commonmeta 0.3.6

//! Schema.org JSON-LD reader and writer.

use serde::{Deserialize, Serialize};
use serde_json::Value;

use crate::data::{
    Affiliation, Contributor, Data, Description, Identifier, License, Publisher, Subject, Title,
};
use crate::doi_utils::{normalize_doi, validate_doi};
use crate::error::{Error, Result};
use crate::utils::{
    normalize_cc_url, normalize_id, normalize_orcid, normalize_ror, normalize_url, sanitize,
    url_to_spdx, validate_id,
};
use crate::vocab::CONTRIBUTOR_ROLES;

// ── Input structs ─────────────────────────────────────────────────────────────

/// Flexible affiliation — @id (optional) + sameAs (optional) + name.
#[derive(Deserialize, Default, Clone)]
pub(crate) struct SoOrganization {
    #[serde(rename = "@id", default)]
    pub(crate) id: String,
    #[serde(rename = "sameAs", default)]
    pub(crate) same_as: String,
    #[serde(default)]
    pub(crate) name: String,
}

/// Schema.org contributor (author / creator / editor / contributor).
#[derive(Deserialize, Default, Clone)]
pub(crate) struct SoContributor {
    #[serde(rename = "@id", default)]
    pub(crate) id: String,
    #[serde(rename = "@type", default)]
    pub(crate) type_: String,
    #[serde(rename = "givenName", default)]
    pub(crate) given_name: String,
    #[serde(rename = "familyName", default)]
    pub(crate) family_name: String,
    #[serde(default)]
    pub(crate) name: String,
    #[serde(default)]
    pub(crate) affiliation: Option<Value>,
}

/// Citation / reference entry.
#[allow(dead_code)]
#[derive(Deserialize, Default)]
struct SoCitation {
    #[serde(rename = "@id", default)]
    id: String,
    #[serde(rename = "@type", default)]
    type_: String,
    #[serde(default)]
    name: String,
}

/// Periodical (journal) embedded in a ScholarlyArticle.
#[allow(dead_code)]
#[derive(Deserialize, Default)]
struct SoPeriodical {
    #[serde(rename = "@id", default)]
    _id: String,
    #[serde(rename = "@type", default)]
    _type: String,
    #[serde(default)]
    name: String,
    #[serde(default)]
    issn: String,
}

/// Publisher embedded in Schema.org JSON-LD.
#[derive(Deserialize, Default)]
struct SoPublisher {
    #[serde(rename = "@type", default)]
    _type: String,
    #[serde(default)]
    name: String,
}

/// The main Schema.org content struct.
/// Polymorphic fields (author, identifier, keywords, version) are kept as
/// `Option<Value>` so they can be either a single object/string or an array.
#[derive(Deserialize, Default)]
struct SoContent {
    #[serde(rename = "@context", default)]
    _context: String,
    #[serde(rename = "@id", default)]
    id: String,
    #[serde(rename = "@type", default)]
    type_: String,
    #[serde(rename = "additionalType", default)]
    additional_type: String,
    // author / creator / contributor can be object or array
    #[serde(default)]
    author: Option<Value>,
    #[serde(default)]
    creator: Option<Value>,
    #[serde(default)]
    contributor: Option<Value>,
    // editor is treated separately (role = "Editor")
    #[serde(default)]
    editor: Option<Value>,
    // citation / references list
    #[allow(dead_code)]
    #[serde(default)]
    citation: Vec<SoCitation>,
    #[serde(rename = "dateCreated", default)]
    date_created: String,
    #[serde(rename = "datePublished", default)]
    date_published: String,
    #[serde(rename = "dateModified", default)]
    date_modified: String,
    #[serde(default)]
    description: String,
    // headline is an alternative title field
    #[serde(default)]
    headline: String,
    // identifier can be a single string or array
    #[serde(default)]
    identifier: Option<Value>,
    #[serde(rename = "inLanguage", default)]
    in_language: String,
    // keywords can be comma-separated string or array
    #[serde(default)]
    keywords: Option<Value>,
    #[serde(default)]
    license: String,
    #[serde(default)]
    name: String,
    #[allow(dead_code)]
    #[serde(default)]
    periodical: Option<SoPeriodical>,
    #[serde(default)]
    publisher: Option<SoPublisher>,
    #[serde(default)]
    url: String,
    // version can be string or number
    #[serde(default)]
    version: Option<Value>,
}

