weave_content/

entity.rs

use std::fmt;

use crate::parser::{ParseError, SectionKind};

/// Maximum entities per file.
const MAX_ENTITIES_PER_FILE: usize = 50;

/// Maximum length of an entity name.
const MAX_NAME_LEN: usize = 300;

/// Label derived from the section an entity appears in.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Label {
    Person,
    Organization,
    Event,
    Document,
    Asset,
}

impl fmt::Display for Label {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        match self {
            Self::Person => write!(f, "person"),
            Self::Organization => write!(f, "organization"),
            Self::Event => write!(f, "event"),
            Self::Document => write!(f, "document"),
            Self::Asset => write!(f, "asset"),
        }
    }
}

impl Label {
    pub fn from_section(kind: SectionKind) -> Option<Self> {
        match kind {
            SectionKind::People => Some(Self::Person),
            SectionKind::Organizations => Some(Self::Organization),
            SectionKind::Events => Some(Self::Event),
            SectionKind::Documents => Some(Self::Document),
            SectionKind::Assets => Some(Self::Asset),
            _ => None,
        }
    }
}

/// A parsed entity with its name, label, and field map.
#[derive(Debug, Clone)]
pub struct Entity {
    pub name: String,
    pub label: Label,
    pub fields: Vec<(String, FieldValue)>,
    /// Stored NULID from `id:` field (None if not yet generated).
    pub id: Option<String>,
    /// Line number (1-indexed) of the H3 heading.
    pub line: usize,
    /// Tags from front matter (empty for inline entities).
    pub tags: Vec<String>,
    /// File-path slug (e.g. `people/id/harvey-moeis`). Only set for
    /// registry entities that have standalone files.
    pub slug: Option<String>,
}

/// A field value: either a single string or a list of strings.
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum FieldValue {
    Single(String),
    List(Vec<String>),
}

/// Parse a single entity from a standalone entity file body.
/// The body is the text after the H1 heading (bullet fields, no H3 headings).
/// `label` is determined by the file's directory (people/ or organizations/).
/// `id` comes from the front matter (may be None).
pub fn parse_entity_file_body(
    name: &str,
    body: &str,
    label: Label,
    id: Option<String>,
    title_line: usize,
    errors: &mut Vec<ParseError>,
) -> Entity {
    let section_kind = match label {
        Label::Person => SectionKind::People,
        Label::Organization => SectionKind::Organizations,
        Label::Event => SectionKind::Events,
        Label::Document => SectionKind::Documents,
        Label::Asset => SectionKind::Assets,
    };

    // Wrap the body with a fake H3 heading so we can reuse parse_entities
    let wrapped = format!("### {name}\n{body}");
    let mut entities = parse_entities(&wrapped, section_kind, title_line.saturating_sub(1), errors);

    if let Some(mut entity) = entities.pop() {
        entity.id = id;
        entity.line = title_line;
        entity
    } else {
        Entity {
            name: name.to_string(),
            label,
            fields: Vec::new(),
            id,
            line: title_line,
            tags: Vec::new(),
            slug: None,
        }
    }
}

/// Parse entities from an entity section (People, Organizations, Events).
/// The `body` is the text between the H2 heading and the next H2 heading.
/// `section_start_line` is the line number of the H2 heading in the original file.
#[allow(clippy::too_many_lines)]
pub fn parse_entities(
    body: &str,
    section_kind: SectionKind,
    section_start_line: usize,
    errors: &mut Vec<ParseError>,
) -> Vec<Entity> {
    let Some(label) = Label::from_section(section_kind) else {
        return Vec::new();
    };

    let lines: Vec<&str> = body.lines().collect();
    let mut entities: Vec<Entity> = Vec::new();
    let mut current_name: Option<String> = None;
    let mut current_line: usize = 0;
    let mut current_fields: Vec<(String, FieldValue)> = Vec::new();
    // Track multi-line value continuation and nested list building
    let mut pending_list_key: Option<String> = None;
    let mut pending_list_items: Vec<String> = Vec::new();

    for (i, line) in lines.iter().enumerate() {
        let file_line = section_start_line + 1 + i; // +1 because body starts after the H2 heading line

        // Check for H3 heading
        if let Some(name) = strip_h3(line) {
            // Flush pending list
            flush_pending_list(
                &mut pending_list_key,
                &mut pending_list_items,
                &mut current_fields,
            );

            // Flush previous entity
            if let Some(entity_name) = current_name.take() {
                let entity = build_entity(
                    entity_name,
                    label,
                    current_line,
                    &mut current_fields,
                    errors,
                );
                entities.push(entity);
            }

            current_name = Some(name.to_string());
            current_line = file_line;
            current_fields.clear();
            continue;
        }

