Have you found clojure-lsp struggling with your large codebase? We did, so we made some improvements to our favorite development tool to cut startup time in half and memory allocation by two thirds.

Metabase is an open-source business intelligence (BI) tool for data visualization and analytics. Readers can find documentation, tutorials, and community forums to learn about using Metabase for data exploration, dashboard creation, and generating insights from data. Additionally, they can learn about connecting Metabase to various data sources, building custom dashboards, and sharing reports with stakeholders.

Metabase

A Metabase engineer investigated why clojure-lsp took over 3 minutes to analyze the Metabase codebase. Using VisualVM and clj-async-profiler allocation profiling, they identified key bottlenecks in clj-kondo (which accounts for 95% of initialization work). Optimizations included: improving the deep-merge function to skip unnecessary assoc operations, removing a dynamic variable on a hot path in rewrite-clj, and replacing clojure.core/memoize with a specialized implementation using nested maps for two-argument functions. The result was a 2x reduction in startup time and nearly 3x reduction in heap allocations (from 145GB to 51.6GB total allocated). Changes are merged and will ship in the next clojure-lsp release.

Improving the performance of the popular Clojure development tool clojure-lsp