alpenwerk-hr/lib/reports.ts

import { fmtName, todayIso, yearsBetweenIso } from "./format";
import type { EmploymentStatus, HistoryEventType, Weekday } from "./types";

export { todayIso };

// ── Bestand (point-in-time snapshot) ──────────────────────────────

export type Measure = "headcount" | "fte" | "parttime_rate" | "avg_age" | "avg_tenure" | "female_share" | "avg_dependents";

export type GroupDimension =
  | "division"
  | "department"
  | "team"
  | "location"
  | "status"
  | "employment_type"
  | "contract_type"
  | "entry_year"
  | "source"
  | "paygrade"
  | "worker_type"
  | "collective_agreement"
  | "betriebsrat"
  | "dienstwagen"
  | "laterale_fuehrung"
  | "c_level"
  | "has_dependents"
  | "teilzeit_art"
  | "weekday";

export const MEASURE_LABELS: Record<Measure, string> = {
  headcount: "Headcount",
  fte: "FTE",
  parttime_rate: "Teilzeitquote",
  avg_age: "Ø Alter",
  avg_tenure: "Ø Zugehörigkeit",
  female_share: "Frauenanteil",
  avg_dependents: "Ø Angehörige",
};

export const GROUP_LABELS: Record<GroupDimension, string> = {
  division: "Bereich",
  department: "Abteilung",
  team: "Team",
  location: "Standort",
  status: "Status",
  employment_type: "Beschäftigung",
  contract_type: "Vertragsart",
  entry_year: "Eintrittsjahr",
  source: "Intern/Extern",
  paygrade: "Paygrade",
  worker_type: "Beschäftigtengruppe",
  collective_agreement: "Kollektivvertrag",
  betriebsrat: "Betriebsrat",
  dienstwagen: "Dienstwagen",
  laterale_fuehrung: "Laterale Führung",
  c_level: "C-Level",
  has_dependents: "Hat Angehörige",
  teilzeit_art: "Teilzeitvariante",
  weekday: "Wochentag",
};

export const AVERAGE_MEASURES: Measure[] = ["parttime_rate", "avg_age", "avg_tenure", "female_share", "avg_dependents"];
const SUM_MEASURES: Measure[] = ["headcount", "fte"];

export const STATUS_OPTIONS: EmploymentStatus[] = ["Aktiv", "Karenz", "Geplant", "Ausgetreten"];
export const DEFAULT_STATUSES: EmploymentStatus[] = ["Aktiv", "Karenz"];

// `status` filters travel through the URL/exports as a comma-joined list
// (e.g. "Aktiv,Karenz"); this is the one place that turns that string back
// into a validated status set, defaulting to Aktiv+Karenz when unset — used
// by both the Bestand pivot and the full data export so they can never
// silently disagree on which statuses "no filter" means.
export function parseStatuses(status: string | undefined): EmploymentStatus[] {
  if (!status) return DEFAULT_STATUSES;
  const requested = status.split(",");
  const valid = STATUS_OPTIONS.filter((s) => requested.includes(s));
  return valid.length > 0 ? valid : DEFAULT_STATUSES;
}

export type ReportEmployee = {
  id: string;
  first_name: string;
  last_name: string;
  job_title: string;
  /** Die Einheit der Planstelle; null, wenn zum Stichtag keine besetzt war. */
  org_unit_id: string | null;
  location_id: string;
  status: string;
  employment_type: string;
  contract_type: string;
  entry_date: string;
  exit_date: string | null;
  weekly_hours: number;
  source: string;
  paygrade: string;
  birth_date: string;
  gender: string;
  worker_type: string;
  collective_agreement: string;
  work_days: Weekday[];
  is_betriebsrat: boolean;
  has_dienstwagen: boolean;
  is_laterale_fuehrung: boolean;
  is_c_level: boolean;
  teilzeit_art: string | null;
  dependents_count: number;
};

// Alle drei sind über die *Einheit* der Planstelle geschlüsselt, nicht über
// drei verschiedene Fremdschlüssel: welcher Bereich, welche Abteilung und
// welches Team zu einer Einheit gehören, ergibt sich aus ihrer Vorfahrenkette
// und wird einmal vorberechnet.
/**
 * Eine Organisationseinheit als Filterwert — jede Ebene, nicht nur die
 * Bereiche. `depth` dient der Einrückung in der Auswahlliste.
 */
export type UnitOption = { id: string; name: string; depth: number; unitType: string };

export type OrgLookups = {
  divisionName: Map<string, string>;
  departmentName: Map<string, string>;
  teamName: Map<string, string>;
  locationName: Map<string, string>;
};