// ── Type mapping ──────────────────────────────────────────────────────────────

pub(crate) fn so_to_cm_type(so: &str) -> &'static str {
    match so {
        "Article"           => "Article",
        "BlogPosting"       => "BlogPost",
        "Book"              => "Book",
        "BookChapter"       => "BookChapter",
        "CreativeWork"      => "Other",
        "Dataset"           => "Dataset",
        "DigitalDocument"   => "Document",
        "Dissertation"      => "Dissertation",
        "Instrument"        => "Instrument",
        "NewsArticle"       => "Article",
        "Legislation"       => "LegalDocument",
        "Report"            => "Report",
        "ScholarlyArticle"  => "JournalArticle",
        "SoftwareSourceCode" => "Software",
        _                   => "",
    }
}

// ── Contributor helpers ───────────────────────────────────────────────────────

/// Deserialise a `Value` that is either a single SoContributor object or an
/// array of them.
pub(crate) fn value_to_contributors(v: &Value) -> Vec<SoContributor> {
    if v.is_null() {
        return vec![];
    }
    if let Ok(single) = serde_json::from_value::<SoContributor>(v.clone()) {
        // Only treat as a single contributor if it has at least one name field
        if !single.name.is_empty() || !single.given_name.is_empty() || !single.family_name.is_empty() {
            return vec![single];
        }
    }
    serde_json::from_value::<Vec<SoContributor>>(v.clone()).unwrap_or_default()
}

/// Convert a raw JSON affiliation value into an `Affiliation`.
pub(crate) fn parse_affiliation(v: &Value) -> Option<Affiliation> {
    // affiliation can be an Organization object or a plain string
    if let Some(s) = v.as_str() {
        if !s.is_empty() {
            return Some(Affiliation { name: s.to_string(), ..Default::default() });
        }
        return None;
    }
    if let Ok(org) = serde_json::from_value::<SoOrganization>(v.clone()) {
        if org.name.is_empty() {
            return None;
        }
        let id = {
            let ror = normalize_ror(&org.id);
            if !ror.is_empty() {
                ror
            } else {
                normalize_ror(&org.same_as)
            }
        };
        return Some(Affiliation { id, name: org.name, ..Default::default() });
    }
    None
}

/// Mirror of Go `GetContributor`.
pub(crate) fn get_contributor(v: SoContributor, default_role: &str) -> Contributor {
    let mut type_ = v.type_.clone();
    let mut id = String::new();

    // Resolve @id — try ORCID first, then ROR
    if !v.id.is_empty() {
        let orcid = normalize_orcid(&v.id);
        if !orcid.is_empty() {
            id = orcid;
            type_ = "Person".to_string();
        } else {
            let ror = normalize_ror(&v.id);
            if !ror.is_empty() {
                id = ror;
                type_ = "Organization".to_string();
            }
        }
    }

    let mut name = v.name.clone();
    let mut given_name = v.given_name.clone();
    let mut family_name = v.family_name.clone();

    // Infer type from names if still unknown
    if type_.is_empty() {
        if !given_name.is_empty() || !family_name.is_empty() {
            type_ = "Person".to_string();
        } else {
            type_ = "Organization".to_string();
        }
    }

