Strange "echo" behavior in shell script

Question

I want to print the content that I have obtained from the split of an array in this way:

string="abc test;ABCtest.it"

IFS=';' read -a array <<< "$string"
name="${array[0]}"
url="${array[1]}"

echo -ne "\n$url,$name" >> "$outputDir/$fileName"

But the output file doesn't contain the URL part.

I think that the problem is the ., but I don't know how to fix it.

If I try this

echo $url

it works!

I've tried also with printf and with an hardcoded filename, but nothing!

printf '%s %s\n' "$url" "$name"  >> test.txt

It seems that when I try to concatenate another thing after the variable $url, some part of the variable is deleted or overwritten into the output file.

For example, if I try with this:

printf '%s %s\n' "$url" "pp"  >> test.txt

What I get with a simple cat test.txt is this:

 pptest.it

but the content of the variable $url must be ABCTest.it.

Answer 1

To complement chepner's helpful answer:

• If output doesn't look what you expect it to look like, it's always worth examining its contents to look for hidden control characters that may change data's appearance on output.

• \r, a CR (carriage return; ASCII value 13) is a notorious example for two reasons:

  • (As @chepner has stated) it moves the cursor to the beginning of the line mid-string, effectively erasing at least part of what came before it; e.g.:
    • echo $'abcd\refg' prints efgd: the \r causes everything after to restart printing at the beginning of the line, with only the d from the string before the \r surviving, because it happened to be 1 char. longer than the string that came after.
      (Note: The $'...' syntax is a so-called ANSI C-quoted string, which allows use of escape sequences such as \r in $'...\r...' to create actual control characters.)
  • Files with unexpected \r chars. occur frequently when interfacing with the Windows world, where line breaks aren't just \n chars., but \r\n sequences, and such files behave strangely in the Unix world.
• A simple way to examine data is to pipe it to cat -et, which highlights control characters as ^<char> sequences:
  • ^M represents a \r (CR)
  • ^I represents a \t (tab. char)
  • ^[ represents an ESC char.
  • ... # see 'man cat'
  • The end of a line is represented as $
  • Thus, a file with Windows-style line endings will show ^M$ at the end of the lines output by cat -et.
• cat -et applied to the above example yields the following, which makes it easy to diagnose the problem:
  • echo $'abcd\refg' | cat -et # -> 'abcd^Mefg$' - note the ^M
• dos2unix is the go-to tool for converting Windows-style line endings (\r\n) to Unix ones (\r\n), but this tool doesn't come preinstalled on most Unix-like platforms, and it's easy to use standard POSIX utilities to perform such a conversion:
  • awk 'sub("\r$", "")+1' win.txt > unix.txt
  • Note that this POSIX-compliant command doesn't allow you to replace the file in-place, however:
    • If you have GNU sed, the following would perform the conversion in place:
      • sed -i 's/\r$//' winIn_UnixOut.txt
    • Ditto with BSD sed (also used on OSX), from bash, ksh, or zsh:
      • sed -i '' $'s/\r$//' winIn_UnixOut.txt