Binary text and Unicode boundaries

The idea

Elixir strings are UTF-8 binaries, while charlists are lists of code points. Byte size measures storage and may differ from a text operation’s character count. String.length counts grapheme clusters, which can include more than one code point. Choose the unit from the task: network framing often uses bytes, while user-facing cursor or length behavior may require graphemes.

Binary patterns can extract fields with explicit sizes and types, making byte-oriented protocols readable. Text decoding still needs an encoding contract; slicing arbitrary bytes can split a UTF-8 sequence. Atoms serve as bounded program identifiers, not an unlimited storage format for untrusted strings. Avoid converting arbitrary external labels into newly created atoms when ordinary string keys meet the requirement.

text = "é"
IO.inspect(byte_size(text))
IO.inspect(String.length(text))
<<first::8, second::8>> = text
IO.inspect({first, second})

2
1
{195, 169}

The UTF-8 encoding contains two bytes while the displayed grapheme is one character for String.length.