    // Split combined name for Persons
    if type_ == "Person" && !name.is_empty() && given_name.is_empty() && family_name.is_empty() {
        // Try "Family, Given"
        if let Some(comma) = name.find(',') {
            family_name = name[..comma].trim().to_string();
            given_name = name[comma + 1..].trim().to_string();
            name = String::new();
        } else {
            // "Given Family" — first token is given, rest is family
            let parts: Vec<&str> = name.splitn(2, ' ').collect();
            if parts.len() == 2 {
                given_name = parts[0].to_string();
                family_name = parts[1].to_string();
                name = String::new();
            }
        }
    }

    // Affiliation
    let affiliations: Vec<Affiliation> = if let Some(aff_val) = &v.affiliation {
        // Can be a single object/string or an array
        if let Ok(list) = serde_json::from_value::<Vec<Value>>(aff_val.clone()) {
            list.iter().filter_map(parse_affiliation).collect()
        } else {
            parse_affiliation(aff_val).into_iter().collect()
        }
    } else {
        vec![]
    };

    // Role: honour if the @type is a known contributor role, else default
    let role = if CONTRIBUTOR_ROLES.contains(&v.type_.as_str()) {
        v.type_.clone()
    } else {
        default_role.to_string()
    };

    Contributor {
        id,
        type_,
        name,
        given_name,
        family_name,
        affiliations,
        contributor_roles: vec![role],
    }
}

// ── Core conversion ───────────────────────────────────────────────────────────

fn from_content(content: SoContent) -> Data {
    // ID and Type computed up front for struct init
    let id = normalize_id(&content.id);
    let cm_type = so_to_cm_type(&content.type_);
    let type_ = if cm_type.is_empty() {
        "WebPage".to_string()
    } else {
        cm_type.to_string()
    };
    let mut data = Data {
        id,
        type_,
        additional_type: content.additional_type.clone(),
        ..Data::default()
    };

    // Contributors from author/creator (role = Author)
    let author_val = content.author.or(content.creator).unwrap_or(Value::Null);
    for v in value_to_contributors(&author_val) {
        if v.name.is_empty() && v.given_name.is_empty() && v.family_name.is_empty() {
            continue;
        }
        let contrib = get_contributor(v, "Author");
        let dup = !contrib.id.is_empty()
            && data.contributors.iter().any(|c| !c.id.is_empty() && c.id == contrib.id);
        if !dup {
            data.contributors.push(contrib);
        }
    }

    // Contributors from contributor (role = Author unless type matches)
    if let Some(ref contrib_val) = content.contributor {
        for v in value_to_contributors(contrib_val) {
            if v.name.is_empty() && v.given_name.is_empty() && v.family_name.is_empty() {
                continue;
            }
            let contrib = get_contributor(v, "Author");
            let dup = !contrib.id.is_empty()
                && data.contributors.iter().any(|c| !c.id.is_empty() && c.id == contrib.id);
            if !dup {
                data.contributors.push(contrib);
            }
        }
    }

    // Contributors from editor (role = Editor)
    if let Some(ref editor_val) = content.editor {
        for v in value_to_contributors(editor_val) {
            if v.name.is_empty() && v.given_name.is_empty() && v.family_name.is_empty() {
                continue;
            }
            let contrib = get_contributor(v, "Editor");
            let dup = !contrib.id.is_empty()
                && data.contributors.iter().any(|c| !c.id.is_empty() && c.id == contrib.id);
            if !dup {
                data.contributors.push(contrib);
            }
        }
    }

    // Dates
    if !content.date_published.is_empty() {
        data.date.published = content.date_published.clone();
    }
    if !content.date_modified.is_empty() {
        data.date.updated = content.date_modified.clone();
    }
    if !content.date_created.is_empty() {
        data.date.created = content.date_created.clone();
    }

    // Description
    if !content.description.is_empty() {
        data.descriptions.push(Description {
            description: sanitize(&content.description),
            type_: "Abstract".to_string(),
            language: String::new(),
        });
    }

