Show HN: Icebug-format: immutable, interoperable graph standard

Icebug-format, an open-source graph data standard built on Apache Arrow and Compressed Sparse Rows (CSR), was posted to Hacker News on August 20, 2026 by author adsharma under a Show HN submission linking to the GitHub repository. The project is MIT-licensed and targets large-scale graph workloads, with the largest converted sample graph reaching 3 billion edges.

The format pairs an on-disk representation based on Apache Parquet with an in-memory format based on Apache Arrow, per the Show HN post. Bindings are available in Python, TypeScript, and Rust. The package ships with convenience scripts to convert flat table files — such as vertex.parquet and edges.parquet — into the CSR layout under RAM- and disk-constrained conditions. Sample graphs converted from the LDBC Graphalytics benchmark datasets are published on Hugging Face under the ladybugdb organization.

The project is developed under the Ladybug-Memory GitHub organization. No founding date, team size, or funding information is mentioned in the source material. The stated motivation, per the Show HN post, is that existing graph analytics packages rely on mutable, heap-allocated vector structures that scale poorly: loading a billion-edge graph via repeated G.add_edge() calls is slow. CSR is already in use in scipy, cugraph, and columnar graph databases on both CPUs and GPUs, and the project frames itself as codifying existing practice rather than introducing a new standard.

The approach leans on formats that already have broad ecosystem support. Apache Arrow provides a language-agnostic in-memory columnar layout, and Apache Parquet offers efficient columnar storage on disk — both well-established in data engineering pipelines. By mapping graph data onto these primitives using the CSR layout, icebug-format aims to make large graphs portable across tools without requiring format conversion at read time.

The quick-start workflow, as shown in the Show HN post, involves installing via uv tool install icebug-format, running a conversion script against a source directory, and then querying the result with a lbug CLI tool using a Cypher schema file. No roadmap, contributor count, or forward-looking plans are described in the available source material.