We implement a crate that unifies the types involved in the transit path between cargo publish, the Cargo registry API, and the Cargo index, and the conversions between them. The hope is that once this is polished up some more, cargo, crates.io, and the crates-io and crates-index crates will all be able to make use of a single crate for their type definitions.

The code at the end of the stream is available at https://github.com/jonhoo/cargo-index-transit.

0:00:00 Introduction
0:03:40 cargo publish
0:15:41 The crates.io index
0:34:23 Gathering type definitions
1:43:04 Trimming down the manifest type
2:00:00 The crates-index crate
2:08:29 Undoing type-erasing
2:19:23 Biobreak
2:22:04 The .crate manifest module
3:05:15 The publish module
4:02:26 The index module
4:55:42 Trying to hide from borrowck
5:08:09 Index conversion methods
5:49:52 Writing an end-to-end test
6:01:46 Aside: A missing pub in Cargo
6:10:34 Writing an end-to-end test (part 2)
6:25:37 toml_edit and DeserializeOwned
6:57:22 Writing an end-to-end test (part 3)
7:11:05 Re-using the test setup

Live version with chat: https://youtube.com/live/NL4vCWrzzg4

Jon Gjengset

A live coding stream exploring how cargo interacts with crates.io and package registries. The host walks through the full lifecycle of cargo publish: how cargo package creates a .crate tarball, the JSON metadata sent to the registry publish endpoint, and how the registry index (both git-based and sparse HTTP-based) stores version information. The motivation is to build a new foundational Rust crate that unifies the data structure definitions and conversions currently scattered across cargo, crates.io, and crates-index — enabling shared logic for publishing, indexing, and consuming crates. The stream includes live code exploration of cargo and crates.io source, discussion of sparse registries, and the start of implementing the new crate.

From cargo to crates.io and back again