    // Identifiers — string or array
    let idents: Vec<String> = match &content.identifier {
        Some(Value::String(s)) => vec![s.clone()],
        Some(Value::Array(arr)) => arr
            .iter()
            .filter_map(|v| v.as_str().map(str::to_string))
            .collect(),
        _ => vec![],
    };
    for id_str in &idents {
        if id_str != &data.id {
            let (identifier, identifier_type) = validate_id(id_str);
            if !identifier.is_empty() {
                let identifier = if identifier_type == "DOI" {
                    normalize_doi(&identifier)
                } else {
                    identifier
                };
                data.identifiers.push(Identifier { identifier, identifier_type: identifier_type.to_string() });
            }
        }
    }

    // Language
    data.language = content.in_language.clone();

    // License
    if !content.license.is_empty() {
        let (url, ok) = normalize_cc_url(&content.license);
        if ok {
            let id = url_to_spdx(&url);
            data.license = License { id, url };
        }
    }

    // Provider — inferred from DOI RA
    if let Some(bare) = validate_doi(&data.id) {
        use crate::doi_utils::get_doi_ra_sync;
        if let Some(ra) = get_doi_ra_sync(&bare) {
            data.provider = ra;
        }
    }

    // Publisher
    if let Some(pub_) = content.publisher
        && !pub_.name.is_empty() {
            data.publisher = Publisher { name: pub_.name, ..Default::default() };
        }

    // Subjects (keywords — string or array)
    let keywords: Vec<String> = match &content.keywords {
        Some(Value::String(s)) => s.split(',').map(|k| k.trim().to_string()).collect(),
        Some(Value::Array(arr)) => arr
            .iter()
            .filter_map(|v| v.as_str().map(str::to_string))
            .collect(),
        _ => vec![],
    };
    for kw in keywords {
        if !kw.is_empty() {
            data.subjects.push(Subject { subject: kw });
        }
    }

    // Titles
    if !content.name.is_empty() {
        data.titles.push(Title { title: content.name.clone(), ..Default::default() });
    } else if !content.headline.is_empty() {
        data.titles.push(Title { title: content.headline.clone(), ..Default::default() });
    }

    // URL
    if let Some(url) = normalize_url(&content.url, true, false) {
        data.url = url;
    }

    // Version — string or number
    if let Some(v) = &content.version {
        data.version = match v {
            Value::String(s) => s.clone(),
            Value::Number(n) => format!("{}", n),
            _ => String::new(),
        };
    }

    data
}

// ── Public API ────────────────────────────────────────────────────────────────

/// Read a Schema.org JSON-LD string directly into `Data`.
pub fn read_json(input: &str) -> Result<Data> {
    let content: SoContent =
        serde_json::from_str(input).map_err(|e| Error::Parse(e.to_string()))?;
    Ok(from_content(content))
}

/// Fetch a URL, extract its JSON-LD, and parse into `Data`.
///
/// falls back to `<meta>` tags, then dispatches to Crossref or DataCite when the
/// embedded DOI belongs to one of those registrars.
pub fn fetch(url: &str) -> Result<Data> {
    let client = reqwest::blocking::Client::builder()
        .user_agent(format!(
            "commonmeta-rs/{} (https://github.com/front-matter/commonmeta-rs; mailto:info@front-matter.de)",
            env!("CARGO_PKG_VERSION")
        ))
        .redirect(reqwest::redirect::Policy::limited(5))
        .timeout(std::time::Duration::from_secs(10))
        .build()
        .map_err(|e| Error::Http(e.to_string()))?;

    let html = client
        .get(url)
        .send()
        .map_err(|e| Error::Http(e.to_string()))?
        .error_for_status()
        .map_err(|e| Error::Http(e.to_string()))?
        .text()
        .map_err(|e| Error::Http(e.to_string()))?;

    let content = extract_content(&html)?;

