defter 0.8.3

//! The public profile at `/<orcid>`: a user's ORCID record, read by commonmeta
//! and laid out as on commonmeta.org's person page, and their My Publications,
//! as zotero.org publishes it. The page itself is rendered in `ui`.

use std::collections::HashMap;
use std::sync::{Arc, Mutex};
use std::time::{Duration, Instant};

use commonmeta::Person;
use serde_json::{Value, json};

use crate::app::AppState;
use crate::countries;
use crate::error::AppError;
use crate::i18n::Lang;
use crate::store;

/// Works per page of My Publications.
pub const PAGE_SIZE: usize = 10;

/// Featured works shown at most, which the profile opens with.
pub const FEATURED_MAX: usize = 10;

/// Works in one export.
pub const EXPORT_MAX: usize = 10_000;

/// Values shown per facet, as on commonmeta.org.
pub const FACET_VALUES: usize = 10;

/// The length of a search term taken from one request.
pub const SEARCH_MAX: usize = 200;

/// Values of a facet taken from one request.
pub const FACET_PICKS_MAX: usize = 20;

/// Websites listed on the person card, as on commonmeta.org.
const URLS_MAX: usize = 5;

/// How long a fetched ORCID record is shown before it is fetched again.
const PERSON_TTL: Duration = Duration::from_secs(3600);

/// How long after a failed fetch the next one is tried, so that page views
/// during an ORCID outage do not each wait for it.
const PERSON_RETRY: Duration = Duration::from_secs(300);

/// How long a page view waits for ORCID.
const PERSON_TIMEOUT: Duration = Duration::from_secs(15);

/// ORCID records kept at most.
const PEOPLE_MAX: usize = 256;

/// The `meta` key that recorded whose profile it is, in a library from before
/// the control database (see `migrate`).
pub const OWNER_KEY: &str = "profile_owner_orcid";
/// The `meta` key that recorded that the owner hid the profile, likewise.
pub const HIDDEN_KEY: &str = "profile_hidden";

/// The bare iD (`0000-0003-1419-2405`) of an ORCID iD or URL, if valid.
pub fn normalize_orcid(input: &str) -> Option<String> {
    commonmeta::utils::validate_orcid(input.trim()).map(|id| id.replace(' ', "-"))
}

/// The ORCID records fetched, by iD, and when to fetch each again.
#[derive(Default)]
pub struct PersonCache {
    slots: Mutex<HashMap<String, PersonSlot>>,
    /// Held while fetching, so that concurrent page views fetch only once.
    fetching: tokio::sync::Mutex<()>,
}

#[derive(Default)]
struct PersonSlot {
    person: Option<Arc<Person>>,
    next_fetch: Option<Instant>,
}

impl PersonCache {
    fn slots(&self) -> std::sync::MutexGuard<'_, HashMap<String, PersonSlot>> {
        self.slots.lock().unwrap_or_else(|e| e.into_inner())
    }

    pub fn set(&self, orcid: &str, person: Person) -> Arc<Person> {
        let person = Arc::new(person);
        let mut slots = self.slots();
        slots.insert(
            orcid.to_string(),
            PersonSlot {
                person: Some(person.clone()),
                next_fetch: Some(Instant::now() + PERSON_TTL),
            },
        );
        // Past the limit, the record due soonest goes: it would be fetched again
        // first anyway.
        while slots.len() > PEOPLE_MAX {
            let Some(soonest) = slots
                .iter()
                .min_by_key(|(_, slot)| slot.next_fetch)
                .map(|(orcid, _)| orcid.clone())
            else {
                break;
            };
            slots.remove(&soonest);
        }
        person
    }

    /// Keep what there is, and fetch again after a while.
    fn fetch_failed(&self, orcid: &str) {
        self.slots()
            .entry(orcid.to_string())
            .or_default()
            .next_fetch = Some(Instant::now() + PERSON_RETRY);
    }

    /// What to show until the next fetch is due: the record, if there is one;
    /// `None` when it is time to fetch.
    fn settled(&self, orcid: &str) -> Option<Option<Arc<Person>>> {
        let slots = self.slots();
        let slot = slots.get(orcid)?;
        let due = slot.next_fetch.is_none_or(|at| Instant::now() >= at);
        (!due).then(|| slot.person.clone())
    }

    fn stale(&self, orcid: &str) -> Option<Arc<Person>> {
        self.slots().get(orcid).and_then(|slot| slot.person.clone())
    }

    /// The record held for `orcid`, fresh or not, and whether it is due to be
    /// fetched (again): for a list of people, which must not wait on ORCID.
    pub fn cached(&self, orcid: &str) -> (Option<Arc<Person>>, bool) {
        (self.stale(orcid), self.settled(orcid).is_none())
    }
}

fn unavailable(orcid: &str) -> AppError {
    AppError::Upstream(format!("ORCID record {orcid} is not available"))
}

/// The ORCID record of `orcid`, from the cache while it is fresh. When ORCID
/// cannot be reached, a stale copy is better than no page.
pub async fn person(app: &AppState, orcid: &str) -> Result<Arc<Person>, AppError> {
    if let Some(person) = app.person.settled(orcid) {
        return person.ok_or_else(|| unavailable(orcid));
    }
    let _fetching = app.person.fetching.lock().await;
    // Another page view may have fetched it in the meantime.
    if let Some(person) = app.person.settled(orcid) {
        return person.ok_or_else(|| unavailable(orcid));
    }
    let id = orcid.to_string();
    let fetch = tokio::task::spawn_blocking(move || commonmeta::fetch_orcid(&id));
    let error = match tokio::time::timeout(PERSON_TIMEOUT, fetch).await {
        Ok(Ok(Ok(person))) => return Ok(app.person.set(orcid, person)),
        Ok(Ok(Err(error))) => error.to_string(),
        Ok(Err(error)) => format!("blocking task: {error}"),
        Err(_) => "timed out".to_string(),
    };
    tracing::warn!(orcid, %error, "fetching the ORCID record");
    app.person.fetch_failed(orcid);
    app.person.stale(orcid).ok_or_else(|| unavailable(orcid))
}

