55

The file data.txt contains the following:

1.00 1.23 54.4 213.2 3.4

The output of the scripts are supposed to be:

ave: 54.646

Some simple scripts are preferred.

Chris Seymour
  • 83,387
  • 30
  • 160
  • 202
JackWM
  • 10,085
  • 22
  • 65
  • 92

4 Answers4

93

Here is one method:

$ awk '{s+=$1}END{print "ave:",s/NR}' RS=" " file
ave: 54.646
Chris Seymour
  • 83,387
  • 30
  • 160
  • 202
  • 5
    That will produce a division by zero error on an empty file. You need something like `awk '{s+=$1}END{print "ave:",(NR?s/NR:"NaN")}' RS=" " file` – Ed Morton Jun 07 '16 at 12:11
  • 1
    As someone new to awk, I run this on this list: 8.20 7.10 8.10 7.40 6.50 8.40 7.90 8.50 9.30 8.80 9.80 9.50 8.20 9.20 8.30 9.10 and get an average of 8.2. If I sum the values, awk '{s+=$1}END{print "sum:" s}' I get 134.3. Divide this by the number of values and you get 8.39... which matches calculating it by spreadsheet. Does awk reduce the number of significant figures? – Montag Mar 16 '23 at 02:18
50

Another option is to use jq:

$ seq 100|jq -s add/length
50.5

-s (--slurp) creates an array for the input lines after parsing each line as JSON, or as a number in this case.

Edit: awk is faster and it doesn't require reading the whole input to memory:

$ time seq 1e6|awk '{x+=$0}END{print x/NR}'>/dev/null

real  0m0.145s
user  0m0.148s
sys   0m0.008s
$ time seq 1e6|jq -s add/length>/dev/null

real  0m0.685s
user  0m0.669s
sys   0m0.024s
nisetama
  • 7,764
  • 1
  • 34
  • 21
  • 6
    probably one of the most creative ways to use `jq` i've seen! – morganbaz Apr 14 '19 at 16:50
  • this has the advantage of being able to be used after having used jq to filter out the necessary numbers, like: ` jq -r '.SpotPriceHistory[].SpotPrice' | jq -s add/length` – dimisjim Oct 19 '20 at 15:15
  • FYI if you're averaging a lot of numbers. This is definitely way slower than the awk method. I can't be bothered to benchmark it, but for 14 million integers, `awk` returns nearly instantly while the `jq` method takes awhile... – WattsInABox Oct 18 '21 at 14:51
  • Yeah although this is super neat and although it's not *that* slow, this is indeed pretty slow, only suitable for processing very small datasets. – Steven Lu Nov 02 '21 at 22:59
  • @nisetama I had a billion lines or so to work with, other solutions finished, jq's didn't because I gave up on it :) – WattsInABox Nov 08 '21 at 16:49
  • https://stedolan.github.io/jq/manual/#Advancedfeatures – Chris Smith Oct 13 '22 at 21:40
6
perl -lane '$a+=$_ for(@F);print "ave: ".$a/scalar(@F)' file

if you have multiple lines and you just need a single average:

perl -lane '$a+=$_ for(@F);$f+=scalar(@F);END{print "ave: ".$a/$f}' file
Vijay
  • 65,327
  • 90
  • 227
  • 319
0

I had to find the average (time took for each ping request)

ping -c 5 www.youtube.com | grep time= | awk -F '[= ]' '{ sum += $11 } END { printf("%.2f ms\n", sum/NR) }'

  • -c in ping request is for numbers of icmp requests to send.

  • -F in grep used to specify the field separators.

  • We can use printf to control number of precision we want to display after the decimal points in our case it is 2 decimal points(%.2f).

  • “NR” is a special built-in variable of AWK that stands for "number of records". This variable is used to deal with the number of records present in the specified files.

  • We are calculating average by diving sum with NR

You can also find the sum (sum of all time took by the ping request)

ping -c 5 www.youtube.com | grep time= | awk -F '[= ]' '{ sum += $11 } END { printf("%.2f ms\n", sum) }'
Udesh
  • 2,415
  • 2
  • 22
  • 32