    // If DOI belongs to a known RA, hand off to that reader
    if let Some(bare) = validate_doi(&content.id) {
        use crate::doi_utils::get_doi_ra_sync;
        if let Some(ra) = get_doi_ra_sync(&bare) {
            match ra.as_str() {
                "Crossref" => return crate::formats::crossref::fetch(&content.id),
                "DataCite" => return crate::formats::datacite::fetch(&content.id),
                _ => {}
            }
        }
    }

    Ok(from_content(content))
}

/// Extract Schema.org JSON-LD (+ meta-tag fallbacks) from raw HTML.
fn extract_content(html: &str) -> Result<SoContent> {
    use scraper::{Html, Selector};

    let doc = Html::parse_document(html);

    // ── JSON-LD ───────────────────────────────────────────────────────────────
    let script_sel =
        Selector::parse("script[type='application/ld+json']").expect("valid selector");
    let mut content = SoContent::default();
    for el in doc.select(&script_sel) {
        let text = el.text().collect::<String>();
        if let Ok(parsed) = serde_json::from_str::<SoContent>(&text) {
            content = parsed;
            break;
        }
    }

    // ── @id / DOI fallbacks ───────────────────────────────────────────────────
    if content.id.is_empty() {
        let meta_names = [
            "meta[name='citation_doi']",
            "meta[name='dc.identifier']",
            "meta[name='DC.identifier']",
            "meta[name='bepress_citation_doi']",
        ];
        for sel_str in &meta_names {
            if let Ok(sel) = Selector::parse(sel_str)
                && let Some(el) = doc.select(&sel).next()
                    && let Some(val) = el.value().attr("content") {
                        content.id = val.to_string();
                        break;
                    }
        }
    }

    // ── Type fallback ─────────────────────────────────────────────────────────
    if content.type_.is_empty() {
        let type_metas = [
            "meta[property='og:type']",
            "meta[name='dc.type']",
            "meta[name='DC.type']",
        ];
        for sel_str in &type_metas {
            if let Ok(sel) = Selector::parse(sel_str)
                && let Some(el) = doc.select(&sel).next()
                    && let Some(val) = el.value().attr("content") {
                        content.type_ = val.to_string();
                        break;
                    }
        }
    }

    // ── Name / headline fallbacks ─────────────────────────────────────────────
    if content.name.is_empty() {
        let name_metas = [
            "meta[name='citation_title']",
            "meta[name='dc.title']",
            "meta[name='DC.title']",
            "meta[property='og:title']",
            "meta[name='twitter:title']",
        ];
        for sel_str in &name_metas {
            if let Ok(sel) = Selector::parse(sel_str)
                && let Some(el) = doc.select(&sel).next() {
                    let val = el
                        .value()
                        .attr("content")
                        .unwrap_or("")
                        .trim()
                        .to_string();
                    if !val.is_empty() {
                        content.name = val;
                        break;
                    }
                }
        }
    }

    // ── Description fallbacks ─────────────────────────────────────────────────
    if content.description.is_empty() {
        let desc_metas = [
            "meta[name='citation_abstract']",
            "meta[name='dc.description']",
            "meta[property='og:description']",
            "meta[name='twitter:description']",
        ];
        for sel_str in &desc_metas {
            if let Ok(sel) = Selector::parse(sel_str)
                && let Some(el) = doc.select(&sel).next() {
                    let val = el
                        .value()
                        .attr("content")
                        .unwrap_or("")
                        .trim()
                        .to_string();
                    if !val.is_empty() {
                        content.description = val;
                        break;
                    }
                }
        }
    }

    // ── Date published fallbacks ──────────────────────────────────────────────
    if content.date_published.is_empty() {
        let date_metas = [
            "meta[name='citation_publication_date']",
            "meta[name='citation_date']",
            "meta[name='dc.date']",
            "meta[property='article:published_time']",
        ];
        for sel_str in &date_metas {
            if let Ok(sel) = Selector::parse(sel_str)
                && let Some(el) = doc.select(&sel).next()
                    && let Some(val) = el.value().attr("content") {
                        content.date_published = val.to_string();
                        break;
                    }
        }
    }

