I want to invoke a command for every line of the standard input, much like xargs, but the line should be passed as standard input, not as a command-line argument:
cat some-file.txt | <magic> | wc -c What are my options?
How about a plain loop
while IFS= read -r line ; do printf "%s" "$line" | wc -c done < some-file.txt 
read -r line, not just read line. A while-read loop is the most clear. If you want to use xargs to do something for each line, you may end up with a monstrosity like this:
printf "%s\n" "foo bar" one " 1 2 3" | xargs -d '\n' -n 1 -I LINE bash -c 'wc -c <<< "LINE"' 8 4 7 Pretty expensive since you have to spawn a bash process for each line.
cat file.txt | while IFS= read -r i; do echo -n "$i" | wc -c; done ## or (better): while IFS= read -r i; do echo -n "$i" | wc -c; done < file.txt However, this will just print the number of characters on a line, one line at a time. If you want something more readable, you might want one of the following:
## Prints the contents of each line underneath the character count: while IFS= read -r i; do echo -n "$i" | wc -c; echo "$i"; done < file.txt ## Prints the line number alongside the character count: n=0; while IFS= read -r i; do n=$((n+1)); echo -n "line number $n : "; echo -n "$i" | wc -c; done < file.txt For greater portability you could use printf '%s' "$i" instead of all the echo -ns
Storing lines into a variable can be avoided using option -e
setopt pipefail cat some-file.txt | while read -re | wc -c do done -r reads \s as they are
-e echoes the input instead of assigning it to any variable
pipefail makes read -re | wc -c exit with non-zero status code as soon as so does read -re, which it will at the end of the file
Unfortunately, this option isn't available in Bash
--unbufferedflag these days but didn't back then).jqat the end of the question, as it implies that the data is in fact not line-based at all but structured. Any answer using line-based tools would therefore be inadequate in various ways. Also note that although the question is almost ten years old, answers may be added that solves it using up-to-date tools.awk '{ print length }'as the second and final stage of the pipeline solve the particular problem here? Or were you stuck having to callwc -cspecifically? The question is whether the issue is philosophical (and the commands shown are examples) or actual (and you actually want to count the length of each line in a particular way).wcwas just a random example.