A visual guide to Rust async

2026-09-07T00:00:00+00:00

One trie, three jobs, zero benchmarks won

2026-08-25T00:00:00+00:00

I dusted off an old project of mine and, with the help of AI, freshened it up and made some improvements. This article is written with the help of AI too, but it's my project, my design, and I can explain every line.

I recently open-sourced wordtree</a>, a compact trie for word lists. First of all, here is what it is not: it is not the fastest at anything. I benchmarked it against a specialist crate for each job it does, and each specialist beat it in its own domain. Exact lookup is slower than a HashMap</code>. The file is three times larger than an FST. Spelling correction is an order of magnitude slower than symspell.

So why did I do it? Because I needed it! A structure that loses every micro-benchmark can still be the right dependency.

The whole comparative study is reproducible — every number below comes fromcomparisons/REPORT.md</a>, regenerable with four cargo</code> commands against the word lists bundled in the repo.

Three jobs</h2> wordtree came out of a translation app that needed three things from one big word list, all at once, on devices where startup time and memory both mattered: A browsable index. Group the words into folders (~100 per folder) so a UI can page through them. path_of("apricot")</code> returns the folder path.</li> Exact lookup. Resolve a word to the index of its expression in O(word length)</code>. index_of("apple")</code> → Some(1)</code>.</li> Typo-tolerant autocomplete. Frequency-ranked, as-you-type suggestions that both extend a prefix ("ap"</code> → apple</code>, apply</code>) and fix a single typo — substitution, transposition, insertion, or deletion at Damerau-Levenshtein distance ≤ 1 ("aple"</code> → apple</code>). suggestions("aple", …)</code>.</li> </ol> Each of those jobs has a specialist crate that does it better. What almost nothing does is all three from one structure, from one file that loads with zero parsing. That last constraint is the whole story, so I'll start there. The structure: 8 bytes a node</h2> The tree is a width-first array of fixed-size nodes: a node is immediately followed by all its siblings, so "next sibling" is the next slot and "first child" is one 24-bit index.

Henrik Åkesson

A visual guide to Rust async

One trie, three jobs, zero benchmarks won

The structure: 8 bytes a node</h2>
The tree is a width-first array of fixed-size nodes: a node is immediately followed by all its siblings, so "next sibling" is the next slot and "first child" is one 24-bit index.</p>

Dusting off my archives