    // ── Date modified fallbacks ───────────────────────────────────────────────
    if content.date_modified.is_empty() {
        let mod_metas = [
            "meta[name='og:updated_time']",
            "meta[name='article:modified_time']",
        ];
        for sel_str in &mod_metas {
            if let Ok(sel) = Selector::parse(sel_str)
                && let Some(el) = doc.select(&sel).next()
                    && let Some(val) = el.value().attr("content") {
                        content.date_modified = val.to_string();
                        break;
                    }
        }
    }

    // ── Language fallback ─────────────────────────────────────────────────────
    if content.in_language.is_empty()
        && let Ok(sel) = Selector::parse("html")
            && let Some(el) = doc.select(&sel).next()
                && let Some(lang) = el.value().attr("lang") {
                    content.in_language = lang.to_string();
                }

    // ── License fallback ──────────────────────────────────────────────────────
    if content.license.is_empty()
        && let Ok(sel) = Selector::parse("link[rel='license']")
            && let Some(el) = doc.select(&sel).next()
                && let Some(href) = el.value().attr("href") {
                    content.license = href.to_string();
                }

    // ── author/creator synonyms ───────────────────────────────────────────────
    if content.author.is_none()
        && let Some(creator) = content.creator.take() {
            content.author = Some(creator);
        }

    Ok(content)
}

// ── Writer ────────────────────────────────────────────────────────────────────

// ── Output structs ────────────────────────────────────────────────────────────

#[derive(Serialize)]
struct OutOrganization {
    #[serde(rename = "@id", skip_serializing_if = "String::is_empty")]
    id: String,
    #[serde(rename = "@type")]
    type_: &'static str,
    #[serde(skip_serializing_if = "String::is_empty")]
    name: String,
}

#[derive(Serialize)]
struct OutContributor {
    #[serde(rename = "@id", skip_serializing_if = "String::is_empty")]
    id: String,
    #[serde(rename = "@type")]
    type_: &'static str,
    #[serde(rename = "givenName", skip_serializing_if = "String::is_empty")]
    given_name: String,
    #[serde(rename = "familyName", skip_serializing_if = "String::is_empty")]
    family_name: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    name: String,
    #[serde(skip_serializing_if = "Option::is_none")]
    affiliation: Option<OutOrganization>,
}

#[derive(Serialize)]
struct OutCitation {
    #[serde(rename = "@id")]
    id: String,
    #[serde(rename = "@type")]
    type_: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    name: String,
}

#[derive(Serialize, Default)]
struct OutPeriodical {
    #[serde(rename = "@id", skip_serializing_if = "String::is_empty")]
    id: String,
    #[serde(rename = "@type")]
    type_: &'static str,
    #[serde(skip_serializing_if = "String::is_empty")]
    name: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    issn: String,
}

#[derive(Serialize)]
struct OutDataCatalog {
    #[serde(rename = "@id", skip_serializing_if = "String::is_empty")]
    id: String,
    #[serde(rename = "@type")]
    type_: &'static str,
    #[serde(skip_serializing_if = "String::is_empty")]
    name: String,
}

#[derive(Serialize)]
struct OutMediaObject {
    #[serde(rename = "@type")]
    type_: &'static str,
    #[serde(rename = "contentUrl")]
    content_url: String,
    #[serde(rename = "encodingFormat", skip_serializing_if = "String::is_empty")]
    encoding_format: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    name: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    sha256: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    size: String,
}

#[derive(Serialize)]
struct OutProvider {
    #[serde(rename = "@type")]
    type_: &'static str,
    name: String,
}

#[derive(Serialize)]
struct OutPublisher {
    #[serde(rename = "@type")]
    type_: &'static str,
    name: String,
}

