Making sure that line X is printed before line Y in a text file?

CoreOxide · 08-21-2014, 07:31 AM

I did some testing today with an actual log file, had some free time.

Quote:

Originally Posted by grail

Code:

sed -n '/^DistributedPOManager/,/^DistributedPOManager/{/^ControlAction/p}' file

Does not return anything, for some reason...

Quote:

Originally Posted by danielbmartin

Try this ...

Code:

result=$(grep '^DistributedPOManager\|^ControlAction' $InFile \
|cut -c1       \
|paste -sd"\0" \
|grep "DCD"    \
|wc -l)
echo "result=" $result

Daniel B. Martin

This works, but I had to trim the date and time at the beginning of each line.

Thanks!

grail · 08-21-2014, 09:30 AM

The sed assumes all the data is at the start of the line and spelled exactly as you see it, including case, otherwise it will not display anything.

So if your original file has information different to the examples, you either need to provide another example or tweak what has been supplied to match

pan64 · 08-22-2014, 11:22 AM

Quote:

Originally Posted by CoreOxide

This works, but I had to trim the date and time at the beginning of each line.
Thanks!

As I wrote several times avoid writing such chains, like grep|cut|paste|grep|wc, usually it can be implemented using a single grep/sed/awk/perl/whatever.
Especially the last two steps:
... | grep "DCD" | wc -l is equal to ... | grep -c "DCD"
If you still need to trim some data just go ahead and rework it ... (or at least give us some more sample - as it was already mentioned).

From the other hand I would rather go with something like a state-machine:

Code:

awk ' BEGIN { state = 0; count = 0 }
/first line regex/ { if state == 0 state=1 else error }
/second line regex/ { if state == 1 state=2 else error }
/third regex/ { if state == 2 state=3 else error }
state == 3 { count++; state=0 }
END { print count }
'

(just an idea, not tested and also can be refined ....)

CoreOxide · 08-26-2014, 06:30 AM

Quote:

Originally Posted by grail

The sed assumes all the data is at the start of the line and spelled exactly as you see it, including case, otherwise it will not display anything.

So if your original file has information different to the examples, you either need to provide another example or tweak what has been supplied to match

Yeah, my logs have date and time print in the beginning of each line, so I had to trim it. I still need to do some testing on actual logs, but I think it will work

Quote:

Originally Posted by pan64

As I wrote several times avoid writing such chains, like grep|cut|paste|grep|wc, usually it can be implemented using a single grep/sed/awk/perl/whatever.
Especially the last two steps:
... | grep "DCD" | wc -l is equal to ... | grep -c "DCD"
If you still need to trim some data just go ahead and rework it ... (or at least give us some more sample - as it was already mentioned).

From the other hand I would rather go with something like a state-machine:

Code:

awk ' BEGIN { state = 0; count = 0 }
/first line regex/ { if state == 0 state=1 else error }
/second line regex/ { if state == 1 state=2 else error }
/third regex/ { if state == 2 state=3 else error }
state == 3 { count++; state=0 }
END { print count }
'

(just an idea, not tested and also can be refined ....)

Where in the awk command do I supply the input file?

NevemTeve · 08-26-2014, 06:50 AM

like with any other filter:

Code:

awk 'program' inputfile >outputfile