        // Only parse bullet fields if we're inside an entity (after an H3)
        if current_name.is_none() {
            if !line.trim().is_empty() {
                errors.push(ParseError {
                    line: file_line,
                    message: "content before first entity heading (### Name)".into(),
                });
            }
            continue;
        }

        let trimmed = line.trim();

        // Nested list item: `  - value` (2-space indent + dash)
        if let Some(item) = trimmed.strip_prefix("- ") {
            if line.starts_with("  - ") && pending_list_key.is_some() {
                // Nested list item for pending list key
                pending_list_items.push(item.trim().to_string());
                continue;
            }

            // Flush pending list before processing new top-level bullet
            flush_pending_list(
                &mut pending_list_key,
                &mut pending_list_items,
                &mut current_fields,
            );

            // Top-level bullet: `- key: value` or `- key:`
            if let Some((key, value)) = parse_bullet(item) {
                if value.is_empty() {
                    // Start a nested list: `- urls:`
                    pending_list_key = Some(key);
                    pending_list_items.clear();
                } else if is_list_field(&key) && value.contains(',') {
                    // Comma-separated list: `- aliases: A, B, C`
                    let items: Vec<String> = value
                        .split(',')
                        .map(|s| s.trim().to_string())
                        .filter(|s| !s.is_empty())
                        .collect();
                    current_fields.push((key, FieldValue::List(items)));
                } else {
                    current_fields.push((key, FieldValue::Single(value)));
                }
            } else {
                errors.push(ParseError {
                    line: file_line,
                    message: format!(
                        "invalid field syntax: expected `- key: value`, got {trimmed:?}"
                    ),
                });
            }
            continue;
        }

        // Multi-line value continuation (2-space indent, not a bullet)
        if line.starts_with("  ") && !trimmed.is_empty() && !trimmed.starts_with('-') {
            if pending_list_key.is_some() {
                // Could be continuation inside a list context -- treat as error
                errors.push(ParseError {
                    line: file_line,
                    message: "unexpected indented text in list context".into(),
                });
            } else if let Some(last) = current_fields.last_mut() {
                match last.1 {
                    FieldValue::Single(ref mut val) => {
                        val.push('\n');
                        val.push_str(trimmed);
                    }
                    FieldValue::List(ref mut items) => {
                        // Continuation of a comma-separated list field.
                        // Join last item with continuation text, then re-split
                        // in case new commas appear.
                        let tail = items.pop().unwrap_or_default();
                        let joined = if tail.is_empty() {
                            trimmed.to_string()
                        } else {
                            format!("{tail} {trimmed}")
                        };
                        for part in joined.split(',') {
                            let part = part.trim().to_string();
                            if !part.is_empty() {
                                items.push(part);
                            }
                        }
                    }
                }
            }
            continue;
        }

        // Blank line or other content -- ignore
        if !trimmed.is_empty() {
            // Flush pending list on non-indented non-bullet content
            flush_pending_list(
                &mut pending_list_key,
                &mut pending_list_items,
                &mut current_fields,
            );
        }
    }

    // Flush final pending list and entity
    flush_pending_list(
        &mut pending_list_key,
        &mut pending_list_items,
        &mut current_fields,
    );

    if let Some(entity_name) = current_name.take() {
        let entity = build_entity(
            entity_name,
            label,
            current_line,
            &mut current_fields,
            errors,
        );
        entities.push(entity);
    }

    // Boundary check
    if entities.len() > MAX_ENTITIES_PER_FILE {
        errors.push(ParseError {
            line: section_start_line,
            message: format!(
                "too many entities in section (max {MAX_ENTITIES_PER_FILE}, got {})",
                entities.len()
            ),
        });
    }

    entities
}

fn flush_pending_list(
    pending_key: &mut Option<String>,
    pending_items: &mut Vec<String>,
    fields: &mut Vec<(String, FieldValue)>,
) {
    if let Some(key) = pending_key.take() {
        fields.push((key, FieldValue::List(std::mem::take(pending_items))));
    }
}

fn build_entity(
    name: String,
    label: Label,
    line: usize,
    fields: &mut Vec<(String, FieldValue)>,
    errors: &mut Vec<ParseError>,
) -> Entity {
    // Validate name
    if name.trim().is_empty() {
        errors.push(ParseError {
            line,
            message: "entity name must not be empty".into(),
        });
    } else if name.len() > MAX_NAME_LEN {
        errors.push(ParseError {
            line,
            message: format!(
                "entity name exceeds {MAX_NAME_LEN} chars (got {})",
                name.len()
            ),
        });
    }

