BASH的保护性编程技巧-成长的对话

这是我写BASH程序的招式。这里本没有什么新的内容，但是从我的经验来看，人们爱滥用BASH。他们忽略了计算机科学，而从他们的程序中创造的是“大泥球”（译注：指架构不清晰的软件系统）。

在此我告诉你方法，以保护你的程序免于障碍，并保持代码的整洁。

不可改变的全局变量

尽量少用全局变量
以大写命名
只读声明
用全局变量来代替隐晦的$0，$1等
在我的程序中常使用的全局变量：

readonly PROGNAME=$(basename $0)
readonly PROGDIR=$(readlink -m $(dirname $0))
readonly ARGS="$@"

readonly PROGNAME = $ ( basename $ 0 )

readonly PROGDIR = $ ( readlink - m $ ( dirname $ 0 ) )

readonly ARGS = "$@"

一切皆是局部的

所有变量都应为局部的。

change_owner_of_file() {
    local filename=$1
    local user=$2
    local group=$3

    chown $user:$group $filename
}

change_owner_of_file ( ) {

local filename = $ 1

local user = $ 2

local group = $ 3

chown $ user : $ group $ filename

}

change_owner_of_files() {
    local user=$1; shift
    local group=$1; shift
    local files=$@
    local i

    for i in $files
    do
        chown $user:$group $i
    done
}

change_owner_of_files ( ) {

local user = $ 1 ; shift

local group = $ 1 ; shift

local files = $ @

local i

for i in $ files

chown $ user : $ group $ i

done

}

自注释（self documenting）的参数
通常作为循环用的变量i，把它声明为局部变量是很重要的。
局部变量不作用于全局域。

kfir@goofy ~ $ local a
bash: local: can only be used in a function

1 2	kfir @ goofy ~ $ local a bash : local : can only be used in a function

main()

有助于保持所有变量的局部性
直观的函数式编程
代码中唯一的全局命令是：main

main() {
    local files="/tmp/a /tmp/b"
    local i

    for i in $files
    do
        change_owner_of_file kfir users $i
    done
}
main

main ( ) {

local files = "/tmp/a /tmp/b"

local i

for i in $ files

change_owner_of_file kfir users $ i

done

}

main

一切皆是函数

唯一全局性运行的代码是：

- 不可变的全局变量声明

- main()函数

保持代码整洁
过程变得清晰

main() {
    local files=$(ls /tmp | grep pid | grep -v daemon)
}

main ( ) {

local files = $ ( ls / tmp | grep pid | grep - v daemon )

}

temporary_files() {
    local dir=$1

    ls $dir \
        | grep pid \
        | grep -v daemon
}

main() {
    local files=$(temporary_files /tmp)
}

temporary_files ( ) {

local dir = $ 1

ls $ dir \

| grep pid \

| grep - v daemon

}

main ( ) {

local files = $ ( temporary_files / tmp )

}

第二个例子好得多。查找文件是temporary_files()的问题而非main()的。这段代码用temporary_files()的单元测试也是可测试的。
如果你一定要尝试第一个例子，你会得到查找临时文件以和main算法的大杂烩。

