安装php后，我使用这种方式:

URL_ENCODED_DATA=`php -r "echo urlencode('$DATA');"`

有什么能比javascript更好地解析url呢?

node -p "encodeURIComponent('$url')"

下面是一个使用Lua的单行转换，类似于blueyed的答案，除了所有的RFC 3986 Unreserved Characters都没有被编码(就像这个答案):

url=$(echo 'print((arg[1]:gsub("([^%w%-%.%_%~])",function(c)return("%%%02X"):format(c:byte())end)))' | lua - "$1")

此外，您可能需要确保字符串中的换行符从LF转换为CRLF，在这种情况下，您可以插入一个gsub("\r? "“\n”，“\r\n”)在百分比编码之前。

下面是一个变体，在application/x-www-form-urlencoded的非标准风格中，它执行换行规范化，并将空格编码为'+'而不是'%20'(可以使用类似的技术将其添加到Perl代码片段中)。

url=$(echo 'print((arg[1]:gsub("\r?\n", "\r\n"):gsub("([^%w%-%.%_%~ ]))",function(c)return("%%%02X"):format(c:byte())end):gsub(" ","+"))' | lua - "$1")

awk版本的直接链接:http://www.shelldorado.com/scripts/cmds/urlencode 我用了很多年了，效果很好

:
##########################################################################
# Title      :  urlencode - encode URL data
# Author     :  Heiner Steven (heiner.steven@odn.de)
# Date       :  2000-03-15
# Requires   :  awk
# Categories :  File Conversion, WWW, CGI
# SCCS-Id.   :  @(#) urlencode  1.4 06/10/29
##########################################################################
# Description
#   Encode data according to
#       RFC 1738: "Uniform Resource Locators (URL)" and
#       RFC 1866: "Hypertext Markup Language - 2.0" (HTML)
#
#   This encoding is used i.e. for the MIME type
#   "application/x-www-form-urlencoded"
#
# Notes
#    o  The default behaviour is not to encode the line endings. This
#   may not be what was intended, because the result will be
#   multiple lines of output (which cannot be used in an URL or a
#   HTTP "POST" request). If the desired output should be one
#   line, use the "-l" option.
#
#    o  The "-l" option assumes, that the end-of-line is denoted by
#   the character LF (ASCII 10). This is not true for Windows or
#   Mac systems, where the end of a line is denoted by the two
#   characters CR LF (ASCII 13 10).
#   We use this for symmetry; data processed in the following way:
#       cat | urlencode -l | urldecode -l
#   should (and will) result in the original data
#
#    o  Large lines (or binary files) will break many AWK
#       implementations. If you get the message
#       awk: record `...' too long
#        record number xxx
#   consider using GNU AWK (gawk).
#
#    o  urlencode will always terminate it's output with an EOL
#       character
#
# Thanks to Stefan Brozinski for pointing out a bug related to non-standard
# locales.
#
# See also
#   urldecode
##########################################################################

PN=`basename "$0"`          # Program name
VER='1.4'

: ${AWK=awk}

Usage () {
    echo >&2 "$PN - encode URL data, $VER
usage: $PN [-l] [file ...]
    -l:  encode line endings (result will be one line of output)

The default is to encode each input line on its own."
    exit 1
}

Msg () {
    for MsgLine
    do echo "$PN: $MsgLine" >&2
    done
}

Fatal () { Msg "$@"; exit 1; }

set -- `getopt hl "$@" 2>/dev/null` || Usage
[ $# -lt 1 ] && Usage           # "getopt" detected an error

EncodeEOL=no
while [ $# -gt 0 ]
do
    case "$1" in
        -l) EncodeEOL=yes;;
    --) shift; break;;
    -h) Usage;;
    -*) Usage;;
    *)  break;;         # First file name
    esac
    shift
done

LANG=C  export LANG
$AWK '
    BEGIN {
    # We assume an awk implementation that is just plain dumb.
    # We will convert an character to its ASCII value with the
    # table ord[], and produce two-digit hexadecimal output
    # without the printf("%02X") feature.

    EOL = "%0A"     # "end of line" string (encoded)
    split ("1 2 3 4 5 6 7 8 9 A B C D E F", hextab, " ")
    hextab [0] = 0
    for ( i=1; i<=255; ++i ) ord [ sprintf ("%c", i) "" ] = i + 0
    if ("'"$EncodeEOL"'" == "yes") EncodeEOL = 1; else EncodeEOL = 0
    }
    {
    encoded = ""
    for ( i=1; i<=length ($0); ++i ) {
        c = substr ($0, i, 1)
        if ( c ~ /[a-zA-Z0-9.-]/ ) {
        encoded = encoded c     # safe character
        } else if ( c == " " ) {
        encoded = encoded "+"   # special handling
        } else {
        # unsafe character, encode it as a two-digit hex-number
        lo = ord [c] % 16
        hi = int (ord [c] / 16);
        encoded = encoded "%" hextab [hi] hextab [lo]
        }
    }
    if ( EncodeEOL ) {
        printf ("%s", encoded EOL)
    } else {
        print encoded
    }
    }
    END {
        #if ( EncodeEOL ) print ""
    }
' "$@"

Orwellophile给出了一个很好的答案，它包含了一个纯bash选项(函数rawurlencode)，我在我的网站上使用过(基于shell的CGI脚本，响应搜索请求的大量url)。唯一的缺点是高峰期间CPU过高。

我找到了一个改进的解决方案，利用bash的“全局替换”特性。有了这个解决方案，url编码的处理时间快了4倍。解决方案确定要转义的字符，并使用“全局替换”操作符(${var//source/replacement})来处理所有替换。这种速度的提高显然来自于使用bash内部循环，而不是显式循环。

性能:核心i3-8100 3.60Ghz。测试用例:来自堆栈溢出的1000个URL，类似于这个票据:“https://stackoverflow.com/questions/296536/how-to-urlencode-data-for-curl-command”。

现有解决方案:0.807秒 优化方案:0.162秒(5倍加速)

url_encode()
{
    local key="${1}" varname="${2:-_rval}" prefix="${3:-_ENCKEY_}"
    local unsafe=${key//[-_.~a-zA-Z0-9 ]/} 
    local -i key_len=${#unsafe}
    local ch ch1 ch0

    while [ "$unsafe" ] ;do
        ch=${unsafe:0:1}
        ch0="\\$ch"
        printf -v ch1 '%%%02x' "'$ch'" 
        key=${key//$ch0/"$ch1"}
        unsafe=${unsafe//"$ch0"}
    done
    key=${key// /+} 

    REPLY="$key"
    # printf "%s" "$REPLY"
    return 0
}

作为一个次要的额外字符，它使用'+'来编码空格。稍微紧凑的URL。

基准:

function t {
    local key
    for (( i=1 ; i<=$1 ; i++ )) do url_encode "$2" kkk2 ; done
    echo "K=$REPLY"
}

t 1000 "https://stackoverflow.com/questions/296536/how-to-urlencode-data-for-curl-command"

Uni2ascii非常方便:

$ echo -ne '你好世界' | uni2ascii -aJ
%E4%BD%A0%E5%A5%BD%E4%B8%96%E7%95%8C