    // Extract id field before validation (not a schema field)
    let id = extract_id_field(fields);

    // Apply type: shorthand
    apply_type_shorthand(fields, label);

    // Normalize enum field values (lowercase, spaces → underscores)
    normalize_enum_fields(fields);

    // Validate fields against schema
    validate_fields(fields, label, line, errors);

    Entity {
        name,
        label,
        fields: std::mem::take(fields),
        id,
        line,
        tags: Vec::new(),
        slug: None,
    }
}

/// Extract and remove the `id` field from the field list.
fn extract_id_field(fields: &mut Vec<(String, FieldValue)>) -> Option<String> {
    let pos = fields.iter().position(|(k, _)| k == "id")?;
    let (_, value) = fields.remove(pos);
    match value {
        FieldValue::Single(s) if !s.is_empty() => Some(s),
        _ => None,
    }
}

/// Replace `type:` shorthand with the label-specific field name.
fn apply_type_shorthand(fields: &mut [(String, FieldValue)], label: Label) {
    for field in fields.iter_mut() {
        if field.0 == "type" {
            field.0 = match label {
                Label::Organization => "org_type".to_string(),
                Label::Event => "event_type".to_string(),
                Label::Document => "doc_type".to_string(),
                Label::Asset => "asset_type".to_string(),
                Label::Person => "type".to_string(), // will be caught as unknown
            };
        }
    }
}

/// Parse `key: value` from a bullet item (after stripping `- `).
fn parse_bullet(item: &str) -> Option<(String, String)> {
    let colon_pos = item.find(':')?;
    let key = item[..colon_pos].trim();
    if key.is_empty() {
        return None;
    }
    let value = item[colon_pos + 1..].trim();
    Some((key.to_string(), value.to_string()))
}

/// Check if a field name is a list-type field.
fn is_list_field(key: &str) -> bool {
    matches!(key, "aliases" | "urls" | "role")
}

/// Strip an H3 heading prefix. Returns the heading text.
fn strip_h3(line: &str) -> Option<&str> {
    let trimmed = line.trim_start();
    if let Some(rest) = trimmed.strip_prefix("### ") {
        // Must not be H4+
        if !rest.starts_with('#') {
            return Some(rest.trim());
        }
    }
    None
}

// --- Field validation ---

/// Known fields per label (common + label-specific).
const COMMON_FIELDS: &[&str] = &[
    "qualifier",
    "aliases",
    "thumbnail",
    "thumbnail_source",
    "urls",
    "description",
];

const PERSON_FIELDS: &[&str] = &[
    "role",
    "nationality",
    "date_of_birth",
    "place_of_birth",
    "status",
];

const ORGANIZATION_FIELDS: &[&str] = &[
    "org_type",
    "jurisdiction",
    "headquarters",
    "founded_date",
    "registration_number",
    "status",
];

const EVENT_FIELDS: &[&str] = &["event_type", "occurred_at", "jurisdiction", "severity"];

const DOCUMENT_FIELDS: &[&str] = &["doc_type", "issued_at", "issuing_authority", "case_number"];

const ASSET_FIELDS: &[&str] = &["asset_type", "value", "status"];

/// Known enum values — delegated to domain module constants.
use crate::domain;

const ROLE_VALUES: &[&str] = domain::Role::KNOWN;
const ORG_TYPE_VALUES: &[&str] = domain::OrgType::KNOWN;
const EVENT_TYPE_VALUES: &[&str] = domain::EventType::KNOWN;
const DOC_TYPE_VALUES: &[&str] = domain::DocType::KNOWN;
const ASSET_TYPE_VALUES: &[&str] = domain::AssetType::KNOWN;
const SEVERITY_VALUES: &[&str] = domain::Severity::KNOWN;
const PERSON_STATUS_VALUES: &[&str] = domain::PersonStatus::KNOWN;
const ORG_STATUS_VALUES: &[&str] = domain::OrgStatus::KNOWN;
const ASSET_STATUS_VALUES: &[&str] = domain::AssetStatus::KNOWN;