/// Each public work's language and license, as an ISO 639-1 code and an SPDX
/// ID, where Zotero's text names one (see `vocab`).
pub struct WorkTerms(Vec<(String, Option<String>, Option<String>)>);

impl WorkTerms {
    pub fn new(rows: Vec<store::WorkTerms>) -> Self {
        // A library holds few distinct values, each mapped once.
        let mut languages: HashMap<String, Option<String>> = HashMap::new();
        let mut licenses: HashMap<String, Option<String>> = HashMap::new();
        WorkTerms(
            rows.into_iter()
                .map(|row| {
                    let language = languages
                        .entry(row.language)
                        .or_insert_with_key(|text| crate::vocab::language(text))
                        .clone();
                    let license = licenses
                        .entry(row.rights)
                        .or_insert_with_key(|text| crate::vocab::license(text))
                        .clone();
                    (row.key, language, license)
                })
                .collect(),
        )
    }

    fn counts<'a>(values: impl Iterator<Item = &'a str>) -> Vec<(String, usize)> {
        let mut counts: HashMap<&str, usize> = HashMap::new();
        for value in values {
            *counts.entry(value).or_default() += 1;
        }
        let mut counts: Vec<(String, usize)> = counts
            .into_iter()
            .map(|(value, n)| (value.to_string(), n))
            .collect();
        counts.sort_by(|a, b| b.1.cmp(&a.1).then_with(|| a.0.cmp(&b.0)));
        counts.truncate(FACET_VALUES);
        counts
    }

    /// The most common languages, with how many works are in each.
    pub fn languages(&self) -> Vec<(String, usize)> {
        Self::counts(self.0.iter().filter_map(|(_, l, _)| l.as_deref()))
    }

    /// The most common licenses, with how many works are under each.
    pub fn licenses(&self) -> Vec<(String, usize)> {
        Self::counts(self.0.iter().filter_map(|(_, _, l)| l.as_deref()))
    }

    /// The works in one of `languages` and under one of `licenses`, where
    /// either names any; `None` when neither does.
    pub fn keys(&self, languages: &[String], licenses: &[String]) -> Option<Arc<Vec<String>>> {
        if languages.is_empty() && licenses.is_empty() {
            return None;
        }
        let among = |picked: &[String], value: &Option<String>| {
            picked.is_empty() || value.as_ref().is_some_and(|v| picked.contains(v))
        };
        Some(Arc::new(
            self.0
                .iter()
                .filter(|(_, language, license)| {
                    among(languages, language) && among(licenses, license)
                })
                .map(|(key, _, _)| key.clone())
                .collect(),
        ))
    }
}

/// A person's name as their ORCID record gives it, if it does.
pub fn person_name(person: &Person) -> Option<String> {
    let name = if person.name.trim().is_empty() {
        [person.given_name.trim(), person.family_name.trim()]
            .iter()
            .filter(|s| !s.is_empty())
            .copied()
            .collect::<Vec<_>>()
            .join(" ")
    } else {
        person.name.trim().to_string()
    };
    (!name.is_empty()).then_some(name)
}

// ── what the page shows ──────────────────────────────────────────────────────

pub struct Link {
    pub name: String,
    pub url: String,
}

pub struct AffiliationView {
    pub name: String,
    /// The organization's ROR ID, as a URL.
    pub ror: Option<String>,
    pub role: String,
    pub dates: String,
}

/// The person card: what commonmeta.org shows for an ORCID iD.
pub struct PersonView {
    pub orcid_url: String,
    pub name: String,
    pub given_name: String,
    pub family_name: String,
    pub additional_names: String,
    /// The biography, as one paragraph.
    pub description: String,
    /// The start of the biography, for the page's meta description.
    pub summary: String,
    pub urls: Vec<Link>,
    pub country: String,
    pub affiliations: Vec<AffiliationView>,
    /// The record, for the copy button, as on commonmeta.org.
    pub json: String,
}

impl PersonView {
    pub fn new(orcid: &str, person: &Person, lang: Lang) -> Self {
        let name = person_name(person).unwrap_or_default();
        let description = person
            .description
            .split_whitespace()
            .collect::<Vec<_>>()
            .join(" ");
        let country = person.country.trim();
        PersonView {
            orcid_url: format!("https://orcid.org/{orcid}"),
            name: if name.is_empty() {
                orcid.to_string()
            } else {
                name
            },
            given_name: person.given_name.clone(),
            family_name: person.family_name.clone(),
            additional_names: person.additional_names.join(" · "),
            summary: summary(&description, 200),
            description,
            urls: person
                .urls
                .iter()
                .filter(|u| web_url(&u.url))
                .take(URLS_MAX)
                .map(|u| Link {
                    name: if u.name.is_empty() {
                        u.url.clone()
                    } else {
                        u.name.clone()
                    },
                    url: u.url.clone(),
                })
                .collect(),
            country: countries::name(&country.to_ascii_uppercase(), lang)
                .unwrap_or(country)
                .to_string(),
            affiliations: person
                .affiliations
                .iter()
                .filter(|a| !a.name.is_empty())
                .map(|a| AffiliationView {
                    name: a.name.clone(),
                    ror: a
                        .identifier
                        .starts_with("https://ror.org/")
                        .then(|| a.identifier.clone()),
                    role: a.role.clone(),
                    dates: lang.date_range(&a.start_date, &a.end_date),
                })
                .collect(),
            json: serde_json::to_string_pretty(person).unwrap_or_default(),
        }
    }
}

