Skip to content

G0001 core-crate-deps

Rule: A _core crate depends only on other _core crates, the allowed layers and listed vocabulary crates.

Code G0001 (stable, never reused)
Id core-crate-deps
Category graph
Source spirit_level
Fix by hand

Level per tier

Tier Level Reason
small off
small-workspace off
medium warn a _core crate holds shared types and drags nothing in
big deny a _core crate holds shared types and drags nothing in
ai off
embedded off
ui off
egui off
units-strict off
units-paranoid off

Parameters

Name Type Default Meaning
suffix string "_core" the name suffix of type-only crates
layers list of strings [] map layers a _core crate may depend on besides other _core crates
vocabulary list of strings ["serde", "serde_derive", "thiserror"] third-party or workspace crates a _core crate may use

Why

A _core crate holds the types others share: ids, data, errors, the vocabulary of a domain. To stay usable by a CLI, a server, a test or a wasm build it must not drag anything in: it depends only on other _core crates, on the layers named in layers, and on a short list of vocabulary crates (serde, thiserror).

vocabulary may also name a workspace crate: a project whose own vocabulary (a units crate, say) does not end in _core lists it there instead of moving it into the crate map's layers. An edge to the workspace-hack package is exempt, as for layer-edges: the one hakari-package names in .config/hakari.toml (or .guppy/hakari.toml), never a crate merely called workspace-hack.

How to fix

Move what needs the dependency into the crate above the _core crate, or, when the dependency really is part of the types (a units crate, say), list it in vocabulary with an override reason.

Configure

[lints.core-crate-deps]
level = "deny"
suffix = "_core"
layers = []
vocabulary = ["serde", "serde_derive", "thiserror"]
reason = "why this project wants it at this level"

Allow one finding in the code with // spirit-level: allow(core-crate-deps) <reason> (waivers); the code works there too, as in [lints.G0001].