// Reconstructs status as of any date from the columns that actually carry a
// timeline (entry/exit/Karenz), rather than trusting `employees.status`,
// which only ever reflects *today*.
//
// Die Einordnung in die Organisation wird zum selben Stichtag aufgelöst: seit
// dem OM-Modell ist position_assignments zeitabhängig, eine Auswertung
// gruppiert also nach der Einheit von damals. Vorher gab es diese Historie
// nicht, und ein Stichtagsbericht gruppierte nach der heutigen Zuordnung —
// was in der Oberfläche vermerkt werden musste, statt still falsch zu sein.
export function deriveStatusAsOf(
  e: { entry_date: string; exit_date: string | null; karenz_start_date: string | null; karenz_return_date: string | null },
  asOf: string
): EmploymentStatus {
  if (e.entry_date > asOf) return "Geplant";
  if (e.exit_date && e.exit_date <= asOf) return "Ausgetreten";
  if (e.karenz_start_date && e.karenz_start_date <= asOf && (!e.karenz_return_date || asOf < e.karenz_return_date)) return "Karenz";
  return "Aktiv";
}

function tenureYearsAsOf(entryDate: string, exitDate: string | null, asOf: string): number {
  const start = Date.parse(`${entryDate}T00:00:00Z`);
  const end = Date.parse(`${exitDate && exitDate <= asOf ? exitDate : asOf}T00:00:00Z`);
  return Math.max(0, (end - start) / (1000 * 60 * 60 * 24 * 365.25));
}

export function groupKeyFor(e: ReportEmployee, dim: GroupDimension, lookups: OrgLookups): string {
  switch (dim) {
    case "division":
      return e.org_unit_id ? (lookups.divisionName.get(e.org_unit_id) ?? "Unbekannt") : "–";
    case "department":
      return e.org_unit_id ? (lookups.departmentName.get(e.org_unit_id) ?? "–") : "–";
    case "team":
      return e.org_unit_id ? (lookups.teamName.get(e.org_unit_id) ?? "–") : "–";
    case "location":
      return lookups.locationName.get(e.location_id) ?? "Unbekannt";
    case "status":
      return e.status;
    case "employment_type":
      return e.employment_type;
    case "contract_type":
      return e.contract_type;
    case "entry_year":
      return e.entry_date.slice(0, 4);
    case "source":
      return e.source;
    case "paygrade":
      return e.paygrade;
    case "worker_type":
      return e.worker_type;
    case "collective_agreement":
      return e.collective_agreement;
    case "betriebsrat":
      return e.is_betriebsrat ? "Ja" : "Nein";
    case "dienstwagen":
      return e.has_dienstwagen ? "Ja" : "Nein";
    case "laterale_fuehrung":
      return e.is_laterale_fuehrung ? "Ja" : "Nein";
    case "c_level":
      return e.is_c_level ? "Ja" : "Nein";
    case "has_dependents":
      return e.dependents_count > 0 ? "Ja" : "Nein";
    case "teilzeit_art":
      // „Keine" statt „–": die Antwort ist hier eine Aussage, kein Fehlen.
      return e.teilzeit_art ?? "Keine";
    case "weekday":
      // Not a strict partition — see groupKeysFor, which aggregateReport
      // actually uses. This single-key fallback only covers a direct
      // groupKeyFor("weekday", ...) call from outside aggregateReport.
      return e.work_days[0] ?? "–";
    default:
      return "Unbekannt";
  }
}

const WEEKDAY_ORDER: Weekday[] = ["Mo", "Di", "Mi", "Do", "Fr", "Sa", "So"];

function weekdayRank(key: string): number {
  const i = WEEKDAY_ORDER.indexOf(key as Weekday);
  return i === -1 ? WEEKDAY_ORDER.length : i;
}

function sortByWeekday<T extends { key: string }>(items: T[]): T[] {
  return [...items].sort((a, b) => weekdayRank(a.key) - weekdayRank(b.key));
}

// Reused by ReportsPageClient (split legend) and the report export route
// (split columns) to render a `weekday` split chronologically rather than
// in first-encountered order; a no-op for every other dimension.
export function sortKeysForDimension(keys: string[], dim: GroupDimension): string[] {
  return dim === "weekday" ? [...keys].sort((a, b) => weekdayRank(a) - weekdayRank(b)) : keys;
}

// Every dimension other than `weekday` is a strict single-key partition
// (delegates to groupKeyFor); `weekday` returns one key per work day, so an
// employee is counted in every day they work — deliberately not a
// partition, since that's the whole point of the dimension.
export function groupKeysFor(e: ReportEmployee, dim: GroupDimension, lookups: OrgLookups): string[] {
  if (dim === "weekday") return e.work_days.length > 0 ? e.work_days : ["–"];
  return [groupKeyFor(e, dim, lookups)];
}

export function measureValue(rows: ReportEmployee[], measure: Measure, asOf: string = todayIso()): number {
  if (rows.length === 0) return 0;
  switch (measure) {
    case "headcount":
      return rows.length;
    case "fte":
      return rows.reduce((s, e) => s + e.weekly_hours / 38.5, 0);
    case "parttime_rate":
      return (rows.filter((e) => e.employment_type === "Teilzeit").length / rows.length) * 100;
    case "avg_age":
      return rows.reduce((s, e) => s + yearsBetweenIso(e.birth_date, asOf), 0) / rows.length;
    case "avg_tenure":
      return rows.reduce((s, e) => s + tenureYearsAsOf(e.entry_date, e.exit_date, asOf), 0) / rows.length;
    case "female_share":
      return (rows.filter((e) => e.gender === "w").length / rows.length) * 100;
    case "avg_dependents":
      return rows.reduce((s, e) => s + e.dependents_count, 0) / rows.length;
    default:
      return 0;
  }
}