/// A format works can be exported in, as on commonmeta.org.
pub struct ExportFormat {
    pub value: &'static str,
    /// `None` for the formatted citation, whose name is translated.
    pub label: Option<&'static str>,
    extension: &'static str,
    content_type: &'static str,
}

const fn export_format(
    value: &'static str,
    label: Option<&'static str>,
    extension: &'static str,
    content_type: &'static str,
) -> ExportFormat {
    ExportFormat {
        value,
        label,
        extension,
        content_type,
    }
}

const JSON: &str = "application/json";
const XML: &str = "application/xml";
const TEXT: &str = "text/plain; charset=utf-8";

pub const EXPORT_FORMATS: [ExportFormat; 11] = [
    export_format("bibtex", Some("BibTeX"), "bib", TEXT),
    export_format("commonmeta", Some("Commonmeta"), "json", JSON),
    export_format("crossref", Some("Crossref"), "json", JSON),
    export_format("crossref_xml", Some("Crossref XML"), "xml", XML),
    export_format("csl", Some("CSL-JSON"), "json", JSON),
    export_format("datacite", Some("DataCite JSON"), "json", JSON),
    export_format("datacite_xml", Some("DataCite XML"), "xml", XML),
    export_format(CITATION, None, "txt", TEXT),
    export_format("inveniordm", Some("InvenioRDM"), "json", JSON),
    export_format("ris", Some("RIS"), "ris", TEXT),
    export_format("schemaorg", Some("Schema.org"), "json", JSON),
];

pub const CITATION: &str = "citation";

/// What an export starts on, as on commonmeta.org.
pub const DEFAULT_EXPORT_FORMAT: &str = "commonmeta";

/// The citation styles on offer, and the name of each: commonmeta.org's common
/// ones.
pub const CITATION_STYLES: [(&str, &str); 6] = [
    ("apa", "American Psychological Association 7th edition"),
    (
        "chicago-author-date",
        "Chicago Manual of Style 18th Edition (Author-Date)",
    ),
    (
        "harvard-cite-them-right",
        "Cite Them Right 12th Edition (Harvard)",
    ),
    ("ieee", "IEEE"),
    ("modern-language-association", "MLA Handbook 9th Edition"),
    ("vancouver", "Vancouver"),
];

impl ExportFormat {
    pub fn find(value: &str) -> Option<&'static ExportFormat> {
        EXPORT_FORMATS.iter().find(|f| f.value == value)
    }

    pub fn content_type(&self) -> &'static str {
        self.content_type
    }

    pub fn filename(&self, orcid: &str) -> String {
        format!("{orcid}.{}", self.extension)
    }
}

/// CSL's type for a Zotero item type.
fn csl_type(item_type: &str) -> &'static str {
    match item_type {
        "artwork" => "graphic",
        "audioRecording" => "song",
        "bill" => "bill",
        "blogPost" => "post-weblog",
        "book" => "book",
        "bookSection" => "chapter",
        "case" => "legal_case",
        "computerProgram" => "software",
        "conferencePaper" => "paper-conference",
        "dataset" => "dataset",
        "dictionaryEntry" => "entry-dictionary",
        "email" | "instantMessage" | "letter" => "personal_communication",
        "encyclopediaArticle" => "entry-encyclopedia",
        "film" | "videoRecording" => "motion_picture",
        "forumPost" => "post",
        "hearing" => "hearing",
        "interview" => "interview",
        "journalArticle" => "article-journal",
        "magazineArticle" => "article-magazine",
        "manuscript" => "manuscript",
        "map" => "map",
        "newspaperArticle" => "article-newspaper",
        "patent" => "patent",
        "podcast" | "radioBroadcast" | "tvBroadcast" => "broadcast",
        "preprint" => "article",
        "presentation" => "speech",
        "report" => "report",
        "standard" => "standard",
        "statute" => "legislation",
        "thesis" => "thesis",
        "webpage" => "webpage",
        _ => "document",
    }
}

/// A Zotero item as CSL-JSON, which commonmeta reads. Only what the profile
/// publishes goes in: not the item's tags, notes or collections.
fn csl_item(data: &Value) -> Value {
    let names = |creator_type: &str| -> Vec<Value> {
        data.get("creators")
            .and_then(Value::as_array)
            .into_iter()
            .flatten()
            .filter(|c| text(c, "creatorType") == creator_type)
            .map(|c| match text(c, "name") {
                "" => json!({"family": text(c, "lastName"), "given": text(c, "firstName")}),
                full => json!({"literal": full}),
            })
            .collect()
    };
    let date_parts: Vec<i64> = text(data, "date")
        .split(['-', '/'])
        .take(3)
        .map_while(|part| part.trim().parse().ok())
        .collect();
    let year = text(data, "date")
        .split(|c: char| !c.is_ascii_digit())
        .find(|part| part.len() == 4)
        .and_then(|year| year.parse::<i64>().ok());
    let issued = match (date_parts.first(), year) {
        (Some(y), _) if *y > 999 => date_parts.clone(),
        (_, Some(year)) => vec![year],
        _ => Vec::new(),
    };
    let mut item = json!({
        "id": match text(data, "DOI") {
            "" => text(data, "key").to_string(),
            doi => format!("https://doi.org/{doi}"),
        },
        "type": csl_type(text(data, "itemType")),
        "title": text(data, "title"),
    });
    let fields = [
        ("DOI", text(data, "DOI")),
        ("URL", text(data, "url")),
        (
            "container-title",
            CONTAINER_FIELDS
                .iter()
                .map(|field| text(data, field))
                .find(|value| !value.is_empty())
                .unwrap_or(""),
        ),
        ("publisher", text(data, "publisher")),
        ("abstract", text(data, "abstractNote")),
        ("language", text(data, "language")),
        ("volume", text(data, "volume")),
        ("issue", text(data, "issue")),
        ("page", text(data, "pages")),
        ("ISSN", text(data, "ISSN")),
        ("ISBN", text(data, "ISBN")),
    ];
    for (field, value) in fields {
        if !value.is_empty() {
            item[field] = json!(value);
        }
    }
    for (field, creator_type) in [("author", "author"), ("editor", "editor")] {
        let names = names(creator_type);
        if !names.is_empty() {
            item[field] = json!(names);
        }
    }
    if !issued.is_empty() {
        item["issued"] = json!({"date-parts": [issued]});
    }
    item
}

