Ndtyjky

I would like to

(i) replace blank space between characters only if these characters are single; i.e. for instance

Down [Enter] p s -- a u x [Delete] 

should become

Down [Enter] ps -- aux [Delete] 

(ii) remove words that are consecutively repeated more than X times until any other thing which is not the word, so that (say X=2)

 [Delete] [Delete] [Delete] [Delete] [Delete] [Delete] ab inition [Delete] [Delete] [Delete] [Delete] [Delete] [Delete] ab definitio

becomes

 [Delete] [Delete] ab initio [Delete] [Delete] ab definitio

thanks!

You did not get a lot of responses. I think the main reason is the combination of two different questions, both non-trivial. Normally it helps to show your own effort, but I anderstand your effort might have been thinkong for hours "where to start".

The first question, removing spaces between single characters, can be done with a loop in sed:

echo 'Down [Enter] p s -- a u x [Delete] ' | 

   sed -r ':a;s/( [^ ]|r) ([^ ])( |$)/12r3/;ta; s/r//g'

Down [Enter] ps -- aux [Delete]

Explanation:
With a direct approach a u x will be changed into au x after the first replace, and the other space will be forgotten. You need to go over the replacements more than once and remember that the letter u in au x was a singleton in the original string.

For remembering the places where a replacement has been done, we use a r (and remove it later).

:a; Label to return for the next replacement.
( [^ ]|r) A space followed by a letter OR our temporary r marker
([^ ]) A space followed by a letter
( |$) A space or end-of-line
/12r3/ Replace with the two remembered characters, insert a special marker and a space when it was not the last charater of the line.
ta Go back to the start-of-loop tag :a when something was replaced
s/r//g' Remove our temporary markers.

The second question is difficult too. The next solution is close but incorrect:

for (( X=2; X<8; X++)); do

  echo "X=$X (incorrect solution)"

  echo 'some some some some some some some some some some some input' |

     sed -r 's/([^ ]+[ ]+)(1{'${X}'})(1+)/2/g'

done

The problem is when the repeated string also appears on another place, as in
some some some input some some some or worse some some some input input input.

I do not see an easy fix for the sed solution, but awk will help here.

For counting repeated fields my solution is considering each word as one record.

for (( X=2; X<8; X++)); do

   echo "X=$X"

   echo 'some some some some some some some some some some some input some some some some' |

      awk -v x=$X 'BEGIN {RS="[ n]"; ORS='n'; repeated=1}

         { if (last==$0)

             repeated++;

           else

             repeated=1;

         }

         {last=$0}

         repeated <= x {print $0" "}

         END {print "n"}

      '

done