对于那些正在寻找不需要perl的解决方案的人来说，这里有一个只需要hexdump和awk的解决方案:

url_encode() {
 [ $# -lt 1 ] && { return; }

 encodedurl="$1";

 # make sure hexdump exists, if not, just give back the url
 [ ! -x "/usr/bin/hexdump" ] && { return; }

 encodedurl=`
   echo $encodedurl | hexdump -v -e '1/1 "%02x\t"' -e '1/1 "%_c\n"' |
   LANG=C awk '
     $1 == "20"                    { printf("%s",   "+"); next } # space becomes plus
     $1 ~  /0[adAD]/               {                      next } # strip newlines
     $2 ~  /^[a-zA-Z0-9.*()\/-]$/  { printf("%s",   $2);  next } # pass through what we can
                                   { printf("%%%s", $1)        } # take hex value of everything else
   '`
}

通过网络上的几个地方和当地的一些试验和错误拼接在一起。效果很好!

下面是一个POSIX函数:

url_encode() {
   awk 'BEGIN {
      for (n = 0; n < 125; n++) {
         m[sprintf("%c", n)] = n
      }
      n = 1
      while (1) {
         s = substr(ARGV[1], n, 1)
         if (s == "") {
            break
         }
         t = s ~ /[[:alnum:]_.!~*\47()-]/ ? t s : t sprintf("%%%02X", m[s])
         n++
      }
      print t
   }' "$1"
}

例子:

value=$(url_encode "$2")

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 ($0, i, 1)
        if ( c ~ /[a-zA-Z0-9.-]/ ) {
        encoded = encoded c     # safe character
        } else if ( c == " " ) {
        encoded = encoded "+"   # special handling
        } else {
        # unsafe character, encode it as a two-digit hex-number
        lo = ord [c] % 16
        hi = int (ord [c] / 16);
        encoded = encoded "%" hextab [hi] hextab [lo]
        }
    }
    if ( EncodeEOL ) {
        printf ("%s", encoded EOL)
    } else {
        print encoded
    }
    }
    END {
        #if ( EncodeEOL ) print ""
    }
' "$@"

使用curl——data-urlencode;来自man curl:

它发布数据，类似于其他——data选项，不同的是它执行url编码。为了与cgi兼容，<data>部分应该以名称开头，后面跟着分隔符和内容规范。

使用示例:

curl \
    --data-urlencode "paramName=value" \
    --data-urlencode "secondParam=value" \
    http://example.com

有关更多信息，请参阅手册页。

这需要curl 7.18.0或更新版本(发布于2008年1月)。使用curl -V来检查您拥有的版本。

你也可以对查询字符串进行编码:

curl --get \
    --data-urlencode "p1=value 1" \
    --data-urlencode "p2=value 2" \
    http://example.com
    # http://example.com?p1=value%201&p2=value%202

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"

在bash脚本的第二行中使用Perl的URI::Escape模块和uri_escape函数:

...

value="$(perl -MURI::Escape -e 'print uri_escape($ARGV[0]);' "$2")"
...

编辑:修复引用问题，正如Chris Johnsen在评论中建议的那样。谢谢!