Why does 'read' strip out whitespace even with the -N flag?

Question

According to the manual for read (and since it is a builtin, not man read, but help read):

-N nchars Return only after reading exactly NCHARS characters, unless EOF is encountered or read times out, ignoring any delimiter

However, even if the -N flag is used, read still seems to want to strip out the delimeter from input data. Take the following data as an example:

# readfail.sh - Reads 5 characters at a time, and adds "RegEx style" start and end characters for clarity
while read -r -N 5 data; do
    echo "^${data}\$"
done;

echo "^${data}\$"  # There will be data left over just because of the way `read` uses exit codes

And here is the output:

$ echo -n "Lorem ipsum dolor sit amet" | ./readfail.sh
^Lorem$
^ipsu$
^m dol$
^or si$
^t ame$
^t$

Notice the second line is only 4 characters long. That is because there is supposed to be a space between Lorem and ipsum, resulting in the following five-character output: ^ ipsu$

I can fix this problem by setting IFS= before the loop, however, isn't read supposed to be ignoring any delimiters because of the -N flag? Is this a bug, or intended behavior?

Answer 1

-N causes the end-of-line delimiter to be ignored. Bash reads exactly the specified number of characters no matter what these characters are. This is independent from the way what had been read is reported to the script in the form of variable content. That uses the usual IFS splitting mechanism.