/// Field max lengths.
struct FieldConstraint {
    max_len: usize,
    /// If Some, the field is an enum with these known values.
    enum_values: Option<&'static [&'static str]>,
}

fn field_constraint(key: &str) -> Option<FieldConstraint> {
    match key {
        "description" => Some(FieldConstraint {
            max_len: 2000,
            enum_values: None,
        }),
        "thumbnail" | "thumbnail_source" => Some(FieldConstraint {
            max_len: 2048,
            enum_values: None,
        }),
        "occurred_at" | "date_of_birth" | "founded_date" | "issued_at" | "opened_at"
        | "closed_at" => Some(FieldConstraint {
            max_len: 10,
            enum_values: None,
        }),
        "place_of_birth" | "headquarters" | "issuing_authority" | "value" => {
            Some(FieldConstraint {
                max_len: 200,
                enum_values: None,
            })
        }
        "jurisdiction" => Some(FieldConstraint {
            // jurisdiction: ID or ID/South Sulawesi (country + optional subdivision)
            max_len: 203, // 2 + 1 + 200
            enum_values: None,
        }),
        "role" => Some(FieldConstraint {
            max_len: 100,
            enum_values: Some(ROLE_VALUES),
        }),
        "org_type" => Some(FieldConstraint {
            max_len: 100,
            enum_values: Some(ORG_TYPE_VALUES),
        }),
        "event_type" => Some(FieldConstraint {
            max_len: 100,
            enum_values: Some(EVENT_TYPE_VALUES),
        }),
        "doc_type" => Some(FieldConstraint {
            max_len: 100,
            enum_values: Some(DOC_TYPE_VALUES),
        }),
        "asset_type" => Some(FieldConstraint {
            max_len: 100,
            enum_values: Some(ASSET_TYPE_VALUES),
        }),
        "severity" => Some(FieldConstraint {
            max_len: 20,
            enum_values: Some(SEVERITY_VALUES),
        }),
        "status" => Some(FieldConstraint {
            // Status validation is context-dependent (Person vs Org vs Asset),
            // handled separately in validate_fields.
            max_len: 30,
            enum_values: None,
        }),
        "qualifier" | "nationality" | "case_number" | "registration_number" => {
            Some(FieldConstraint {
                max_len: 100,
                enum_values: None,
            })
        }
        // List fields validated separately
        _ => None,
    }
}

/// Maximum items in list fields.
const MAX_ALIASES: usize = 10;
const MAX_ALIAS_LEN: usize = 200;
const MAX_URLS: usize = 10;
const MAX_URL_LEN: usize = 2048;

/// Normalize enum field values in-place: lowercase and replace spaces with
/// underscores. Values with the `custom:` prefix are left unchanged.
/// Handles both single-value and list-value enum fields.
fn normalize_enum_fields(fields: &mut [(String, FieldValue)]) {
    for (key, value) in fields.iter_mut() {
        let is_enum = field_constraint(key).and_then(|c| c.enum_values).is_some();

        match value {
            FieldValue::Single(val) if is_enum && !val.starts_with("custom:") => {
                let normalized = val.to_lowercase().replace(' ', "_");
                if normalized != *val {
                    *val = normalized;
                }
            }
            FieldValue::List(items) if is_enum => {
                for item in items.iter_mut() {
                    if !item.starts_with("custom:") {
                        let normalized = item.to_lowercase().replace(' ', "_");
                        if normalized != *item {
                            *item = normalized;
                        }
                    }
                }
            }
            _ => {}
        }
    }
}

#[allow(clippy::too_many_lines)]
fn validate_fields(
    fields: &[(String, FieldValue)],
    label: Label,
    line: usize,
    errors: &mut Vec<ParseError>,
) {
    let label_fields: &[&str] = match label {
        Label::Person => PERSON_FIELDS,
        Label::Organization => ORGANIZATION_FIELDS,
        Label::Event => EVENT_FIELDS,
        Label::Document => DOCUMENT_FIELDS,
        Label::Asset => ASSET_FIELDS,
    };

    for (key, value) in fields {
        // Check if field is known
        if !COMMON_FIELDS.contains(&key.as_str()) && !label_fields.contains(&key.as_str()) {
            errors.push(ParseError {
                line,
                message: format!("unknown field {key:?} for {label}"),
            });
            continue;
        }

        match value {
            FieldValue::Single(val) => {
                if let Some(constraint) = field_constraint(key) {
                    if val.len() > constraint.max_len {
                        errors.push(ParseError {
                            line,
                            message: format!(
                                "field {key:?} exceeds {} chars (got {})",
                                constraint.max_len,
                                val.len()
                            ),
                        });
                    }

                    // Validate enum values
                    if let Some(allowed) = constraint.enum_values {
                        validate_enum_value(key, val, allowed, line, errors);
                    }

