Why is xkvbd eating newlines when I it is fed from xsel?

Question

I have this script that I assign t oa keyboard shortcut to simulate pasting through middle click:

#!/bin/bash
aa=0
for randstring in `xsel`
do
  if [[ "$randstring" =~ [ěščřžýáíéúůóťďň] ]]
  then
      xxx=`xsel|sed 's/ě/\\\[ecaron]/g' |sed 's/š/\\\[scaron]/g' |sed 's/č/\\\[ccaron] g' |sed 's/ř/\\\[rcaron]/g' |sed 's/ž/\\\[zcaron]/g' |sed 's/ý/\\\[yacute]/g' |sed 's/á/\\\[aacute]/g' |sed 's/í/\\\[iacute]/g' |sed 's/é/\\\[eacute]/g' |sed 's/ú/\\\[uacute]/g' |sed 's/ů/\\\[uring]/g'  |sed 's/ó/\\\[oacute]/g' |sed 's/ď/\\\[dcaron]/g' |sed 's/ň/\\\[ncaron]/g' |sed 's/ť/\\\[tcaron]/g'  |sed ':a;N;$!ba;s/\n/\\n/g'`
      xvkbd -text "$xxx" 2>/dev/null
      aa=1
      break
  else
    aa=0
  fi
done
if [[ $aa -eq 0 ]]
    then
        xsel | xvkbd  -file - 2>/dev/null
fi

I use -text with xvkbd when the text is in Czech (my language) because xvkbd does not understand diacritics like ě but only in form like \[ecaron]. Now, with this option, if there is a newline int xsel, it does not get printed with xvkbd. However, when I do

xx="---8<-----\nToday date is: $(date +%Y%m%d)\n---8<-----" 
xvkbd -text "$xx"  2>/dev/null

Newlines do get printed.

I suspect the trouble is in the last sed expression sed ':a;N;$!ba;s/\n/\\n/g', but I do not know how to make it better. I think that I need to take care for \ns somehow?

Answer 1

EDIT: I found the problem, but the rest may still be worth reading for when you actually do want a trailing newline `...

You are missing an extra \\ or even just \

  sed ':a;N;$!ba;s/\n/\\\\n/g'`  

MORE INFO: Regarding the issue of $( ) vs backquotes for command substitution, mentioned a comment I put to this question, here is an excerpt from man bash.

When  the  old-style  backquote form of substitution is used, backslash
retains its literal meaning except when followed by $, `,  or  \.   The
first backquote not preceded by a backslash terminates the command sub‐
stitution.  When using the $(command) form, all characters between  the
parentheses make up the command; none are treated specially.

---

-- Original Post --

It won't print the last \n but it adds intermediate \n into $xx ...

Here is a simplified version of your last sed call:

printf '%s\n' 'a\[ecaron]' b c '\[rcaron]d' |
   sed ':a;N;$!ba;s/\n/\\n/g' 

Output:

a\[ecaron]\nb\nc\n\[rcaron]d

The reason it doesn't have a final \n (in the above example) is that there simply won't be a trailing newline char in the pattern-space for the lastline (only ones from previous line-endings, via N, will be there)... sed will subsequently output the final newline char as it exits, but even so, it will be gobbled up by the xx=$(command substitution) either by $( ) or backticks ...

To include a final \n, it needs just one final substitution.

sed ':a;N;$!ba;s/\n/\\n/g;s/$/\\n/'

Or is it a case of you are not getting even the intermediate \ns?

---

Just a side note: You don't need to have umpteen seds calling that many processes. You can concatenate them via ; (colon), eg. sed 's/ě/\\\[ecaron]/g; s/š/\\\[scaron]/g; .... or not even use the colons and just put each substitution expression on a new line.. which allows them to line up nicely...