test_temporary_files() {
    local dir=/tmp

    touch $dir/a-pid1232.tmp
    touch $dir/a-pid1232-daemon.tmp

    returns "$dir/a-pid1232.tmp" temporary_files $dir

    touch $dir/b-pid1534.tmp

    returns "$dir/a-pid1232.tmp $dir/b-pid1534.tmp" temporary_files $dir
}

test_temporary_files ( ) {

local dir = / tmp

touch $ dir / a - pid1232 . tmp

touch $ dir / a - pid1232 - daemon . tmp

returns "$dir/a-pid1232.tmp" temporary _files $ dir

touch $ dir / b - pid1534 . tmp

returns "$dir/a-pid1232.tmp $dir/b-pid1534.tmp" temporary _files $ dir

}

如你所见，这个测试不关心main()。

调试函数

带-x标志运行程序：

bash -x my_prog.sh

1	bash - x my_prog . sh

只调试一小段代码，使用set-x和set+x，会只对被set -x和set +x包含的当前代码打印调试信息。

temporary_files() {
    local dir=$1

    set -x
    ls $dir \
        | grep pid \
        | grep -v daemon
    set +x
}

temporary_files ( ) {

local dir = $ 1

set - x

ls $ dir \

| grep pid \

| grep - v daemon

set + x

}

打印函数名和它的参数：

temporary_files() {
    echo $FUNCNAME $@
    local dir=$1

    ls $dir \
        | grep pid \
        | grep -v daemon
}

temporary_files ( ) {

echo $ FUNCNAME $ @

local dir = $ 1

ls $ dir \

| grep pid \

| grep - v daemon

}

调用函数：

temporary_files /tmp

1	temporary_files / tmp

会打印到标准输出：

temporary_files /tmp

1	temporary_files / tmp

代码的清晰度

这段代码做了什么？

main() {
    local dir=/tmp

    [[ -z $dir ]] \
        && do_something...

    [[ -n $dir ]] \
        && do_something...

    [[ -f $dir ]] \
        && do_something...

    [[ -d $dir ]] \
        && do_something...
}
main

main ( ) {

local dir = / tmp

[ [ - z $ dir ] ] \

&& do_something . . .

[ [ - n $ dir ] ] \

&& do_something . . .

[ [ - f $ dir ] ] \

&& do_something . . .

[ [ - d $ dir ] ] \

&& do_something . . .

}

main

让你的代码说话：

is_empty() {
    local var=$1

    [[ -z $var ]]
}

is_not_empty() {
    local var=$1

    [[ -n $var ]]
}

is_file() {
    local file=$1

    [[ -f $file ]]
}

is_dir() {
    local dir=$1

    [[ -d $dir ]]
}

main() {
    local dir=/tmp

    is_empty $dir \
        && do_something...

    is_not_empty $dir \
        && do_something...

    is_file $dir \
        && do_something...

    is_dir $dir \
        && do_something...
}
main

is_empty ( ) {

local var = $ 1

[ [ - z $ var ] ]

}

is_not_empty ( ) {

local var = $ 1

[ [ - n $ var ] ]

}

is_file ( ) {

local file = $ 1

[ [ - f $ file ] ]

}

is_dir ( ) {

local dir = $ 1

[ [ - d $ dir ] ]

}

main ( ) {

local dir = / tmp

is _empty $ dir \

&& do_something . . .

is_not _empty $ dir \

&& do_something . . .

is _file $ dir \

&& do_something . . .

is _dir $ dir \

&& do_something . . .

}

main

每一行只做一件事

用反斜杠\来作分隔符。例如：

temporary_files() {
    local dir=$1

    ls $dir | grep pid | grep -v daemon
}

temporary_files ( ) {

local dir = $ 1

ls $ dir | grep pid | grep - v daemon

}

可以写得简洁得多：

temporary_files() {
    local dir=$1

    ls $dir \
        | grep pid \
        | grep -v daemon
}

temporary_files ( ) {

local dir = $ 1

ls $ dir \

| grep pid \

| grep - v daemon

}

符号在缩进行的开始

符号在行末的坏例子：（译注：原文在此例中用了temporary_files()代码段，疑似是贴错了。结合上下文，应为print_dir_if_not_empty()）

print_dir_if_not_empty() {
    local dir=$1
 
    is_empty $dir && \
        echo "dir is empty" || \
        echo "dir=$dir"
}

print_dir_if_not_empty ( ) {

local dir = $ 1

is _empty $ dir && \

echo "dir is empty" || \

echo "dir=$dir"

}

好的例子：我们可以清晰看到行和连接符号之间的联系。

print_dir_if_not_empty() {
    local dir=$1

    is_empty $dir \
        && echo "dir is empty" \
        || echo "dir=$dir"
}

print_dir_if_not_empty ( ) {

local dir = $ 1

is _empty $ dir \

&& echo "dir is empty" \

|| echo "dir=$dir"

}

打印用法

不要这样做：

echo "this prog does:..."
echo "flags:"
echo "-h print help"

echo "this prog does:..."

echo "flags:"

echo "-h print help"

它应该是个函数：

usage() {
    echo "this prog does:..."
    echo "flags:"
    echo "-h print help"
}

usage ( ) {

echo "this prog does:..."

echo "flags:"

echo "-h print help"

}

echo在每一行重复。因此我们得到了这个文档：

usage() {
    cat <<- EOF
    usage: $PROGNAME options
    
    Program deletes files from filesystems to release space. 
    It gets config file that define fileystem paths to work on, and whitelist rules to 
    keep certain files.