                    // Validate date format
                    if matches!(
                        key.as_str(),
                        "occurred_at"
                            | "date_of_birth"
                            | "founded_date"
                            | "issued_at"
                            | "opened_at"
                            | "closed_at"
                    ) && !val.is_empty()
                    {
                        validate_date_format(key, val, line, errors);
                    }

                    // Validate URL fields
                    if matches!(key.as_str(), "thumbnail" | "thumbnail_source")
                        && !val.is_empty()
                        && !val.starts_with("https://")
                    {
                        errors.push(ParseError {
                            line,
                            message: format!("field {key:?} must be HTTPS URL"),
                        });
                    }
                }

                // Context-dependent status validation
                if key == "status" {
                    validate_status(val, label, line, errors);
                }

                // Validate jurisdiction format: `XX` or `XX/Subdivision`
                if key == "jurisdiction" && !val.is_empty() {
                    validate_jurisdiction(val, line, errors);
                }

                // Validate money format: `amount currency "display"`
                if key == "value" && !val.is_empty() {
                    validate_money(val, line, errors);
                }
            }
            FieldValue::List(items) => match key.as_str() {
                "aliases" => {
                    if items.len() > MAX_ALIASES {
                        errors.push(ParseError {
                            line,
                            message: format!(
                                "aliases exceeds {MAX_ALIASES} items (got {})",
                                items.len()
                            ),
                        });
                    }
                    for item in items {
                        if item.len() > MAX_ALIAS_LEN {
                            errors.push(ParseError {
                                line,
                                message: format!("alias exceeds {MAX_ALIAS_LEN} chars: {item:?}"),
                            });
                        }
                    }
                }
                "urls" => {
                    if items.len() > MAX_URLS {
                        errors.push(ParseError {
                            line,
                            message: format!("urls exceeds {MAX_URLS} items (got {})", items.len()),
                        });
                    }
                    for item in items {
                        if item.len() > MAX_URL_LEN {
                            errors.push(ParseError {
                                line,
                                message: format!("url exceeds {MAX_URL_LEN} chars: {item:?}"),
                            });
                        }
                        if !item.starts_with("https://") {
                            errors.push(ParseError {
                                line,
                                message: format!("url must be HTTPS: {item:?}"),
                            });
                        }
                    }
                }
                "role" => {
                    if items.len() > MAX_ROLES {
                        errors.push(ParseError {
                            line,
                            message: format!(
                                "role exceeds {MAX_ROLES} items (got {})",
                                items.len()
                            ),
                        });
                    }
                    for item in items {
                        validate_enum_value("role", item, ROLE_VALUES, line, errors);
                    }
                }
                _ => {}
            },
        }
    }

    // Required field checks
    if label == Label::Organization && !fields.iter().any(|(k, _)| k == "org_type") {
        errors.push(ParseError {
            line,
            message: "organization entity missing required field \"org_type\"".into(),
        });
    }
}

/// Maximum roles per person.
const MAX_ROLES: usize = 10;

/// Validate status value based on entity label context.
fn validate_status(value: &str, label: Label, line: usize, errors: &mut Vec<ParseError>) {
    let allowed: &[&str] = match label {
        Label::Person => PERSON_STATUS_VALUES,
        Label::Organization => ORG_STATUS_VALUES,
        Label::Asset => ASSET_STATUS_VALUES,
        _ => {
            errors.push(ParseError {
                line,
                message: format!("field \"status\" is not valid for {label}"),
            });
            return;
        }
    };

    let normalized = value.to_lowercase().replace(' ', "_");
    if !allowed.contains(&normalized.as_str()) {
        errors.push(ParseError {
            line,
            message: format!(
                "invalid status {value:?} for {label} (known: {})",
                allowed.join(", ")
            ),
        });
    }
}

/// Validate jurisdiction format: `XX` or `XX/Subdivision`.
fn validate_jurisdiction(value: &str, line: usize, errors: &mut Vec<ParseError>) {
    if let Some(slash_pos) = value.find('/') {
        let country = &value[..slash_pos];
        let subdivision = &value[slash_pos + 1..];
        if country.len() != 2 || !country.chars().all(|c| c.is_ascii_uppercase()) {
            errors.push(ParseError {
                line,
                message: format!(
                    "jurisdiction country must be 2-letter uppercase ISO code, got {country:?}"
                ),
            });
        }
        if subdivision.is_empty() || subdivision.len() > domain::MAX_SUBDIVISION_LEN {
            errors.push(ParseError {
                line,
                message: format!(
                    "jurisdiction subdivision must be 1-{} chars",
                    domain::MAX_SUBDIVISION_LEN
                ),
            });
        }
    } else {
        // Just country code
        if value.len() != 2 || !value.chars().all(|c| c.is_ascii_uppercase()) {
            errors.push(ParseError {
                line,
                message: format!(
                    "jurisdiction must be 2-letter uppercase ISO code or CODE/Subdivision, got {value:?}"
                ),
            });
        }
    }
}