/// Works in an export format, with commonmeta's writers. Slow for many works:
/// run it off the async threads.
pub fn export(
    works: &[Value],
    format: &ExportFormat,
    style: &str,
    lang: Lang,
) -> Result<Vec<u8>, String> {
    let entities = works
        .iter()
        .map(|work| commonmeta::read("csl", &csl_item(work).to_string()))
        .collect::<Result<Vec<_>, _>>()
        .map_err(|e| e.to_string())?;
    let bytes = commonmeta::write_list_citation(
        &entities,
        format.value,
        Some(style),
        Some(lang.citation_locale()),
    )
    .map_err(|e| e.to_string())?;
    if format.value != CITATION {
        return Ok(bytes);
    }
    // The citation writer renders HTML, one entry per line.
    let entries: Vec<String> = String::from_utf8_lossy(&bytes)
        .lines()
        .map(|line| decode_entities(&plain_text(line)))
        .filter(|line| !line.is_empty())
        .collect();
    Ok(entries.join("\n\n").into_bytes())
}

/// The Creative Commons icons for a Rights field, which is free text: an
/// identifier (`CC-BY-4.0`, `CC BY 4.0`, `CC0`), a name (`Creative Commons
/// Attribution 4.0`) or a URL. None for anything else.
fn cc_icons(rights: &str) -> Vec<&'static str> {
    let rights = rights.to_ascii_uppercase();
    let rights = rights
        .strip_prefix("HTTPS://")
        .or_else(|| rights.strip_prefix("HTTP://"))
        .map_or(rights.as_str(), |url| url.trim_start_matches("WWW."));
    let is_url = rights.starts_with("CREATIVECOMMONS.ORG/");
    if !is_url && !rights.starts_with("CC") && !rights.starts_with("CREATIVE COMMONS") {
        return Vec::new();
    }
    let words: Vec<&str> = rights.split(|c: char| !c.is_ascii_alphanumeric()).collect();
    if !is_url && !matches!(words[0], "CC" | "CC0" | "CREATIVE") {
        return Vec::new();
    }
    let has = |names: &[&str]| words.iter().any(|word| names.contains(word));
    let mut icons = vec!["cc"];
    if has(&["CC0", "ZERO"]) {
        icons.push("zero");
    } else {
        // In the order commonmeta.org has them.
        for (icon, names) in [
            ("by", &["BY", "ATTRIBUTION"]),
            ("sa", &["SA", "SHAREALIKE"]),
            ("nc", &["NC", "NONCOMMERCIAL"]),
            ("nd", &["ND", "NODERIVATIVES"]),
        ] {
            if has(names) {
                icons.push(icon);
            }
        }
    }
    icons
}

/// A Creative Commons license's full name, as its legal code calls it, and the
/// URL of that legal code, from its SPDX ID: `CC-BY-4.0` is the "Creative
/// Commons Attribution 4.0 License" at `…/licenses/by/4.0/legalcode`.
fn cc_license(spdx: &str, lang: Lang) -> Option<(String, String)> {
    if let Some(version) = spdx.strip_prefix("CC0-") {
        let name = match lang {
            Lang::En => format!("CC0 {version} Universal Public Domain Dedication"),
            Lang::De => format!("Public-Domain-Widmung CC0 {version} Universell"),
        };
        let url = format!("https://creativecommons.org/publicdomain/zero/{version}/legalcode");
        return Some((name, url));
    }
    let mut parts = spdx.strip_prefix("CC-")?.split('-');
    let mut elements = Vec::new();
    let version = loop {
        match parts.next()? {
            part @ ("BY" | "NC" | "SA" | "ND") => elements.push(part),
            part if part.chars().all(|c| c.is_ascii_digit() || c == '.') => break part,
            _ => return None,
        }
    };
    let jurisdiction = parts.next();
    if elements.is_empty() || version.is_empty() || parts.next().is_some() {
        return None;
    }
    // Before 4.0, NoDerivatives was NoDerivs.
    let current = version.starts_with('4');
    let names: Vec<&str> = elements
        .iter()
        .map(|element| match (lang, *element) {
            (Lang::En, "BY") => "Attribution",
            (Lang::En, "NC") => "NonCommercial",
            (Lang::En, "SA") => "ShareAlike",
            (Lang::En, _) if current => "NoDerivatives",
            (Lang::En, _) => "NoDerivs",
            (Lang::De, "BY") => "Namensnennung",
            (Lang::De, "NC") => "Nicht kommerziell",
            (Lang::De, "SA") => "Weitergabe unter gleichen Bedingungen",
            (Lang::De, _) if current => "Keine Bearbeitungen",
            (Lang::De, _) => "Keine Bearbeitung",
        })
        .collect();
    let place = jurisdiction
        .map(|code| format!(" {}", countries::name(code, lang).unwrap_or(code)))
        .unwrap_or_default();
    let name = match lang {
        Lang::En => format!(
            "Creative Commons {} {version}{place} License",
            names.join("-")
        ),
        Lang::De => format!(
            "Creative-Commons-Lizenz {} {version}{place}",
            names.join(" – ")
        ),
    };
    let path = jurisdiction
        .map(|code| format!("{}/", code.to_ascii_lowercase()))
        .unwrap_or_default();
    let url = format!(
        "https://creativecommons.org/licenses/{}/{version}/{path}legalcode",
        elements.join("-").to_ascii_lowercase()
    );
    Some((name, url))
}

