A developer released Rust Glancer, a new Language Server Protocol implementation for Rust built around one goal: using far less memory than the standard rust-analyzer, according to the project's own announcement.
Rust Glancer supports the usual LSP features, goto definition, hover, inlay hints and completions, with full type inference and trait solving. Instead of rust-analyzer's incremental, lazy computation held in memory, it indexes a workspace once, writes the results to disk, and loads only what each query needs. The author said this keeps peak memory under 100 megabytes for typical projects, down from the 16 gigabytes he said rust-analyzer had been consuming across his own open projects.
In benchmarks on a MacBook Pro M1 with 8GB of RAM, the author reported base indexing in six seconds against rust-analyzer's seven, and full indexing in nine seconds against 14.
The project was built by GitHub user popzxc, a seven-year Rust contributor with prior work on rustc, clippy and rust-analyzer itself, over about four months with heavy use of large language model assistance, according to the announcement.
Memory pressure from rust-analyzer is a common complaint on large Rust codebases, one often blamed for stalled editors on lower-spec machines. A disk-backed alternative that trades a bit of query latency for a fraction of the RAM is worth testing for anyone hitting that wall, even before it matches every feature of the incumbent.