My Favorite Bugs: Invalid Surrogate Pairs

4 hours ago

The author's favorite bug involved invalid surrogate pairs causing silent sync failures in a collaborative editor.
The bug was triggered by specific edits like inserting an emoji next to another, which split surrogate pairs.
Debugging revealed it was due to JavaScript's string methods operating on code units, not code points or graphemes.
The issue occurred in lib0's splice method using .slice(), leading to orphaned surrogates and URI errors.
A temporary fix included an error listener and making emoji atomic nodes, while lib0 was eventually patched.
The modern solution is using Intl.Segmenter for grapheme-aware string manipulation to avoid such bugs.
This bug highlights the pitfalls of UTF-16 in JavaScript and how Unicode complexities can break applications.

Hasty Briefsbeta