#[derive(Serialize)]
struct OutPayload {
    #[serde(rename = "@context")]
    context: &'static str,
    #[serde(rename = "@id")]
    id: String,
    #[serde(rename = "@type")]
    type_: String,
    #[serde(rename = "additionalType", skip_serializing_if = "String::is_empty")]
    additional_type: String,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    author: Vec<OutContributor>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    editor: Vec<OutContributor>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    citation: Vec<OutCitation>,
    #[serde(rename = "includedInDataCatalog", skip_serializing_if = "Option::is_none")]
    included_in_data_catalog: Option<OutDataCatalog>,
    #[serde(skip_serializing_if = "Option::is_none")]
    periodical: Option<OutPeriodical>,
    #[serde(rename = "dateCreated", skip_serializing_if = "String::is_empty")]
    date_created: String,
    #[serde(rename = "datePublished", skip_serializing_if = "String::is_empty")]
    date_published: String,
    #[serde(rename = "dateModified", skip_serializing_if = "String::is_empty")]
    date_modified: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    description: String,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    distribution: Vec<OutMediaObject>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    encoding: Vec<OutMediaObject>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    identifier: Vec<String>,
    #[serde(rename = "inLanguage", skip_serializing_if = "String::is_empty")]
    in_language: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    keywords: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    license: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    name: String,
    #[serde(rename = "pageStart", skip_serializing_if = "String::is_empty")]
    page_start: String,
    #[serde(rename = "pageEnd", skip_serializing_if = "String::is_empty")]
    page_end: String,
    provider: OutProvider,
    publisher: OutPublisher,
    #[serde(skip_serializing_if = "String::is_empty")]
    url: String,
    #[serde(skip_serializing_if = "String::is_empty")]
    version: String,
}

// ── CM → Schema.org type mapping ─────────────────────────────────────────────

fn cm_to_so_type(cm: &str) -> &'static str {
    match cm {
        "Article"               => "Article",
        "Audiovisual"           => "CreativeWork",
        "BlogPost"              => "BlogPosting",
        "Book"                  => "Book",
        "BookChapter"           => "BookChapter",
        "Collection"            => "CreativeWork",
        "Dataset"               => "Dataset",
        "Dissertation"          => "Dissertation",
        "Document"              => "CreativeWork",
        "Entry"                 => "CreativeWork",
        "Event"                 => "CreativeWork",
        "Figure"                => "CreativeWork",
        "Image"                 => "CreativeWork",
        "Instrument"            => "Instrument",
        "JournalArticle"        => "ScholarlyArticle",
        "LegalDocument"         => "Legislation",
        "Software"              => "SoftwareSourceCode",
        "Presentation"          => "PresentationDigitalDocument",
        _                       => "CreativeWork",
    }
}

// ── Conversion ────────────────────────────────────────────────────────────────

fn convert(data: &crate::data::Data) -> OutPayload {
    // Authors and editors
    let mut authors: Vec<OutContributor> = Vec::new();
    let mut editors: Vec<OutContributor> = Vec::new();

    for c in &data.contributors {
        let is_author = c.contributor_roles.contains(&"Author".to_string());
        let is_editor = c.contributor_roles.contains(&"Editor".to_string());
        if !is_author && !is_editor {
            continue;
        }

        let affiliation = c.affiliations.first().map(|a| OutOrganization {
            id: a.id.clone(),
            type_: "Organization",
            name: a.name.clone(),
        });

        let out = if c.type_ == "Organization" {
            OutContributor {
                id: c.id.clone(),
                type_: "Organization",
                given_name: String::new(),
                family_name: String::new(),
                name: c.name.clone(),
                affiliation: None,
            }
        } else {
            OutContributor {
                id: c.id.clone(),
                type_: "Person",
                given_name: c.given_name.clone(),
                family_name: c.family_name.clone(),
                name: String::new(),
                affiliation,
            }
        };

        if is_author {
            authors.push(out);
        } else {
            editors.push(out);
        }
    }