/// The short name of a Creative Commons license URL:
/// `…/licenses/by-nc/4.0/` → "CC BY-NC 4.0", `…/publicdomain/zero/1.0/` → "CC0 1.0".
fn cc_name(rights: &str) -> Option<String> {
    let (_, path) = rights.split_once("creativecommons.org/")?;
    let mut parts = path.split('/');
    match (parts.next()?, parts.next()?, parts.next()?) {
        ("licenses", kind, version) if !version.is_empty() => {
            Some(format!("CC {} {version}", kind.to_ascii_uppercase()))
        }
        ("publicdomain", "zero", version) if !version.is_empty() => Some(format!("CC0 {version}")),
        _ => None,
    }
}

/// The entities a citation's HTML is escaped with.
fn decode_entities(text: &str) -> String {
    text.replace("&lt;", "<")
        .replace("&gt;", ">")
        .replace("&quot;", "\"")
        .replace("&#39;", "'")
        .replace("&#x27;", "'")
        .replace("&amp;", "&")
}

/// A work in the lists, as on commonmeta.org's work cards.
pub struct WorkView {
    /// For copying it, in the format picked for exports.
    pub key: String,
    pub title: String,
    /// The DOI as a URL, else the item's URL.
    pub link: Option<String>,
    pub has_doi: bool,
    pub authors: String,
    /// When it was published, for people.
    pub date: String,
    pub container: String,
    pub kind: String,
    /// The name of the work's language.
    pub language: String,
    /// The abstract, without markup.
    pub description: String,
    /// A PDF to read it in, where it was published.
    pub pdf: Option<String>,
    /// The Rights field: what it says, or a Creative Commons license's short
    /// name for its URL, and where it leads when it is a URL.
    pub license: String,
    /// It names a license, rather than saying something else about rights,
    /// such as "All rights reserved".
    pub is_license: bool,
    pub license_url: Option<String>,
    /// The Creative Commons icons for the Rights field, in order.
    pub license_icons: Vec<&'static str>,
}

fn web_url(url: &str) -> bool {
    url.starts_with("https://") || url.starts_with("http://")
}

fn text<'a>(data: &'a Value, field: &str) -> &'a str {
    data.get(field).and_then(Value::as_str).unwrap_or("").trim()
}

/// Where a work appeared: the first container field its item type has.
const CONTAINER_FIELDS: &[&str] = &[
    "publicationTitle",
    "blogTitle",
    "websiteTitle",
    "bookTitle",
    "proceedingsTitle",
    "meetingName",
    "repository",
    "institution",
    "university",
    "publisher",
];

/// Authors listed before "et al.".
const AUTHORS_MAX: usize = 5;

impl WorkView {
    pub fn new(data: &Value, lang: Lang) -> Self {
        let doi = text(data, "DOI");
        let url = text(data, "url");
        let link = (!doi.is_empty())
            .then(|| format!("https://doi.org/{doi}"))
            .or_else(|| web_url(url).then(|| url.to_string()));
        let creators: Vec<&Value> = data
            .get("creators")
            .and_then(Value::as_array)
            .map(|a| a.iter().collect())
            .unwrap_or_default();
        // Authors when there are any, else whoever is credited first.
        let main_type = if creators.iter().any(|c| text(c, "creatorType") == "author") {
            "author"
        } else {
            creators
                .first()
                .map(|c| text(c, "creatorType"))
                .unwrap_or("")
        };
        let names: Vec<String> = creators
            .iter()
            .filter(|c| text(c, "creatorType") == main_type)
            .map(|c| match text(c, "name") {
                "" => format!("{} {}", text(c, "firstName"), text(c, "lastName"))
                    .trim()
                    .to_string(),
                full => full.to_string(),
            })
            .filter(|n| !n.is_empty())
            .collect();
        let mut authors = names
            .iter()
            .take(AUTHORS_MAX)
            .cloned()
            .collect::<Vec<_>>()
            .join(", ");
        if names.len() > AUTHORS_MAX {
            authors.push_str(" et al.");
        }
        let title = match text(data, "title") {
            "" => "Untitled",
            title => title,
        }
        .to_string();
        let date = text(data, "date");
        let container = CONTAINER_FIELDS
            .iter()
            .map(|field| text(data, field))
            .find(|value| !value.is_empty())
            .unwrap_or("")
            .to_string();
        let rights = text(data, "rights");
        WorkView {
            key: text(data, "key").to_string(),
            title,
            link,
            has_doi: !doi.is_empty(),
            authors,
            date: lang.date(date),
            container,
            kind: lang.item_type(text(data, "itemType")),
            language: lang.language(text(data, "language")),
            description: plain_text(text(data, "abstractNote")),
            pdf: None,
            license: cc_name(rights).unwrap_or_else(|| rights.trim_end_matches('.').to_string()),
            is_license: web_url(rights) || !cc_icons(rights).is_empty(),
            license_url: web_url(rights).then(|| rights.to_string()),
            license_icons: cc_icons(rights),
        }
        .with_cc_license(rights, lang)
    }

    /// A Creative Commons license, however `rights` names it, by its full name,
    /// with its icons, linked to its legal code.
    fn with_cc_license(mut self, rights: &str, lang: Lang) -> Self {
        let spdx = crate::vocab::license(rights.trim_end_matches('.'));
        if let Some((name, url)) = spdx.as_deref().and_then(|id| cc_license(id, lang)) {
            self.license = name;
            self.is_license = true;
            self.license_url = Some(url);
            self.license_icons = cc_icons(spdx.as_deref().unwrap_or_default());
        }
        self
    }
}