/// Validate money DSL format: `amount currency "display"`.
/// Example: `500000000000 IDR "Rp 500 billion"`
fn validate_money(value: &str, line: usize, errors: &mut Vec<ParseError>) {
    // Split: amount currency "display"
    let parts: Vec<&str> = value.splitn(3, ' ').collect();
    if parts.len() < 3 {
        errors.push(ParseError {
            line,
            message: format!(
                "invalid money format: expected `amount currency \"display\"`, got {value:?}"
            ),
        });
        return;
    }

    // Validate amount is a valid integer
    if parts[0].parse::<i64>().is_err() {
        errors.push(ParseError {
            line,
            message: format!("money amount must be an integer, got {:?}", parts[0]),
        });
    }

    // Validate currency is 3-letter uppercase
    let currency = parts[1];
    if currency.len() != 3 || !currency.chars().all(|c| c.is_ascii_uppercase()) {
        errors.push(ParseError {
            line,
            message: format!(
                "money currency must be 3-letter uppercase ISO code, got {currency:?}"
            ),
        });
    }

    // Validate display is quoted
    let display = parts[2];
    if !display.starts_with('"') || !display.ends_with('"') {
        errors.push(ParseError {
            line,
            message: format!("money display must be quoted, got {display:?}"),
        });
    } else {
        let inner = &display[1..display.len() - 1];
        if inner.len() > domain::MAX_MONEY_DISPLAY_LEN {
            errors.push(ParseError {
                line,
                message: format!(
                    "money display exceeds {} chars (got {})",
                    domain::MAX_MONEY_DISPLAY_LEN,
                    inner.len()
                ),
            });
        }
    }
}

fn validate_enum_value(
    key: &str,
    value: &str,
    allowed: &[&str],
    line: usize,
    errors: &mut Vec<ParseError>,
) {
    // custom: prefix is always valid (if non-empty after prefix, max 100 chars)
    if let Some(custom) = value.strip_prefix("custom:") {
        if custom.is_empty() || custom.len() > 100 {
            errors.push(ParseError {
                line,
                message: format!(
                    "field {key:?} custom value must be 1-100 chars, got {}",
                    custom.len()
                ),
            });
        }
        return;
    }

    let normalized = value.to_lowercase().replace(' ', "_");
    if !allowed.contains(&normalized.as_str()) {
        errors.push(ParseError {
            line,
            message: format!(
                "invalid {key} value {value:?} (known: {}; use \"custom:Value\" for custom)",
                allowed.join(", ")
            ),
        });
    }
}

fn validate_date_format(key: &str, value: &str, line: usize, errors: &mut Vec<ParseError>) {
    // Valid formats: YYYY, YYYY-MM, YYYY-MM-DD
    let valid = matches!(value.len(), 4 | 7 | 10)
        && value.chars().enumerate().all(|(i, c)| match i {
            4 | 7 => c == '-',
            _ => c.is_ascii_digit(),
        });

    if !valid {
        errors.push(ParseError {
            line,
            message: format!("field {key:?} must be YYYY, YYYY-MM, or YYYY-MM-DD, got {value:?}"),
        });
    }
}

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

    #[test]
    fn parse_person_entity() {
        let body = [
            "",
            "### Mark Bonnick",
            "- qualifier: Arsenal Kit Manager",
            "- nationality: GB",
            "- role: custom:Kit Manager",
            "- date_of_birth: 1962",
            "- description: Academy kit manager at Arsenal FC for 22 years",
            "  (2001-2024). Age 62 at time of dismissal.",
            "",
        ]
        .join("\n");

        let mut errors = Vec::new();
        let entities = parse_entities(&body, SectionKind::People, 10, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        assert_eq!(entities.len(), 1);

        let e = &entities[0];
        assert_eq!(e.name, "Mark Bonnick");
        assert_eq!(e.label, Label::Person);
        assert_eq!(e.fields.len(), 5);

        // Check multi-line description
        let desc = e
            .fields
            .iter()
            .find(|(k, _)| k == "description")
            .map(|(_, v)| v);
        assert_eq!(
            desc,
            Some(&FieldValue::Single(
                "Academy kit manager at Arsenal FC for 22 years\n(2001-2024). Age 62 at time of dismissal.".into()
            ))
        );
    }

