将长单行命令转换为bash shell脚本

Question

Im working on the following command in crontab:

nice --20 iperf3 -c somelocation.com -f k | while IFS= read -r line; do echo "$(date) $line"; done | tee onespeed.txt | tee -a speeds.txt; sleep 30 ;cat onespeed.txt | grep sender >> concentrated.txt; sleep 2 ;cat onespeed.txt | grep sender | awk -F' ' '{print $13}' | while IFS= read -r line; do echo "$(date) $line"; done | tee onerawspeed.txt | tee -a rawspeeds.txt

I understand that the job could be done in a far more elegant way but Im looking to find out how I go about breaking this down so that its structured properly to act as a bash shell script.

Specifically Im focused on structuring it over multiple lines so that it is more readable.

Answer 1

Since it's not clear what you really want to put in all those output files, I'll give an example of how to write separate values to different files.

The bitrate p/sec per interval and the average Bitrate that iperf3 produces can be put -separely- into a file like this:

$ cat tst.awk
BEGIN {OFS="\t";f1="speed.txt";f2="sender.txt"}
$6~/MBytes/{print strftime(), $7 >> f1; close(f1)}
/sender/{print strftime(), $7 >> f2; close(f2)}

which can be called like this:

$ iperf3 -c server -f k --forceflush | awk -f tst.awk

or, in one line:

 iperf3 -c server -f k --forceflush | awk -v f1=speed.txt -v f2=sender.txt \
    'BEGIN{OFS="\t"}                                          \
     $6~/MBytes/{print strftime(), $7 >> f1; close(f1)}       \
     /sender/{print strftime(), $7 >> f2; close(f2)}'

which print sysdate and Bitrate to each file, split by \\t. You'll need the forceflush option of iperf3 to prevent buffering of the output, which would cause the awk script to print the same timestamp for each line.

Answer 2

Wrap the code in an arbitrary function n() { one-liner_code_here ; } n() { one-liner_code_here ; } , and run type n . bash will flesh a lot of it out -- particularly useful is how it'll indent loops. But type won't break up the pipes, and leaves semicolons behind, so here's some code to automate the whole business...

Suppose the one-liner is a file src . The following puts src into $x , then makes a function n for it, (this requires an eval which for a quoted string in a function should be harmless), and type s it, feeds that to sed which ignores the surrounding brackets, breaks up the longer pipes and removes the leading indents & trailing ; s, then unset s n :

read x <src; eval n\(\) \{ "$x" \;\}
type n | sed -n '/^ /{s/^    //;/^[^ ]/s/| /|\n/g;s/\;$//;p}' ; unset -f n

Output:

nice --20 iperf3 -c somelocation.com -f k |
while IFS= read -r line; do
    echo "$(date) $line"
done |
tee onespeed.txt |
tee -a speeds.txt
sleep 30
cat onespeed.txt |
grep sender >> concentrated.txt
sleep 2
cat onespeed.txt |
grep sender |
awk -F' ' '{print $13}' |
while IFS= read -r line; do
    echo "$(date) $line"
done |
tee onerawspeed.txt |
tee -a rawspeeds.txt

---

Note: type will also convert any alias in the current environment to code. So if grep is alias ed to grep --color=auto , each grep above would be replaced with grep --color=auto .

Answer 3

nice --20 iperf3 -c somelocation.com -f k | while IFS= read -r line; do 
echo "$(date) $line" 
done | tee onespeed.txt | tee -a speeds.txt
sleep 30 
grep "sender" onespeed.txt >> concentrated.txt
sleep 2 
awk -F' ' '/sender/ {print $13}' onespeed.txt | while IFS= read -r line; do
echo "$(date) $line"
done | tee onerawspeed.txt | tee -a rawspeeds.txt

EDIT : I have updated the script, following the comments advices