/// Whether this leads to a copy of the owner's, in Zotero's storage, an object
/// store or this server, rather than to where the work was published.
fn stored_copy(url: &str, public_url: &str) -> bool {
    let host = |url: &str| -> String {
        let rest = url.split_once("://").map_or(url, |(_, rest)| rest);
        let authority = rest.split(['/', '?', '#']).next().unwrap_or("");
        let host = authority.rsplit('@').next().unwrap_or("");
        host.split(':').next().unwrap_or("").to_ascii_lowercase()
    };
    let target = host(url);
    let within = |domain: &str| target == domain || target.ends_with(&format!(".{domain}"));
    target == host(public_url)
        || within("zotero.org")
        || within("your-objectstorage.com")
        || within("amazonaws.com")
        || within("r2.cloudflarestorage.com")
        // A pre-signed URL, from any object store.
        || url.to_ascii_lowercase().contains("x-amz-signature=")
}

/// An endpoint for licensed text mining, which publishers list beside the copy
/// for readers and which does not answer a reader following a link.
fn mining_url(url: &str) -> bool {
    let url = url.to_ascii_lowercase();
    ["//api.", "/tdm/", "/chorus/", "text-and-data-mining"]
        .iter()
        .any(|part| url.contains(part))
}

/// The one PDF of a work to link to, as on commonmeta.org: of its attached PDFs'
/// source URLs, then its own URL if that is a PDF, a copy for readers if there
/// is one, else the first. Never a stored copy.
fn preferred_pdf(own: &str, attached: &[String], public_url: &str) -> Option<String> {
    let path = own.split(['?', '#']).next().unwrap_or("");
    let own_is_pdf = path.to_ascii_lowercase().ends_with(".pdf");
    let pdfs: Vec<&str> = attached
        .iter()
        .map(String::as_str)
        .chain(own_is_pdf.then_some(own))
        .filter(|url| web_url(url) && !stored_copy(url, public_url))
        .collect();
    pdfs.iter()
        .find(|url| !mining_url(url))
        .or(pdfs.first())
        .map(|url| url.to_string())
}

/// The PDF to link to of each public work that has one, by work: what the links
/// and the Content facet both go by.
pub fn work_pdfs(sources: &[store::PdfSources], public_url: &str) -> HashMap<String, String> {
    sources
        .iter()
        .filter_map(|s| {
            let pdf = preferred_pdf(s.url.trim(), &s.attached, public_url)?;
            Some((s.key.clone(), pdf))
        })
        .collect()
}

/// `text` without tags: abstracts can hold JATS or HTML markup. A lone `<`, as
/// in "p < 0.05", stays.
pub(crate) fn plain_text(text: &str) -> String {
    let mut out = String::with_capacity(text.len());
    let mut rest = text;
    while let Some(at) = rest.find('<') {
        let (head, tail) = rest.split_at(at);
        out.push_str(head);
        let opens_tag = tail[1..]
            .chars()
            .next()
            .is_some_and(|c| c.is_ascii_alphabetic() || c == '/');
        match tail.find('>').filter(|_| opens_tag) {
            Some(end) => rest = &tail[end + 1..],
            None => {
                out.push('<');
                rest = &tail[1..];
            }
        }
    }
    out.push_str(rest);
    out.trim().to_string()
}

/// The first `max` characters or so of `text`, cut at a word.
fn summary(text: &str, max: usize) -> String {
    if text.chars().count() <= max {
        return text.to_string();
    }
    let cut: String = text.chars().take(max).collect();
    let cut = cut.rsplit_once(' ').map_or(cut.as_str(), |(head, _)| head);
    format!("{}…", cut.trim_end_matches([',', ';', ':', '.']))
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn normalizes_orcid_ids() {
        assert_eq!(
            normalize_orcid("https://orcid.org/0000-0003-1419-2405").as_deref(),
            Some("0000-0003-1419-2405")
        );
        assert_eq!(
            normalize_orcid("0000-0003-1419-2405").as_deref(),
            Some("0000-0003-1419-2405")
        );
        // Wrong check digit.
        assert_eq!(normalize_orcid("0000-0003-1419-2404"), None);
        assert_eq!(normalize_orcid("1"), None);
    }

    #[test]
    fn a_failed_fetch_is_not_retried_at_once() {
        const ID: &str = "0000-0003-1419-2405";
        let cache = PersonCache::default();
        assert!(cache.settled(ID).is_none(), "nothing fetched yet");
        cache.fetch_failed(ID);
        assert!(matches!(cache.settled(ID), Some(None)));
        cache.set(ID, Person::default());
        assert!(matches!(cache.settled(ID), Some(Some(_))));
        // A failure keeps the record it has.
        cache.fetch_failed(ID);
        assert!(matches!(cache.settled(ID), Some(Some(_))));
        // Each person has their own, and only so many are kept.
        assert!(cache.settled("0000-0002-1825-0097").is_none());
        for n in 0..PEOPLE_MAX + 10 {
            cache.set(&format!("id-{n}"), Person::default());
        }
        assert_eq!(cache.slots().len(), PEOPLE_MAX);
    }

