#autocomplete #fst #wasm-binary #data #stored #targeting

porigon

Lightweight FST-based autocompleter library, targeting WebAssembly and data stored in-memory

4 releases (breaking)

0.4.0 Feb 1, 2022
0.3.0 Jan 13, 2022
0.2.0 May 10, 2021
0.1.0 Mar 29, 2020

#6 in #targeting

MIT license

37KB
598 lines

porigon

Lightweight FST-based autocompleter library written in Rust, targeting WebAssembly and data stored in-memory

Build status

Licensed under MIT.

Intended usecase

The idea of this library is to have a lightweight, yet idiomatic API around the fst crate that allows you to construct, serialize/deserialize and query FSTs in an WebAssembly environment. It's an ideal starting point for building an autocompleter service that can be used on the web, the edge (eg Cloudflare Worker) or the backend-side (node.js).

Existing solutions like eg tantivy are not fitting as they're too heavyweight (wasm binary size is over 1MB) or not compilable to WebAssembly. If you're looking for a more full fledged full-text search engine, take a look at the list of alternatives at the bottom.

Documentation

https://docs.rs/porigon

Installation

Simply add a corresponding entry to your Cargo.toml dependency list:

[dependencies]
porigon = "0.1.0"

Example

This example demonstrates building a Searchable in memory, executing a StartsWith query against it and collecting the top 3 documents with TopScoreCollector.

use porigon::{Searchable, TopScoreCollector};

fn main() -> Result<(), Box<dyn std::error::Error>> {
  let items = vec!(
    ("bar".as_bytes(), 1),
    ("foo".as_bytes(), 2),
    ("foobar".as_bytes(), 3)
  );
  let searchable = Searchable::build_from_iter(items)?;

  let mut collector = TopScoreCollector::new(3);
  collector.consume_stream(
    searchable
      .starts_with("foo")
      .rescore(|_, index, _| index * 2)
  );

  let docs = collector.top_documents();
  assert_eq!(docs[0].index, 3);
  
  Ok(())
}

Check out the documentation or wasm-example for a more examples.

Alternatives

If you're looking for a more general-purpose full-text search engine, take a look at these alternatives:

Dependencies

~2.5MB
~24K SLoC