在Bash中,测试数组是否包含某个值的最简单方法是什么?


当前回答

The answer with most votes is very concise and clean, but it can have false positives when a space is part of one of the array elements. This can be overcome when changing IFS and using "${array[*]}" instead of "${array[@]}". The method is identical, but it looks less clean. By using "${array[*]}", we print all elements of $array, separated by the first character in IFS. So by choosing a correct IFS, you can overcome this particular issue. In this particular case, we decide to set IFS to an uncommon character $'\001' which stands for Start of Heading (SOH)

$ array=("foo bar" "baz" "qux")
$ IFS=$'\001'
$ [[ "$IFS${array[*]}$IFS" =~ "${IFS}foo${IFS}" ]] && echo yes || echo no
no
$ [[ "$IFS${array[*]}$IFS" =~ "${IFS}foo bar${IFS}" ]] && echo yes || echo no
yes
$ unset IFS

这解决了大多数假阳性问题,但需要一个好的IFS选择。

注意:如果之前设置了IFS,最好保存并重新设置,而不是使用未设置的IFS


相关:

访问bash命令行参数$@ vs $*

其他回答

一行的解决方案

printf '%s\0' "${myarray[@]}" | grep -F -x -z -- 'myvalue'

解释

printf语句打印数组中的每个元素,以空字符分隔。

grep语句使用以下标志来匹配一个包含myvalue字符串的项(不多不少):

-z/——null-data -行以0字节而不是换行符结束。 -f /——fixed-strings -将pattern解释为固定字符串,而不是正则表达式。 -x/——line-regexp -只选择与整行完全匹配的匹配项。 ——-标记命令行选项的结束,使Grep处理“myvalue”作为一个非选项参数,即使它以破折号开始

为什么我们使用空字节\0而不是换行符\n?数组的元素实际上可能包含换行符。(如果您知道它没有,请随意删除-z grep选项,并将%s\n替换为您的第一个printf参数。)


使用

把这个放进一个if…然后声明:

if printf '%s\0' "${myarray[@]}" | grep -Fxqz -- 'myvalue'; then
    # ...
fi

我在grep表达式中添加了-q标志,这样它就不会打印匹配项;它只会将匹配的存在视为“真”。

更新:感谢presto8指出——line-regexp标志。谢谢Tino,你指出了数组项中可以存在换行符的情况。

保持简单:

Array1=( "item1" "item2" "item3" "item-4" )
var="item3"

count=$(echo ${Array1[@]} | tr ' ' '\n' | awk '$1 == "'"$var"'"{print $0}' | wc -l)
[ $count -eq 0 ] && echo "Not found" || echo "found"

借鉴Dennis Williamson的答案,下面的解决方案结合了数组、shell-safe引号和正则表达式,以避免需要:遍历循环;使用管道或其他子过程;或者使用非bash实用程序。

declare -a array=('hello, stack' one 'two words' words last)
printf -v array_str -- ',,%q' "${array[@]}"

if [[ "${array_str},," =~ ,,words,, ]]
then
   echo 'Matches'
else
   echo "Doesn't match"
fi

上面的代码通过使用Bash正则表达式来匹配数组内容的字符串化版本。有六个重要的步骤来确保正则表达式匹配不会被数组中的值的巧妙组合所欺骗:

Construct the comparison string by using Bash's built-in printf shell-quoting, %q. Shell-quoting will ensure that special characters become "shell-safe" by being escaped with backslash \. Choose a special character to serve as a value delimiter. The delimiter HAS to be one of the special characters that will become escaped when using %q; that's the only way to guarantee that values within the array can't be constructed in clever ways to fool the regular expression match. I choose comma , because that character is the safest when eval'd or misused in an otherwise unexpected way. Combine all array elements into a single string, using two instances of the special character to serve as delimiter. Using comma as an example, I used ,,%q as the argument to printf. This is important because two instances of the special character can only appear next to each other when they appear as the delimiter; all other instances of the special character will be escaped. Append two trailing instances of the delimiter to the string, to allow matches against the last element of the array. Thus, instead of comparing against ${array_str}, compare against ${array_str},,. If the target string you're searching for is supplied by a user variable, you must escape all instances of the special character with a backslash. Otherwise, the regular expression match becomes vulnerable to being fooled by cleverly-crafted array elements. Perform a Bash regular expression match against the string.

这种方法的优点是不需要遍历所有元素(至少不是显式地)。但是由于array.c中的array_to_string_internal()仍然循环遍历数组元素并将它们连接到一个字符串中,因此它可能并不比所提出的循环解决方案更有效,但它更具可读性。

if [[ " ${array[*]} " =~ " ${value} " ]]; then
    # whatever you want to do when array contains value
fi

if [[ ! " ${array[*]} " =~ " ${value} " ]]; then
    # whatever you want to do when array doesn't contain value
fi

请注意,如果您正在搜索的值是带有空格的数组元素中的某个单词,则会给出假阳性。例如

array=("Jack Brown")
value="Jack"