    #[test]
    fn describes_works_as_citations() {
        let w = WorkView::new(
            &json!({
                "itemType": "journalArticle",
                "title": "ORCID: a system to uniquely identify researchers",
                "date": "2012-10-01",
                "DOI": "10.1087/20120404",
                "url": "https://example.org/orcid",
                "publicationTitle": "Learned Publishing",
                "language": "en-US",
                "rights": "https://creativecommons.org/licenses/by/4.0/",
                "creators": [
                    {"firstName": "Laurel", "lastName": "Haak", "creatorType": "author"},
                    {"name": "ORCID", "creatorType": "author"},
                    {"firstName": "Eve", "lastName": "Editor", "creatorType": "editor"},
                ],
            }),
            Lang::En,
        );
        assert_eq!(w.link.as_deref(), Some("https://doi.org/10.1087/20120404"));
        assert_eq!(w.authors, "Laurel Haak, ORCID");
        assert_eq!(w.date, "October 1, 2012");
        assert_eq!(w.container, "Learned Publishing");
        assert_eq!(w.kind, "Journal article");
        assert_eq!(w.language, "English");
        assert!(w.license_url.is_some());

        let many: Vec<Value> = (0..7)
            .map(|i| json!({"lastName": format!("A{i}"), "creatorType": "author"}))
            .collect();
        let w = WorkView::new(
            &json!({
                "itemType": "blogPost", "url": "javascript:alert(1)", "creators": many,
                "rights": "CC BY 4.0.", "tags": [{"tag": "secret"}],
            }),
            Lang::De,
        );
        assert_eq!(w.link, None);
        assert_eq!(w.title, "Untitled");
        assert_eq!(w.authors, "A0, A1, A2, A3, A4 et al.");
        assert_eq!(w.kind, "Blogbeitrag");
        assert_eq!(w.license, "Creative-Commons-Lizenz Namensnennung 4.0");
        assert_eq!(
            w.license_url.as_deref(),
            Some("https://creativecommons.org/licenses/by/4.0/legalcode")
        );
    }

    #[test]
    fn reads_the_person_card() {
        let person: Person = serde_json::from_value(json!({
            "id": "https://orcid.org/0000-0003-1419-2405",
            "given_name": "Martin",
            "family_name": "Fenner",
            "additional_names": ["Martin Hellmut Fenner"],
            "description": "First paragraph.\nSecond paragraph.",
            "affiliations": [
                {"name": "Front Matter", "role": "Founder", "start_date": "2021-08-01"},
                {"identifier": "https://ror.org/04wxnsj81", "identifier_type": "ROR",
                 "name": "DataCite", "start_date": "2015-08", "end_date": "2021-07-31"},
            ],
            "urls": [
                {"name": "Homepage", "url": "https://front-matter.de/mfenner"},
                {"name": "Bad", "url": "javascript:alert(1)"},
            ],
            "country": "DE",
        }))
        .unwrap();
        let view = PersonView::new("0000-0003-1419-2405", &person, Lang::En);
        assert_eq!(view.name, "Martin Fenner");
        assert_eq!(view.description, "First paragraph. Second paragraph.");
        assert_eq!(view.urls.len(), 1);
        assert_eq!(view.country, "Germany");
        assert_eq!(view.affiliations[0].dates, "Started August 1, 2021");
        assert_eq!(view.affiliations[0].ror, None);
        assert_eq!(
            view.affiliations[1].ror.as_deref(),
            Some("https://ror.org/04wxnsj81")
        );
        assert!(view.json.contains("Fenner"));
        let view = PersonView::new("0000-0003-1419-2405", &person, Lang::De);
        assert_eq!(view.country, "Deutschland");
        assert_eq!(view.affiliations[0].dates, "Beginn 1. August 2021");
    }

    #[test]
    fn exports_only_what_the_profile_publishes() {
        let work = json!({
            "key": "ARTICLEA", "itemType": "journalArticle", "title": "A <i>title</i>",
            "date": "2012-10-01", "DOI": "10.1087/20120404", "publicationTitle": "Journal",
            "creators": [
                {"firstName": "Laurel", "lastName": "Haak", "creatorType": "author"},
                {"name": "ORCID", "creatorType": "author"},
            ],
            "tags": [{"tag": "secret tag"}], "collections": ["SECRETCL"],
            "extra": "secret extra",
        });
        let item = csl_item(&work);
        assert_eq!(item["type"], "article-journal");
        assert_eq!(item["issued"]["date-parts"], json!([[2012, 10, 1]]));
        assert_eq!(item["author"][1]["literal"], "ORCID");
        assert_eq!(
            csl_item(&json!({"date": "circa 1990"}))["issued"]["date-parts"],
            json!([[1990]])
        );
        for format in &EXPORT_FORMATS {
            let bytes = export(std::slice::from_ref(&work), format, "apa", Lang::En)
                .unwrap_or_else(|e| panic!("{}: {e}", format.value));
            let output = String::from_utf8_lossy(&bytes);
            assert!(output.contains("20120404"), "{}: {output}", format.value);
            assert!(!output.contains("secret"), "{}: {output}", format.value);
            assert!(!output.contains("SECRETCL"), "{}", format.value);
        }
        let citation = ExportFormat::find(CITATION).unwrap();
        let bytes = export(std::slice::from_ref(&work), citation, "apa", Lang::En).unwrap();
        let output = String::from_utf8_lossy(&bytes);
        assert!(output.contains("Haak"), "{output}");
        assert!(!output.contains('<'), "{output}");
    }

    #[test]
    fn reads_creative_commons_licenses() {
        for (rights, icons) in [
            ("CC-BY-4.0", vec!["cc", "by"]),
            ("CC BY 4.0", vec!["cc", "by"]),
            ("cc-by-nc-sa-4.0", vec!["cc", "by", "sa", "nc"]),
            ("CC BY-ND 4.0", vec!["cc", "by", "nd"]),
            ("CC0", vec!["cc", "zero"]),
            ("CC0 1.0", vec!["cc", "zero"]),
            (
                "https://creativecommons.org/licenses/by/4.0/legalcode",
                vec!["cc", "by"],
            ),
            (
                "https://www.creativecommons.org/licenses/by-nc/4.0/",
                vec!["cc", "by", "nc"],
            ),
            (
                "https://creativecommons.org/publicdomain/zero/1.0/",
                vec!["cc", "zero"],
            ),
            (
                "Creative Commons Attribution 4.0 International",
                vec!["cc", "by"],
            ),
            ("", vec![]),
            ("All rights reserved", vec![]),
            ("MIT", vec![]),
            // Not licenses, though they start alike.
            ("CCC permission required", vec![]),
            ("https://example.org/licenses/by/4.0/", vec![]),
        ] {
            assert_eq!(cc_icons(rights), icons, "{rights}");
        }
    }