    #[test]
    fn parse_person_with_role_list() {
        let body = "### Test\n- role: politician, executive\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        let roles = entities[0].fields.iter().find(|(k, _)| k == "role");
        assert_eq!(
            roles.map(|(_, v)| v),
            Some(&FieldValue::List(vec![
                "politician".into(),
                "executive".into(),
            ]))
        );
    }

    #[test]
    fn parse_person_with_status() {
        let body = "### Test\n- status: imprisoned\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
    }

    #[test]
    fn reject_invalid_person_status() {
        let body = "### Test\n- status: unknown_status\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("invalid status")));
    }

    #[test]
    fn parse_organization_with_type_shorthand() {
        let body = [
            "",
            "### Arsenal FC",
            "- type: sports_club",
            "- jurisdiction: GB",
            "- aliases: Arsenal, The Gunners, Arsenal Football Club",
            "- urls:",
            "  - https://www.arsenal.com",
            "  - https://en.wikipedia.org/wiki/Arsenal_F.C.",
            "",
        ]
        .join("\n");

        let mut errors = Vec::new();
        let entities = parse_entities(&body, SectionKind::Organizations, 20, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        assert_eq!(entities.len(), 1);

        let e = &entities[0];
        assert_eq!(e.name, "Arsenal FC");
        assert_eq!(e.label, Label::Organization);

        // type: should have been expanded to org_type:
        let it = e.fields.iter().find(|(k, _)| k == "org_type");
        assert_eq!(
            it.map(|(_, v)| v),
            Some(&FieldValue::Single("sports_club".into()))
        );

        // aliases as comma-separated
        let aliases = e.fields.iter().find(|(k, _)| k == "aliases");
        assert_eq!(
            aliases.map(|(_, v)| v),
            Some(&FieldValue::List(vec![
                "Arsenal".into(),
                "The Gunners".into(),
                "Arsenal Football Club".into(),
            ]))
        );

        // urls as nested list
        let urls = e.fields.iter().find(|(k, _)| k == "urls");
        assert_eq!(
            urls.map(|(_, v)| v),
            Some(&FieldValue::List(vec![
                "https://www.arsenal.com".into(),
                "https://en.wikipedia.org/wiki/Arsenal_F.C.".into(),
            ]))
        );
    }

    #[test]
    fn parse_organization_with_jurisdiction_subdivision() {
        let body = "### Pemkab Bogor\n- org_type: local_government\n- jurisdiction: ID/West Java\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::Organizations, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        let j = entities[0].fields.iter().find(|(k, _)| k == "jurisdiction");
        assert_eq!(
            j.map(|(_, v)| v),
            Some(&FieldValue::Single("ID/West Java".into()))
        );
    }

    #[test]
    fn reject_invalid_jurisdiction() {
        let body = "### Test\n- org_type: corporation\n- jurisdiction: England\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::Organizations, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("jurisdiction")));
    }

    #[test]
    fn parse_event_with_type_shorthand() {
        let body = [
            "",
            "### Bonnick dismissal",
            "- occurred_at: 2024-12-24",
            "- type: dismissal",
            "- description: Arsenal dismisses Bonnick.",
            "",
        ]
        .join("\n");

        let mut errors = Vec::new();
        let entities = parse_entities(&body, SectionKind::Events, 50, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");

        let e = &entities[0];
        assert_eq!(e.label, Label::Event);
        let dt = e.fields.iter().find(|(k, _)| k == "event_type");
        assert_eq!(
            dt.map(|(_, v)| v),
            Some(&FieldValue::Single("dismissal".into()))
        );
    }

    #[test]
    fn parse_event_with_severity() {
        let body =
            "### Test event\n- event_type: bribery\n- severity: major\n- occurred_at: 2024-01-01\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::Events, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
    }

    #[test]
    fn parse_document_entity() {
        let body = [
            "### Indictment No. 123",
            "- doc_type: indictment",
            "- issued_at: 2024-03-15",
            "- issuing_authority: Jakarta District Court",
            "- case_number: 123/Pid.B/2024/PN.Jkt.Pst",
        ]
        .join("\n");
        let mut errors = Vec::new();
        let entities = parse_entities(&body, SectionKind::Documents, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        assert_eq!(entities.len(), 1);
        assert_eq!(entities[0].label, Label::Document);
    }