    // Citations from references
    let citation: Vec<OutCitation> = data
        .references
        .iter()
        .filter(|r| !r.id.is_empty())
        .map(|r| {
            let type_ = if r.type_ == "JournalArticle" {
                "ScholarlyArticle".to_string()
            } else {
                "CreativeWork".to_string()
            };
            OutCitation { id: r.id.clone(), type_, name: r.title.clone() }
        })
        .collect();

    // Container → periodical or data catalog
    let (included_in_data_catalog, periodical) = if data.type_ == "Dataset" {
        let cat = OutDataCatalog {
            id: data.container.identifier.clone(),
            type_: "DataCatalog",
            name: data.container.title.clone(),
        };
        (Some(cat), None)
    } else {
        let (id, issn) = if data.container.identifier_type == "ISSN" {
            (String::new(), data.container.identifier.clone())
        } else {
            (data.container.identifier.clone(), String::new())
        };
        let p = OutPeriodical {
            id,
            type_: "Periodical",
            name: data.container.title.clone(),
            issn,
        };
        // Only include periodical if it has something meaningful
        if p.name.is_empty() && p.issn.is_empty() && p.id.is_empty() {
            (None, None)
        } else {
            (None, Some(p))
        }
    };

    // Files → MediaObject
    let media_objects: Vec<OutMediaObject> = data
        .files
        .iter()
        .map(|f| OutMediaObject {
            type_: "MediaObject",
            content_url: f.url.clone(),
            encoding_format: f.mime_type.clone(),
            name: f.key.clone(),
            sha256: f.checksum.clone(),
            size: if f.size > 0 { f.size.to_string() } else { String::new() },
        })
        .collect();

    let (distribution, encoding) = if data.type_ == "Dataset" {
        (media_objects, vec![])
    } else {
        (vec![], media_objects)
    };

    // Identifiers
    let identifier: Vec<String> = data
        .identifiers
        .iter()
        .map(|i| i.identifier.clone())
        .collect();

    // Keywords
    let keywords = if data.subjects.is_empty() {
        String::new()
    } else {
        data.subjects
            .iter()
            .filter(|s| !s.subject.is_empty())
            .map(|s| s.subject.as_str())
            .collect::<Vec<_>>()
            .join(", ")
    };

    // Title
    let name = data.titles.first().map(|t| t.title.clone()).unwrap_or_default();

    // Description
    let description = data
        .descriptions
        .first()
        .map(|d| d.description.clone())
        .unwrap_or_default();

    OutPayload {
        context: "http://schema.org",
        id: data.id.clone(),
        type_: cm_to_so_type(&data.type_).to_string(),
        additional_type: data.additional_type.clone(),
        author: authors,
        editor: editors,
        citation,
        included_in_data_catalog,
        periodical,
        date_created: data.date.created.clone(),
        date_published: data.date.published.clone(),
        date_modified: data.date.updated.clone(),
        description,
        distribution,
        encoding,
        identifier,
        in_language: data.language.clone(),
        keywords,
        license: data.license.url.clone(),
        name,
        page_start: data.container.first_page.clone(),
        page_end: data.container.last_page.clone(),
        provider: OutProvider { type_: "Organization", name: data.provider.clone() },
        publisher: OutPublisher { type_: "Organization", name: data.publisher.name.clone() },
        url: data.url.clone(),
        version: data.version.clone(),
    }
}

pub fn write(data: &crate::data::Data) -> crate::error::Result<Vec<u8>> {
    let payload = convert(data);
    serde_json::to_vec(&payload).map_err(|e| crate::error::Error::Serialize(e.to_string()))
}

pub fn write_all(list: &[crate::data::Data]) -> crate::error::Result<Vec<u8>> {
    let payloads: Vec<OutPayload> = list.iter().map(convert).collect();
    serde_json::to_vec_pretty(&payloads)
        .map_err(|e| crate::error::Error::Serialize(e.to_string()))
}