Reading metadata using Python?

[molpopgen]

It's currently unclear how to do this. No effort has been made to test it. It seems quite likely that any metadata encoded using JSON would work, but we should actually show that. It is anyone's guess for the other encoding methods, as it's unclear that we can predict the layout.

[molpopgen]

To do this:

1. We need to add "set schema" functions.
2. Figure out how to do some kind of external test
   that runs a binary and outputs a trees file
   containing metadata
3. Add Python-based CI to the tests.
   For sanity, we can do this for Linux only.

[molpopgen]

#294 had some progress on this. But what's missing is a rusty way of managing the schema. We have to put more thought into this.

[molpopgen]

This crate seems helpful re: schema stuff.

[molpopgen]

jsonschema may also be helpful.

[molpopgen]

Working example of schema validation:

use serde_json::json;

#[derive(serde::Serialize, serde::Deserialize, Debug)]
struct Fwdpy11TopLevelMetadata {
    generation: u32,
    model_params: Option<String>,
    seed: Option<u32>,
    demes_graph: Option<String>,
    // Note: without this, we need to encode
    // null as one of the optional valid values.
    #[serde(skip_serializing_if = "Option::is_none")]
    data: Option<String>,
}

fn main() {
    // NOTE: the 'data' field doesn't specify null
    // and the field is one of many not "required".
    // Without extra work, we get error when None
    // serializes to NULL
    let fwdpy11_individual_metadata: serde_json::Value = json!(
        {
            "codec": "json",
            "title": "Top-level metadata for table collection/tree sequence.",
            "type": "object",
            "properties": {
                "generation": {
                    "type": "integer",
                    "description": "The value of pop.generation at the time data were exported to tskit",
                },
                "model_params": {
                    "type": ["string", "null"],
                    "description": "One or more fwdpy11.ModelParams instances.",
                },
                "seed": {
                    "type": ["integer", "null"],
                    "description": "Random number seed.",
                },
                "demes_graph": {
                    "type": ["string", "null"],
                    "description": "A demographic model specified using demes.",
                },
                "data": {
                    "type": ["string", "object"],
                    "description": "This field is reserved for the user to fill.",
                },
            },
            "required": ["generation"],
        }
    );
    let cfg = jsonschema_valid::Config::from_schema(
        &fwdpy11_individual_metadata,
        Some(jsonschema_valid::schemas::Draft::Draft7),
    )
    .unwrap();
    assert!(cfg.validate_schema().is_ok());

    let md = Fwdpy11TopLevelMetadata {
        generation: 0,
        model_params: Some("".to_string()),
        seed: Some(0),
        demes_graph: Some("".to_string()),
        data: Some("".to_string()),
    };
    let data = serde_json::to_string(&md).unwrap();
    let data: serde_json::Value = serde_json::from_str(&data).unwrap();
    println!("{data:?}");
    assert!(cfg.validate(&data).is_ok());

    let md = Fwdpy11TopLevelMetadata {
        generation: 101,
        model_params: None,
        seed: None,
        demes_graph: None,
        data: None,
    };
    let data = serde_json::to_string(&md).unwrap();
    let data: serde_json::Value = serde_json::from_str(&data).unwrap();
    println!("{data:?}");
    //assert!(cfg.validate(&data).is_ok());
    if let Err(errors) = cfg.validate(&data) {
        for error in errors {
            let f = format!("{}", error);
            println!("{f}");
        }
    };

    // Try with other crate
    let compiled = jsonschema::JSONSchema::options()
        .with_draft(jsonschema::Draft::Draft7)
        .compile(&fwdpy11_individual_metadata)
        .expect("A valid schema");
    let result = compiled.validate(&data);
    if let Err(errors) = result {
        for error in errors {
            println!("Validation error: {}", error);
            println!("Instance path: {}", error.instance_path);
        }
    }
    let md: Fwdpy11TopLevelMetadata =
        serde_json::from_str(&serde_json::to_string(&md).unwrap()).unwrap();
    println!("decoded = {md:?}");
}