    #[test]
    fn parse_asset_entity() {
        let body = "### Bribe payment\n- asset_type: cash\n- value: 500000000000 IDR \"Rp 500 billion\"\n- status: seized\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::Assets, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        assert_eq!(entities.len(), 1);
        assert_eq!(entities[0].label, Label::Asset);
    }

    #[test]
    fn reject_invalid_money_format() {
        let body = "### Test\n- asset_type: cash\n- value: lots of money\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::Assets, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("money")));
    }

    #[test]
    fn reject_unknown_field() {
        let body = "### Test\n- foobar: value\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("unknown field")));
    }

    #[test]
    fn reject_wrong_label_field() {
        // org_type on a person
        let body = "### Test\n- org_type: court\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("unknown field")));
    }

    #[test]
    fn reject_invalid_enum_value() {
        let body = "### Test\n- role: wizard\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("invalid role")));
    }

    #[test]
    fn accept_custom_enum_value() {
        let body = "### Test\n- role: custom:Kit Manager\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        assert_eq!(entities.len(), 1);
    }

    #[test]
    fn normalize_enum_value_spaces_to_underscores() {
        let body = "### Test\n- role: civil servant\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        assert_eq!(entities.len(), 1);
        let val = entities[0]
            .fields
            .iter()
            .find(|(k, _)| k == "role")
            .map(|(_, v)| match v {
                FieldValue::Single(s) => s.as_str(),
                _ => "",
            });
        assert_eq!(val, Some("civil_servant"));
    }

    #[test]
    fn normalize_enum_list_values() {
        let body = "### Test\n- role: civil servant, law enforcement\n";
        let mut errors = Vec::new();
        let entities = parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        let roles = entities[0].fields.iter().find(|(k, _)| k == "role");
        assert_eq!(
            roles.map(|(_, v)| v),
            Some(&FieldValue::List(vec![
                "civil_servant".into(),
                "law_enforcement".into(),
            ]))
        );
    }

    #[test]
    fn reject_invalid_date_format() {
        let body = "### Test\n- date_of_birth: January 1990\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("YYYY")));
    }

    #[test]
    fn accept_valid_date_formats() {
        for date in &["2024", "2024-01", "2024-01-15"] {
            let body = format!("### Test\n- date_of_birth: {date}\n");
            let mut errors = Vec::new();
            parse_entities(&body, SectionKind::People, 1, &mut errors);
            assert!(
                errors.is_empty(),
                "date {date:?} should be valid: {errors:?}"
            );
        }
    }

    #[test]
    fn reject_non_https_url() {
        let body = "### Test\n- urls:\n  - http://example.com\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("HTTPS")));
    }

    #[test]
    fn reject_non_https_thumbnail() {
        let body = "### Test\n- thumbnail: http://example.com/img.jpg\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("HTTPS")));
    }

    #[test]
    fn multiple_entities() {
        let body = [
            "",
            "### Alice",
            "- nationality: NL",
            "",
            "### Bob",
            "- nationality: GB",
            "",
        ]
        .join("\n");

        let mut errors = Vec::new();
        let entities = parse_entities(&body, SectionKind::People, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
        assert_eq!(entities.len(), 2);
        assert_eq!(entities[0].name, "Alice");
        assert_eq!(entities[1].name, "Bob");
    }

    #[test]
    fn field_max_length_violation() {
        let long_val = "a".repeat(201);
        let body = format!("### Test\n- nationality: {long_val}\n");
        let mut errors = Vec::new();
        parse_entities(&body, SectionKind::People, 1, &mut errors);
        assert!(
            errors
                .iter()
                .any(|e| e.message.contains("exceeds 100 chars"))
        );
    }

    #[test]
    fn too_many_aliases() {
        let aliases: Vec<String> = (0..11).map(|i| format!("Alias{i}")).collect();
        let body = format!("### Test\n- aliases: {}\n", aliases.join(", "));
        let mut errors = Vec::new();
        parse_entities(&body, SectionKind::People, 1, &mut errors);
        assert!(errors.iter().any(|e| e.message.contains("exceeds 10")));
    }

    #[test]
    fn require_org_type_for_organizations() {
        let body = "### Test Corp\n- qualifier: Test\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::Organizations, 1, &mut errors);
        assert!(
            errors
                .iter()
                .any(|e| { e.message.contains("missing required field \"org_type\"") })
        );
    }

    #[test]
    fn accept_organization_with_type() {
        let body = "### Test Corp\n- qualifier: Test\n- org_type: corporation\n";
        let mut errors = Vec::new();
        parse_entities(body, SectionKind::Organizations, 1, &mut errors);
        assert!(errors.is_empty(), "errors: {errors:?}");
    }
}