    OPTIONS:
       -c --config              configuration file containing the rules. use --help-config to see the syntax.
       -n --pretend             do not really delete, just how what you are going to do.
       -t --test                run unit test to check the program
       -v --verbose             Verbose. You can specify more then one -v to have more verbose
       -x --debug               debug
       -h --help                show this help
          --help-config         configuration help

    
    Examples:
       Run all tests:
       $PROGNAME --test all

       Run specific test:
       $PROGNAME --test test_string.sh

       Run:
       $PROGNAME --config /path/to/config/$PROGNAME.conf

       Just show what you are going to do:
       $PROGNAME -vn -c /path/to/config/$PROGNAME.conf
    EOF
}

usage ( ) {

cat << - EOF

usage : $ PROGNAME options

Program deletes files from filesystems to release space .

It gets config file that define fileystem paths to work on , and whitelist rules to

keep certain files .

OPTIONS :

- c -- config configuration file containing the rules . use -- help - config to see the syntax .

- n -- pretend do not really delete , just how what you are going to do .

- t -- test run unit test to check the program

- v -- verbose Verbose . You can specify more then one - v to have more verbose

- x -- debug debug

- h -- help show this help

-- help - config configuration help

Examples :

Run all tests :

$ PROGNAME -- test all

Run specific test :

$ PROGNAME -- test test_string . sh

Run :

$ PROGNAME -- config / path / to / config / $ PROGNAME . conf

Just show what you are going to do :

$ PROGNAME - vn - c / path / to / config / $ PROGNAME . conf

EOF

}

注意在每一行的行首应该有一个真正的制表符‘\t’。

在vim里，如果你的tab是4个空格，你可以用这个替换命令：

:s/^    /\t/

1	: s / ^ / \ t /

命令行参数

这里是一个例子，完成了上面usage函数的用法。我从Kirk’s blog post – bash shell script to use getopts with gnu style long positional parameters得到这段代码

cmdline() {
    # got this idea from here:
    # http://kirk.webfinish.com/2009/10/bash-<a href="http://www.ttlsa.com/shell/" title="shell"target="_blank">shell</a>-script-to-use-getopts-with-gnu-style-long-positional-parameters/
    local arg=
    for arg
    do
        local delim=""
        case "$arg" in
            #translate --gnu-long-options to -g (short options)
            --config)         args="${args}-c ";;
            --pretend)        args="${args}-n ";;
            --test)           args="${args}-t ";;
            --help-config)    usage_config && exit 0;;
            --help)           args="${args}-h ";;
            --verbose)        args="${args}-v ";;
            --debug)          args="${args}-x ";;
            #pass through anything else
            *) [[ "${arg:0:1}" == "-" ]] || delim="\""
                args="${args}${delim}${arg}${delim} ";;
        esac
    done

    #Reset the positional parameters to the short options
    eval set -- $args

    while getopts "nvhxt:c:" OPTION
    do
         case $OPTION in
         v)
             readonly VERBOSE=1
             ;;
         h)
             usage
             exit 0
             ;;
         x)
             readonly DEBUG='-x'
             set -x
             ;;
         t)
             RUN_TESTS=$OPTARG
             verbose VINFO "Running tests"
             ;;
         c)
             readonly CONFIG_FILE=$OPTARG
             ;;
         n)
             readonly PRETEND=1
             ;;
        esac
    done

    if [[ $recursive_testing || -z $RUN_TESTS ]]; then
        [[ ! -f $CONFIG_FILE ]] \
            && eexit "You must provide --config file"
    fi
    return 0
}

cmdline ( ) {

# got this idea from here:

# http://kirk.webfinish.com/2009/10/bash-shell-script-to-use-getopts-with-gnu-style-long-positional-parameters/

local arg =

for arg

local delim = ""

case "$arg" in

#translate --gnu-long-options to -g (short options)

-- config ) args = "${args}-c " ; ;

-- pretend ) args = "${args}-n " ; ;

-- test ) args = "${args}-t " ; ;

-- help - config ) usage_config && exit 0 ; ;

