result=$(echo "$result" | awk '{ string=substr($0, 1, length($0)-1); print string; }' ) ;;
        esac
    fi

    echo $result

    return 0
}

这里的答案并不是每天都能用的。由于在纯bash中很难正确地做到这一点,我建议以下可靠的解决方案(类似于注释中的一个建议):

function relpath() { 
  python -c "import os,sys;print(os.path.relpath(*(sys.argv[1:])))" "$@";
}

然后,你可以得到基于当前目录的相对路径:

echo $(relpath somepath)

或者你可以指定路径相对于给定的目录:

echo $(relpath somepath /etc)  # relative to /etc

一个缺点是这需要python,但是:

它在任何python >= 2.6中工作相同 它不要求文件或目录存在。 文件名可以包含更广泛的特殊字符。 例如,如果文件名包含 空格或其他特殊字符。 它是一个单行函数,不会使脚本混乱。

注意,包含basename或dirname的解决方案不一定更好,因为它们要求安装coreutils。如果有人有可靠而简单的纯bash解决方案(而不是令人费解的好奇心),我会感到惊讶。