BASH的保护性编程技巧

main()

temporary_files() {      local dir=$1        set -x      ls $dir \          | grep pid \          | grep -v daemon      set +x  }

temporary_files() {      echo $FUNCNAME $@      local dir=$1        ls $dir \          | grep pid \          | grep -v daemon  }

调用函数：

temporary_files /tmp

会打印到标准输出：

temporary_files /tmp

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

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

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  }

可以写得简洁得多：

temporary_files() {      local dir=$1        ls $dir \          | grep pid \          | grep -v daemon  }

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"  }

打印用法

不要这样做：

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"  }

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  }

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/

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-shell-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  }

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

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

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  }

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。