正则表达式将“Jack”视为在数组中,即使它不在数组中。所以你必须改变IFS和正则表达式上的分隔符如果你仍然想使用这个解决方案,就像这样

IFS="|"
array=("Jack Brown${IFS}Jack Smith")
value="Jack"

if [[ "${IFS}${array[*]}${IFS}" =~ "${IFS}${value}${IFS}" ]]; then
    echo "true"
else
    echo "false"
fi

unset IFS # or set back to original IFS if previously set

这将打印“false”。

显然,这也可以用作测试语句,允许将其表示为一行程序

[[ " ${array[*]} " =~ " ${value} " ]] && echo "true" || echo "false"
: NeedleInArgs "$needle" "${haystack[@]}"
: NeedleInArgs "$needle" arg1 arg2 .. argN
NeedleInArgs()
{
local a b;
printf -va '\n%q\n' "$1";
printf -vb '%q\n' "${@:2}";
case $'\n'"$b" in (*"$a"*) return 0;; esac;
return 1;
}

使用:

NeedleInArgs "$needle" "${haystack[@]}" && echo "$needle" found || echo "$needle" not found;

对于bash v3.1及以上版本(printf -v支持) 没有分叉,也没有外部程序 没有循环(除了bash中的内部扩展) 适用于所有可能的值和数组,没有异常,没有什么可担心的

也可以直接使用,比如:

if      NeedleInArgs "$input" value1 value2 value3 value4;
then
        : input from the list;
else
        : input not from list;
fi;

对于从v20.5 b到v3.0的bash, printf缺少-v,因此需要额外的2个fork(但不需要执行,因为printf是bash内置的):

NeedleInArgs()
{
case $'\n'"`printf '%q\n' "${@:2}"`" in
(*"`printf '\n%q\n' "$1"`"*) return 0;;
esac;
return 1;
}

注意,我测试了时间:

check call0:  n: t4.43 u4.41 s0.00 f: t3.65 u3.64 s0.00 l: t4.91 u4.90 s0.00 N: t5.28 u5.27 s0.00 F: t2.38 u2.38 s0.00 L: t5.20 u5.20 s0.00
check call1:  n: t3.41 u3.40 s0.00 f: t2.86 u2.84 s0.01 l: t3.72 u3.69 s0.02 N: t4.01 u4.00 s0.00 F: t1.15 u1.15 s0.00 L: t4.05 u4.05 s0.00
check call2:  n: t3.52 u3.50 s0.01 f: t3.74 u3.73 s0.00 l: t3.82 u3.80 s0.01 N: t2.67 u2.67 s0.00 F: t2.64 u2.64 s0.00 L: t2.68 u2.68 s0.00

Call0和call1是对另一个快速pure-bash变体调用的不同变体 Call2在这里。 N=notfound F=firstmatch L=lastmatch 小写字母为短数组,大写字母为长数组

正如您所看到的,这里的这个变体有一个非常稳定的运行时,所以它不太依赖于匹配位置。运行时主要由数组长度决定。搜索变量的运行时高度依赖于匹配位置。所以在边缘情况下,这个变体可以(快得多)。

但非常重要的是,搜索变量的RAM效率更高,因为这里的这个变量总是将整个数组转换为一个大字符串。

所以如果你的内存很紧,你希望大部分比赛都是早期的,那么就不要在这里使用这个。但是,如果您想要一个可预测的运行时,有很长的数组来匹配(期望延迟或根本不匹配),并且双RAM使用也不是太大的问题,那么这里有一些优势。

定时测试脚本:

in_array()
{
    local needle="$1" arrref="$2[@]" item
    for item in "${!arrref}"; do
        [[ "${item}" == "${needle}" ]] && return 0
    done
    return 1
}

NeedleInArgs()
{
local a b;
printf -va '\n%q\n' "$1";
printf -vb '%q\n' "${@:2}";
case $'\n'"$b" in (*"$a"*) return 0;; esac;
return 1;
}

loop1() { for a in {1..100000}; do "$@"; done }
loop2() { for a in {1..1000}; do "$@"; done }

run()
{
  needle="$5"
  arr=("${@:6}")

  out="$( ( time -p "loop$2" "$3" ) 2>&1 )"

  ret="$?"
  got="${out}"
  syst="${got##*sys }"
  got="${got%"sys $syst"}"
  got="${got%$'\n'}"
  user="${got##*user }"
  got="${got%"user $user"}"
  got="${got%$'\n'}"
  real="${got##*real }"
  got="${got%"real $real"}"
  got="${got%$'\n'}"
  printf ' %s: t%q u%q s%q' "$1" "$real" "$user" "$syst"
  [ -z "$rest" ] && [ "$ret" = "$4" ] && return
  printf 'FAIL! expected %q got %q\n' "$4" "$ret"
  printf 'call:   %q\n' "$3"
  printf 'out:    %q\n' "$out"
  printf 'rest:   %q\n' "$rest"
  printf 'needle: %q\n' "$5"
  printf 'arr:   '; printf ' %q' "${@:6}"; printf '\n'
  exit 1
}

check()
{
  printf 'check %q: ' "$1"
  run n 1 "$1" 1 needle a b c d
  run f 1 "$1" 0 needle needle a b c d
  run l 1 "$1" 0 needle a b c d needle
  run N 2 "$1" 1 needle "${rnd[@]}"
  run F 2 "$1" 0 needle needle "${rnd[@]}"
  run L 2 "$1" 0 needle "${rnd[@]}" needle
  printf '\n'
}

call0() { chk=("${arr[@]}"); in_array "$needle" chk; }
call1() { in_array "$needle" arr; }
call2() { NeedleInArgs "$needle" "${arr[@]}"; }

rnd=()
for a in {1..1000}; do rnd+=("$a"); done

check call0
check call1
check call2