Can we provide a json schema for tokei.toml?

[Freed-Wu]

https://www.schemastore.org/ provide schemas for auto completion of json/yaml/toml files.

[ther12k]

Yes, and it's actually straightforward because tokei already uses serde + toml (verified in Cargo.toml: toml = "0.8.19", serde = { version = "1.0.208", features = ["derive", "rc"] }, plus serde_yaml and serde_json as feature-gated extras).

The cleanest approach is schemars:

# Cargo.toml
[dependencies]
schemars = { version = "0.8", features = ["preserve_order"] }

// in src/config.rs (or wherever Config is defined)
use schemars::JsonSchema;

#[derive(Serialize, Deserialize, JsonSchema)]
#[serde(deny_unknown_fields)]
pub struct Config {
    pub columns: Option<Vec<String>>,
    pub hidden: Option<Vec<String>>,
    pub languages: Option<HashMap<String, Language>>,
    pub output: Option<Output>,
    // ...existing fields
}

Then add a small CLI command or a build.rs that dumps the schema:

fn write_schema() -> std::io::Result<()> {
    let schema = schemars::schema_for!(Config);
    std::fs::write("docs/tokei.schema.json", serde_json::to_string_pretty(&schema)?)
}

Two practical notes:

1. Tokei already requires deny_unknown_fields in most deserializers, so the derived schema will be a faithful reflection of what's accepted. Make sure to keep that deny_unknown_fields when you add JsonSchema, otherwise the schema will look more permissive than the actual parser.
2. The schema can also be exposed under a hidden flag like --print-config-schema so editors that don't have a downloaded file can still resolve it (e.g. via a Local schema in VS Code settings).

Once tokei.schema.json is in the repo, the standard pattern is:

• Commit it under docs/ (or schemas/).
• Submit it to https://github.com/SchemaStore/schemastore so editors auto-pick it up via the https://json.schemastore.org/tokei.json URL.
• Add a small note in tokei --help or the README pointing at the schema.

Net: one schemars dep, one JsonSchema derive, one schema-dump command. Happy to send a PR if useful.

[ther12k]

Self-correction after re-reading src/config.rs. Two corrections to my previous reply:

1. Config is NOT marked deny_unknown_fields. Current derive is just #[derive(Debug, Default, Deserialize)] (no Serialize, no deny_unknown_fields). My comment "make sure to keep deny_unknown_fields when you add JsonSchema" was wrong — there is none to keep, and the derived schema will reflect whatever the deserializer actually accepts (which today is permissive — unknown keys are silently dropped).

2. Config is re-exported from the crate root, so the derive lives on the re-exported type, not on tokei::config::Config:

   // src/lib.rs
   pub use self::{ config::Config, /* ... */ };

   That doesn't change the answer (schemars derives just fine), but it does mean the file that needs JsonSchema is src/config.rs line 17 (the #[derive(...)] line), and the use tokei::Config; paths in docs stay correct.

Revised patch:

// src/config.rs
use schemars::JsonSchema;
use serde::{Deserialize, Serialize};   // add Serialize

#[derive(Debug, Default, Serialize, Deserialize, JsonSchema)]
pub struct Config { /* unchanged */ }

# Cargo.toml
schemars = { version = "0.8", features = ["preserve_order"] }

Note the deliberate choice not to add #[serde(deny_unknown_fields)] — that's a separate behavior change (it would start rejecting currently-tolerated unknown keys in tokei.toml), and shouldn't ride along with a docs/schema PR. Sorry for the wrong instruction in the first reply.

[Freed-Wu]

Can you provide a PR? TIA!