4

I have a huge text file, each line contains value or text pattern.

How can I find average or at least sum values between each pattern?

Are there any awk/sed or maybe perl solutions?

Example:

pattern1
200
300
pattern2
200
100
200
pattern3
pattern4

Expected output:

pattern1
250
pattern2
166
pattern3
pattern4
Improve this question
3
  • 2
    awk and perl would be good languages to use to solve this. Not sed, as it doesn't have any facility for arithmetic. What attempts have you made to solve this, and where are you stuck? Commented Oct 30, 2019 at 14:59
  • Will you have values after the last pattern? Commented Oct 30, 2019 at 15:09
  • No values after last pattern Commented Oct 30, 2019 at 15:11

2 Answers 2

Reset to default
4

Awk approach:

awk '/pattern/{ if (cnt) { printf "%d\n", sum/cnt; sum=cnt=0 } print }
     /^[0-9]+$/{ sum += $1; cnt++ }' file

The output:

pattern1
250
pattern2
166
pattern3
pattern4
Improve this answer
8
  • 1
    If there are numbers after the last pattern, you'll need to do something in an END block. Commented Oct 30, 2019 at 15:03
  • @glennjackman, they won't show up anyway, will be ignored in that case Commented Oct 30, 2019 at 15:04
  • Many thanks to @RomanPerekhrest , it works! I just only add /pattern1|pattern2/... because have several quite different patterns Commented Oct 30, 2019 at 15:09
  • @干猕猴桃, how those patterns are actually look like? Commented Oct 30, 2019 at 15:10
  • 1
    It might actually make more sense to test for all numeric lines and next there given the OP says in a comment I just only add /pattern1|pattern2/... because have several quite different patterns to avoid all those different "pattenrs" (ugh!) BUT only the OP will know for sure... Commented Oct 31, 2019 at 12:18
0

Using the Gnu tools sed n dc we can do this:

$ sed -re '
    /^[0-9]+$/,/pattern/!b
    /^[0-9]+$/{H;d;}
    x;y/\n/ /
    s#.*#dc -e "0dsn&[+ln1+snz1<+]s+l+xln/p"#ep
    z;x
' file

Results:

pattern1
250
pattern2
166
pattern3
pattern4

Assuming the trailing line cannot be a number. Also assuming no negative nums.

We setup a range in sed to look for lines comprising only digits to pattern. Outside this all are printed as is. The numbers are appended in hold nd then on tge the pattern line the average is computed using the s///e command of gnu sed utilizing the dc utility.

Improve this answer

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.