我如何使用sed或awk来更正日志时间格式

Question

I am currently working on a script which processes a csv file, and corrects certain aspects of them along the way. One of the things that it does is correct time format if needed. Two type of conversion takes place:

 xx:xx:xx to PTxxHxxMxxS
 10:03:45 to PT10H03M45S

I've have been able to do this using the following (see below) although I am trying to find out how to do it either using sed or awk in order to speed up the process. In addition to the actual conversion process, I would also like to keep count of the changes that are made (so say 4 times values are converted, a counter would be incremented to 4), which I have been able to do easily with the if statement below (though it is not shown), although I would not know much about doing that using sed/awk.

 istimef=$( echo "$Sfcpp6" | grep ".*:.*:.*" )
                    if [ "$istimef" != "" ]; then
                            hs=$( echo "$Sfcpp6" | cut -d ':' -f 1 )
                            mn=$( echo "$Sfcpp6" | cut -d ':' -f 2 )
                            sc=$( echo "$Sfcpp6" | cut -d ':' -f 3 )
                            Sfcpp6=$( echo "PT"$hs"H"$mn"M"$sc"S" )
                            echo "$Sfcp6"
                    fi

which essentially checks if the time value is even there, and then performs the conversion.

Answer 1

It amazing how much processes and subshells you need for this task! I'll always be amazed at people's ingenuity and creativity. I counted 10 subshells, and 4 process spawns.

Look, you can achieve exactly the same without spawning one process and with no subshell whatsoever. We're talking about speed-up here!

First task, given a string in the form xx:yy:zz , transform it into PTxxHyyMzzS as efficiently as possible (look, in only one command! and a builtin! no sed !):

$ string='12:34:56'
$ printf -v transformed 'PT%sH%sM%sS' ${string//:/ }
$ # Done! Don't believe me?
$ echo "$transformed"
PT12H34M56S

Now, before doing this, you probably want to check if the string is of the form xx:yy:zz . Quit grep for that. Just test it thus:

if [[ "$string" = *:*:* ]]; then
    echo "ok"
else
    echo "not ok"
fi

So the part of your script you showed us would be much more efficient thus:

if [[ "$Sfcpp6" = *:*:* ]]; then
    printf -v Sfcp6 'PT%sH%sM%sS' ${Sfcpp6//:/ }
    echo "$Sfcp6"
fi

Total: 0 subshells, 0 processes spawned.

Or if your goal is only to echo the transformed string:

if [[ "$Sfcpp6" = *:*:* ]]; then
    printf 'PT%sH%sM%sS\n' ${Sfcpp6//:/ }
fi

Answer 2

sed解决方案：使用\\(...\\)来捕获数字，字符类[0-9]以匹配任何数字。

sed 's/\([0-9][0-9]\):\([0-9][0-9]\):\([0-9][0-9]\)/PT\1H\2M\3S/'

Answer 3

If you want to count the substituted lines :

perl -pe '
    END{print "count=$count\n"}
    s/(\d{2}):(\d{2}):(\d{2})/PT$1H$2M$3S/ && $count++
' file.txt

Answer 4

The GNU awk equivalent of this sed solution posted by @choroba:

sed 's/\([0-9][0-9]\):\([0-9][0-9]\):\([0-9][0-9]\)/PT\1H\2M\3S/'

would be the very similar:

awk '{print gensub(/([0-9][0-9]):([0-9][0-9]):([0-9][0-9])/,"PT\\1H\\2M\\3S","")}'

but the awk solution can be trivially modified to address your question of "would it be possible to make sed keep count of the changes that It has made?":

awk '{orig=$0; $0=gensub(/([0-9][0-9]):([0-9][0-9]):([0-9][0-9])/,"PT\\1H\\2M\\3S",""); print} $0 != orig{count++} END{printf "%d changes made.\n",count}'

while the sed solution can't.