-- help ) args = "${args}-h " ; ;

-- verbose ) args = "${args}-v " ; ;

-- debug ) args = "${args}-x " ; ;

#pass through anything else

* ) [ [ "${arg:0:1}" == "-" ] ] || delim = "\""

args = "${args}${delim}${arg}${delim} " ; ;

esac

done

#Reset the positional parameters to the short options

eval set -- $ args

while getopts "nvhxt:c:" OPTION

case $ OPTION in

v )

readonly VERBOSE = 1

; ;

h )

usage

exit 0

; ;

x )

readonly DEBUG = '-x'

set - x

; ;

t )

RUN_TESTS = $ OPTARG

verbose VINFO "Running tests"

; ;

c )

readonly CONFIG_FILE = $ OPTARG

; ;

n )

readonly PRETEND = 1

; ;

esac

done

if [ [ $ recursive_testing || - z $ RUN _TESTS ] ] ; then

[ [ ! - f $ CONFIG _FILE ] ] \

&& eexit "You must provide --config file"

return 0

}

你像这样，使用我们在头上定义的不可变的ARGS变量：

main() {
    cmdline $ARGS
}
main

main ( ) {

cmdline $ ARGS

}

main

单元测试

在更高级的语言中很重要。
使用shunit2做单元测试

test_config_line_paths() {
    local s='partition cpm-all, 80-90,'

    returns "/a" "config_line_paths '$s /a, '"
    returns "/a /b/c" "config_line_paths '$s /a:/b/c, '"
    returns "/a /b /c" "config_line_paths '$s   /a  :    /b : /c, '"
}

config_line_paths() {
    local partition_line="$@"

    echo $partition_line \
        | csv_column 3 \
        | delete_spaces \
        | column 1 \
        | colons_to_spaces
}

source /usr/bin/shunit2

test_config_line_paths ( ) {

local s = 'partition cpm-all, 80-90,'

returns "/a" "config_line_paths '$s /a, '"

returns "/a /b/c" "config_line_paths '$s /a:/b/c, '"

returns "/a /b /c" "config_line_paths '$s /a : /b : /c, '"

}

config_line_paths ( ) {

local partition_line = "$@"

echo $ partition _line \

| csv _column 3 \

| delete _spaces \

| column 1 \

| colons_to _spaces

}

source / usr / bin / shunit2

这里是另一个使用df命令的例子：

DF=df

mock_df_with_eols() {
    cat <<- EOF
    Filesystem           1K-blocks      Used Available Use% Mounted on
    /very/long/device/path
                         124628916  23063572 100299192  19% /
    EOF
}

test_disk_size() {
    returns 1000 "disk_size /dev/sda1"

    DF=mock_df_with_eols
    returns 124628916 "disk_size /very/long/device/path"
}

df_column() {
    local disk_device=$1
    local column=$2

    $DF $disk_device \
        | grep -v 'Use%' \
        | tr '\n' ' ' \
        | awk "{print \$$column}"
}

disk_size() {
    local disk_device=$1

    df_column $disk_device 2
}

DF = df

mock_df_with_eols ( ) {

cat << - EOF

Filesystem 1K - blocks Used Available Use % Mounted on

/ very / long / device / path

124628916 23063572 100299192 19 % /

EOF

}

test_disk_size ( ) {

returns 1000 "disk_size /dev/sda1"

DF = mock_df_with_eols

returns 124628916 "disk_size /very/long/device/path"

}

df_column ( ) {

local disk_device = $ 1

local column = $ 2

$ DF $ disk _device \

| grep - v 'Use%' \

| tr '\n' ' ' \

| awk "{print \$$column}"

}

disk_size ( ) {

local disk_device = $ 1

df _column $ disk _device 2

}

这里我有个例外，为了测试，我在全局域中声明了DF为非只读。这是因为shunit2不允许改变全局域函数。

原文：http://www.kfirlavi.com/blog/2012/11/14/defensive-bash-programming/

译文：http://blog.jobbole.com/73257/

转载请注明：成长的对话 » BASH的保护性编程技巧

不可改变的全局变量

一切皆是局部的

main()

一切皆是函数

调试函数

代码的清晰度

每一行只做一件事

打印用法

命令行参数

单元测试

与本文相关的文章