    #[test]
    fn names_licenses_and_other_rights() {
        let view = |rights: &str| WorkView::new(&json!({"rights": rights}), Lang::En);
        for (rights, name, url, icons) in [
            (
                "https://creativecommons.org/licenses/by-nc-sa/4.0/deed.en",
                "Creative Commons Attribution-NonCommercial-ShareAlike 4.0 License",
                "https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode",
                vec!["cc", "by", "sa", "nc"],
            ),
            (
                "CC BY 4.0",
                "Creative Commons Attribution 4.0 License",
                "https://creativecommons.org/licenses/by/4.0/legalcode",
                vec!["cc", "by"],
            ),
            (
                "Attribution 4.0 International (CC BY 4.0)",
                "Creative Commons Attribution 4.0 License",
                "https://creativecommons.org/licenses/by/4.0/legalcode",
                vec!["cc", "by"],
            ),
            (
                "CC BY-ND 3.0",
                "Creative Commons Attribution-NoDerivs 3.0 License",
                "https://creativecommons.org/licenses/by-nd/3.0/legalcode",
                vec!["cc", "by", "nd"],
            ),
            (
                "https://creativecommons.org/licenses/by/3.0/de/",
                "Creative Commons Attribution 3.0 Germany License",
                "https://creativecommons.org/licenses/by/3.0/de/legalcode",
                vec!["cc", "by"],
            ),
            (
                "https://creativecommons.org/publicdomain/zero/1.0/",
                "CC0 1.0 Universal Public Domain Dedication",
                "https://creativecommons.org/publicdomain/zero/1.0/legalcode",
                vec!["cc", "zero"],
            ),
        ] {
            let w = view(rights);
            assert_eq!(w.license, name, "{rights}");
            assert_eq!(w.license_url.as_deref(), Some(url), "{rights}");
            assert_eq!(w.license_icons, icons, "{rights}");
            assert!(w.is_license);
        }
        let w = WorkView::new(&json!({"rights": "CC BY-NC-ND 4.0"}), Lang::De);
        assert_eq!(
            w.license,
            "Creative-Commons-Lizenz Namensnennung – Nicht kommerziell – Keine Bearbeitungen 4.0"
        );
        let w = view("https://example.org/terms");
        assert_eq!(w.license, "https://example.org/terms");
        assert!(w.is_license);
        let w = view("All rights reserved.");
        assert_eq!(w.license, "All rights reserved");
        assert!(!w.is_license);
    }

    #[test]
    fn links_only_to_where_a_pdf_was_published() {
        const PUBLIC: &str = "https://defter.example.org";
        let urls = |urls: &[&str]| urls.iter().map(|u| u.to_string()).collect::<Vec<_>>();
        let pick = |own: &str, attached: &[&str]| preferred_pdf(own, &urls(attached), PUBLIC);
        const READER: &str = "https://onlinelibrary.wiley.com/doi/pdf/10.1/a";
        const MINING: &str = "https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1";
        assert_eq!(pick("", &[READER]).as_deref(), Some(READER));
        // A copy for readers over a mining endpoint, which is the last resort.
        assert_eq!(pick("", &[MINING, READER]).as_deref(), Some(READER));
        assert_eq!(pick("", &[MINING]).as_deref(), Some(MINING));
        // The work's own URL, when it is a PDF.
        assert_eq!(
            pick("https://example.org/a.PDF?download=1", &[]).as_deref(),
            Some("https://example.org/a.PDF?download=1")
        );
        assert_eq!(pick("https://example.org/a", &[]), None);
        assert_eq!(pick("", &[]), None);
        // Never a stored copy, or something that is not a web link.
        for stored in [
            "https://defter.example.org/users/1/items/ABCDEFGH/file",
            "https://www.zotero.org/users/1/items/ABCDEFGH/file/view",
            "https://api.zotero.org/users/1/items/ABCDEFGH/file",
            "https://zotero.fsn1.your-objectstorage.com/ab/cd.pdf",
            "https://bucket.s3.amazonaws.com/ab/cd.pdf",
            "https://x.r2.cloudflarestorage.com/ab/cd.pdf",
            "https://store.example.org/ab.pdf?X-Amz-Signature=abc",
            "javascript:alert(1)//.pdf",
            "file:///Users/me/a.pdf",
        ] {
            assert_eq!(pick("", &[stored]), None, "{stored}");
            assert_eq!(pick(stored, &[]), None, "{stored}");
        }
        // Lookalike hosts are not stored copies.
        assert_eq!(
            pick("", &["https://notzotero.org/a.pdf"]).as_deref(),
            Some("https://notzotero.org/a.pdf")
        );
    }

    #[test]
    fn abstracts_lose_their_markup() {
        assert_eq!(
            plain_text("<jats:p>Results: p < 0.05 for <i>all</i></jats:p>"),
            "Results: p < 0.05 for all"
        );
        assert_eq!(plain_text("unclosed <b"), "unclosed <b");
    }

    #[test]
    fn summaries_end_at_a_word() {
        assert_eq!(summary("short", 200), "short");
        assert_eq!(summary("one two three four", 12), "one two…");
    }
}