ksh - Unix 脚本 - 需要提高性能的建议(shell 脚本)

Question

我有一个输入 csv 文件，实际上我需要在输入文件中选择第 2 列和第 3 列值，并且需要转换两个值的时区(从 PT 到 CT)，转换后我需要替换转换后的时区值到文件。

注意: 所有输入日期值都在太平洋时区，我正在转换为中央时区。

每行有 5 列 - 逗号分隔的文件

CHID-123456323,2017-01-09 17:17:58-08:00,2017-01-09 17:39:25-08:00,hello,123456733
CHID-123456733,2017-01-09 17:16:58-08:00,2017-01-09 18:04:09-08:00,hello,123456734
CHID-123433589,2017-01-09 17:16:55-08:00,2017-01-09 17:40:29-08:00,hello,123456735
CHID-123000789,2017-01-09 17:16:52-08:00,2017-01-09 17:46:41-08:00,hello,123456736

脚本: 我写了一个下面的脚本，这给出了我期望的确切结果。但是当输入记录数增加时，它需要更多的时间。例如，2 万条记录需要 1 小时 15 分钟。

任何人都可以看看这个脚本并建议如何提高性能？

脚本:

while read i
do
    var1=`echo $i | awk -F',' '{ print $2 }'`

    var1_EPOCH=`date --date="${var1}" +%s`
    var1_CTZ=`TZ=":America/Chicago" date +"%Y-%m-%d %T" -d@$var1_EPOCH`
    sed -i "${cnt}s/${var1}/${var1_CTZ}/" filename

    var2=`echo $i | awk -F',' '{ print $3 }'`
    var2_EPOCH=`date --date="${var2}" +%s`
    var2_CTZ=`TZ=":America/Chicago" date +"%Y-%m-%d %T" -d@$var2_EPOCH`
    sed  -i "${cnt}s/${var2}/${var2_CTZ}/" filename

    cnt=$(($cnt+1))
done < filename

这里是预期的输出文件

最终输出文件:

CHID-123456323,2017-01-09 19:17:58,2017-01-09 19:39:25,hello,123456733
CHID-123456733,2017-01-09 19:16:58,2017-01-09 20:04:09,hello,123456734
CHID-123433589,2017-01-09 19:16:55,2017-01-09 19:40:29,hello,123456735
CHID-123000789,2017-01-09 19:16:52,2017-01-09 19:46:41,hello,123456736 

Answer 1

Ksh 为您提供了足够的内置功能。

示例输入文件:

[STEP 100] $ echo $BASH_VERSION
4.4.5(2)-release
[STEP 101] $ cat file
CHID-123456323,2017-01-09 17:17:58-08:00,2017-01-09 17:39:25-08:00,hello,123456733
CHID-123456733,2017-01-09 17:16:58-08:00,2017-01-09 18:04:09-08:00,hello,123456734
CHID-123433589,2017-01-09 17:16:55-08:00,2017-01-09 17:40:29-08:00,hello,123456735
CHID-123000789,2017-01-09 17:16:52-08:00,2017-01-09 17:46:41-08:00,hello,123456736

剧本:

[STEP 102] $ cat time.ksh
tz=America/Chicago
pattern='(.+),(.+),(.+),(.+),(.+)'
while read -r line; do
    if [[ $line =~ $pattern ]]; then
        c1=${.sh.match[1]}
        c2=${.sh.match[2]}
        c3=${.sh.match[3]}
        c4=${.sh.match[4]}
        c5=${.sh.match[5]}

        TZ=$tz printf '%(%Y-%m-%d %T)T' "$c2" | read c2
        TZ=$tz printf '%(%Y-%m-%d %T)T' "$c3" | read c3

        print -r -- "$c1,$c2,$c3,$c4,$c5"
    else
        print -r -- "$line"
    fi
done

示例输出:

[STEP 103] $ ksh time.ksh < file
CHID-123456323,2017-01-09 19:17:58,2017-01-09 19:39:25,hello,123456733
CHID-123456733,2017-01-09 19:16:58,2017-01-09 20:04:09,hello,123456734
CHID-123433589,2017-01-09 19:16:55,2017-01-09 19:40:29,hello,123456735
CHID-123000789,2017-01-09 19:16:52,2017-01-09 19:46:41,hello,123456736

制作一个 20,000 行的文件:

[STEP 104] $ rm -f bigfile
[STEP 105] $ fourlines=$(<file)
[STEP 106] $ for ((i=0; i<5000; ++i)); do printf '%s\n' "$fourlines" >> bigfile; done
[STEP 107] $ wc -l bigfile
   20000 bigfile

让我们对其进行性能测试:

[STEP 108] $ time ksh time.ksh < bigfile > newfile

real    1m36.849s
user    0m27.376s
sys     0m46.741s
[STEP 109] $ tail -n 4 newfile
CHID-123456323,2017-01-09 19:17:58,2017-01-09 19:39:25,hello,123456733
CHID-123456733,2017-01-09 19:16:58,2017-01-09 20:04:09,hello,123456734
CHID-123433589,2017-01-09 19:16:55,2017-01-09 19:40:29,hello,123456735
CHID-123000789,2017-01-09 19:16:52,2017-01-09 19:46:41,hello,123456736
[STEP 110] $ ksh --version
  version         sh (AT&T Research) 93u+ 2012-08-01
[STEP 111] $