I have a file whose columns contain simple arithmetic equations that I would like to merge to the arithmetic result.

Input sample (tab-separated columns):

+104-1+12   6   +3

I would like to compute the arithmetic sum within each column. If one column contains no arithmetic sign, I treat it as it contained a + before the item. Although it would be easy through sed to add a + sign if a column starts with no sign (sed -E 's/(\t)([0-9]*)/\1\t+\2/g' would work, assuming that a row never begins with a digit, as in the example)

The output I would expect is the following:

115 6   3

How can I achieve this in unix? awk/sed solutions are preferred.

  • i made edits in the question text, including just the first line of the given sample. the tab format was not quite right in the code chunk, sorry for that :/ – dovah May 9 '17 at 12:51

You could use perl:

perl -pe 's/[\d+-]+/eval$&/ge' your-file

Or even:

perl -pe 's/[\d+-]+/$&/gee' your-file (thanks Rakesh)

Same with zsh:

set -o extendedglob # for the ## operator (same as ERE +)
while IFS= read -r line; do 
  printf '%s\n' ${line//(#m)[0-9+-]##/$((MATCH))}
done < your-file


zmodload zsh/mapfile
set -o extendedglob
printf %s ${mapfile[your-file]//(#m)[0-9+-]##/$((MATCH))}

In all four, we're looking for sequences of digits, - and + characters and passing them to the interpreter's arithmetic processor (eval in perl (or the ee flag that causes the expansion of the replacement to be evaluated as perl code), $((...)) in zsh).

We're not validating the expressions before passing to the interpreter, so it may cause failures (for instance on sequences like -+- or 3++) but at least, because we're only considering digits and -/+ characters, it shouldn't do much more harm than reporting an error message and aborting the command.

|improve this answer|||||
  • set -o extendedglob requires what bash version? – user218374 May 9 '17 at 13:38
  • @RakeshSharma, as stated, that's zsh code, not bash. If run from bash, wrap the code in zsh -c 'the-code' (though it would make more sense to switch to zsh for the whole script). – Stéphane Chazelas May 9 '17 at 14:08

I won't duplicate the Addition with 'sed' answer; nor did I find a way in awk, but here's a bash version:

while IFS= read -r line
  set -f; set -- $line
  for e in "$@"
    printf "%d " "$(( e ))"
done < input
|improve this answer|||||
  • 2
    Note that it's an arbitrary command injection vulnerability if the content of the file is not tightly controlled. – Stéphane Chazelas May 9 '17 at 12:52
  • @StéphaneChazelas wait... where is the vulnerability here? I tried putting 1+$(echo 1) in the input but bash gives me syntax error: operand expected. – JoL May 9 '17 at 16:04
  • 2
    @jlmg, try putting a[`uname>&2`0] for instance. – Stéphane Chazelas May 9 '17 at 16:06
sed -E 's/(\t)([0-9])/\1+\2/g' data.file |
while IFS= read -r l; do
   set -f; IFS=$'\t'
   printf '0%s\n' $l | bc -l | paste -s -

sed -e 's/\t\([0-9]\)/\t+\1/' data.file |
while IFS= read -r l; do
   set -f; IFS=$'\t'
   printf '0%s\n' $l | bc -c |
   sed -ne '
         y/:@irKW/      /
         s/[^ 0-9]/ & /g
         s/[ ][ ]*/ /g;s/^[ ]*//;s/[ ]*$/p/p
   ' | dc | paste -s -

Here we generate a postfix representation of the math expression and before passing it onto the postfix calculator dc, we clean up the non-math info from the output of the bc -c command.


115     6       3
|improve this answer|||||

Using awk getline from a pipe

awk '{
  for (i=1;i<=NF;i++) {
    cmd = sprintf("echo %s | bc -l", $i); 
    cmd | getline $i; close(cmd);
} 1' file
115 6 3
25 6 2 69 57
|improve this answer|||||
  • 1
    It also amounts to a command injection vulnerability if the content of the file is not tightly controlled. Also note that it runs one sh and one bc command per field. – Stéphane Chazelas May 9 '17 at 13:09

Here's an all-awk solution taking advantage of awk's ability to marshall string representations of numbers into numeric representations, with no use of external executables:

awk -F"\t" \
'BEGIN { OFS="\t" }
 { gsub(/-/,"|-") 
   for(i=1; i<=NF; i++) { ## iterate over columns
     for(j=1; j<=num_parts; j++) ## iterate over arithmetic expression parts
       sums[i] += parts[j]+0 ## Adding zero marshals the string into a numeric
      for(i=1; i<=NF; i++) { 
        if(i>1) printf OFS
        printf sums[i]
      print "" }' file
|improve this answer|||||

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.