Notes on word splitting

Posted on In

Code snippet:

1
2
3
4
5
6
7
8
9
#!/usr/bin/env -S bash

hello="hello:world:hello:world:hello"

IFS="lo:"

read -ra list <<<"$hello"

printf "%s*" "${list[*]}"
  1. Double quotes
  2. Bundling into an array
  3. Array expansion re-splitting

Double quotes

If $hello is wrapped with "", its literal value will be preserved. Without them, IFS characters will be displayed as blanks spots.

With Without
hello:world:hello:world:hello he(4 spaces)w r d he(4 spaces)w r d he

Bundling into an array

read -ra list cares not for whether the string is wrapped in quotes. It will split the string into an array regardless.

However, mind that the read command reads one line at a time. If the string is multiline, read will only read the first line.

Array expansion re-splitting

Both ${list[*]} and ${list[@]} are prone to re-splitting the array.

IFS \ printf "%s*" <..> ${list[*]} ${list[@]} "${list[*]}" "${list[@]}"
IFS="lo:" he****w*r*d*he****w*r*d*he** (1) (1) hellllwlrldlhellllwlrldlhell* (2) he****w*r*d*he****w*r*d*he*** (3)
IFS= he*w*r*d*he*w*r*d*he* (4) (4) (2) (3)
Observation
  • IFS="lo:"
    1. * acts like a delimiter, and fills the gap between each array element. Eg: hel * l * o. Only behaves like this when the last element contains a single delimiter, else the result is the same as (3).
    2. The first character of IFS, which happens to be l, serves as the delimiter for the array elements. The whole string is treated as one element, thus * is appended to the very end.
    3. * acts normally and is appended to each array element. Eg: hel * l * o *
  • IFS=: The effect this has on word splitting seems to be that it squeezes multiple blank arguments into one.