bash script: write string with double quotes and blanks to file - bash

Question

I try to use sed to read a line from an ASCII file, parse it and write it slightly changed to a defined line number in an output file.
The line format in the input file is as follows:
linenumber:designator,"variable text content"
e.g.
3:string1,"this is text of string 1"
So the outfile should look as follows in line 3:
string1,"this is text of string 1"
The line includes the double quotes and the blanks. All old lines are moved one line down.
The user is responsible to provide a proper input file regarding the order of lines and has to consider that lines in the output file are moved down with each new line in the input file. The script does not know about any order except for the line number given in the input file.
A script shall read all lines and put the content of those lines into an outputfile at the given line numbers
including double quotes and blanks
without the line number part and the colon
The command I use successfully with the shell is e.g.:
sed -i '3istring1,"this is text of string 1"' outfile
No trouble with quotes, double quotes and blanks there.
Using the bash script
while read line
do
linenum=$(echo $line | cut -f1 -d:)
linestr=$(echo $line | cut -f2 -d:)
sedcmd="sed -i '"
sedcmd=${sedcmd}${linenum}
sedcmd=${sedcmd}i
sedcmd=${sedcmd}${linestr}
sedcmd=${sedcmd}"' outfile"
echo "---> $sedcmd"
$sedcmd
done < script/new_records.txt
shows exactly the same sed command with echo but returns with:
sed: -e expression #1, char 1: unknown command: `''
Apparently executing the sed command from within a bash script is different from executing it directly in the bash shell.
I tried a variety of escape sequences "\" before quotes, double quotes and blanks...but rather randomly, and neither of those was successful.
What do I have to do in order to write the string including blanks and double quotes to a specified line in a text file?

Answer 1

# Assuming OutFile exist and have enough line
while read ThisLine
do
LineNum=$(echo "${ThisLine}" | cut -f1 -d ":" )
echo "${ThisLine##*:}" > /tmp/LineContent.txt
sed -i -n "${LineNum} !{p;b;};r /tmp/LineContent.txt" OutFile
done < script/new_records.txt
Not the best thing because you assume lot of issue like enough line in outfile, no problem reading the line (what about escaped char in quoted string, ...) could occur

Answer 2

Okay, I'll give it a shot. If I understand what you're trying to do correctly, and if you're certain the code input file is not malformed, then
sed -i -f <(sed 's/:/i/' insertions.txt) datafile.txt
is the most straightforward way. This works because with an input specification of
number:text
all one has to do to is to replace the : with an i to get a sed command that says: "When handling line number, insert text". The <() bit is bash-style command substitution that expands to the name of a FIFO from which the output of the command can be read.
It might be prudent to guard against mistakes by saying something like
sed -i -f <(sed '/^[0-9]\+:/!d; s/:/i/' insertions.txt) datafile.txt
This removes all lines from insertions.txt that don't begin with a number followed by a colon because those are obviously broken.
Note that this all-in-one-go approach treats line numbers as they were in the input file. That is to say, given an insertions file with content
2:foo,"bar "
4:baz,"qux "
baz,"qux " will appear in line 5 of the output (before line 4 of the input). If this is not desired, sed will have to be called multiple times to handle each insertion individually, as in
while read insertion; do
sed -i "${insertion/:/i}" datafile.txt
done < insertions.txt
${insertion/:/i} is another bashism that replaces the first : in a shell variable with i and expands to the result, i.e., if insertion=1:2:3, then ${insertion/:/i} is 1i2:3.