export type ReportPerson = { id: string; name: string; title: string; team: string; entry_date: string };
export type ReportSplitRow = { key: string; value: number; count: number };
export type ReportRow = { key: string; value: number; count: number; people: ReportPerson[]; split?: ReportSplitRow[] };

export function aggregateReport(
  employees: ReportEmployee[],
  measure: Measure,
  group: GroupDimension,
  split: GroupDimension | null,
  lookups: OrgLookups,
  asOf: string = todayIso()
): ReportRow[] {
  const byGroup = new Map<string, ReportEmployee[]>();
  for (const e of employees) {
    for (const key of groupKeysFor(e, group, lookups)) {
      if (!byGroup.has(key)) byGroup.set(key, []);
      byGroup.get(key)!.push(e);
    }
  }

  const rows: ReportRow[] = [];
  for (const [key, rowsForGroup] of byGroup) {
    const value = measureValue(rowsForGroup, measure, asOf);
    const people: ReportPerson[] = rowsForGroup.map((e) => ({
      id: e.id,
      name: fmtName(e.first_name, e.last_name),
      title: e.job_title,
      team: e.org_unit_id ? (lookups.teamName.get(e.org_unit_id) ?? "–") : "–",
      entry_date: e.entry_date,
    }));
    const row: ReportRow = { key, value, count: rowsForGroup.length, people };
    if (split) {
      const bySplit = new Map<string, ReportEmployee[]>();
      for (const e of rowsForGroup) {
        for (const sKey of groupKeysFor(e, split, lookups)) {
          if (!bySplit.has(sKey)) bySplit.set(sKey, []);
          bySplit.get(sKey)!.push(e);
        }
      }
      const splitRows = Array.from(bySplit.entries()).map(([sKey, sRows]) => ({
        key: sKey,
        value: measureValue(sRows, measure, asOf),
        count: sRows.length,
      }));
      row.split = split === "weekday" ? sortByWeekday(splitRows) : splitRows;
    }
    rows.push(row);
  }
  return group === "weekday" ? sortByWeekday(rows) : rows.sort((a, b) => b.value - a.value);
}

export function sumValues(rows: { value: number }[]): number {
  return rows.reduce((s, r) => s + r.value, 0);
}

// headcount/fte sum across groups; averages/ratios are weighted by each
// group's underlying record count for a sensible overall figure.
export function totalForRows(rows: { value: number; count: number }[], measure: Measure): number {
  if (SUM_MEASURES.includes(measure)) return sumValues(rows);
  const totalCount = rows.reduce((s, r) => s + r.count, 0);
  if (totalCount === 0) return 0;
  return rows.reduce((s, r) => s + r.value * r.count, 0) / totalCount;
}

export const REPORT_PRESETS: { name: string; measure: Measure; group: GroupDimension; split?: GroupDimension }[] = [
  { name: "Headcount nach Bereich", measure: "headcount", group: "division" },
  { name: "Frauenanteil nach Bereich", measure: "female_share", group: "division" },
  { name: "Headcount nach Paygrade", measure: "headcount", group: "paygrade" },
  { name: "Teilzeitquote nach Standort", measure: "parttime_rate", group: "location" },
  { name: "Headcount nach Wochentag", measure: "headcount", group: "weekday" },
  { name: "Headcount nach C-Level", measure: "headcount", group: "c_level" },
  { name: "Ø Angehörige nach Bereich", measure: "avg_dependents", group: "division" },
];

// ── Ereignisse (events over a period) ─────────────────────────────
// Backed by employee_history, the append-only log — unlike Bestand, this
// covers every event type (not just Eintritt/Austritt), survives an
// employee's entry_date being overwritten by a later rehire, and each event
// keeps its own date/description regardless of the employee's current state.

// Sentinel for "von"/"bis" — distinct from "" (unset, falls back to the
// current-year default) or a real date. Written to the URL/exports as the
// literal string "open".
export const EVENT_DATE_OPEN = "open";

export type EventGroupDimension = "event_type" | "division" | "department" | "team" | "location" | "event_year";

export const EVENT_GROUP_LABELS: Record<EventGroupDimension, string> = {
  event_type: "Ereignistyp",
  division: "Bereich",
  department: "Abteilung",
  team: "Team",
  location: "Standort",
  event_year: "Jahr",
};

export const EVENT_TYPE_LABELS: Record<HistoryEventType, string> = {
  Eintritt: "Eintritt",
  Beförderung: "Beförderung",
  Versetzung: "Versetzung",
  // Stored value stays 'Karenz'; the label follows the renamed concept.
  Karenz: "Langzeitabwesenheit",
  Vertragsänderung: "Vertragsänderung",
  Stammdatenänderung: "Stammdatenänderung",
  Austritt: "Austritt",
  Wiedereintritt: "Wiedereintritt",
  Reorganisation: "Reorganisation",
  Gehaltsanpassung: "Gehaltsanpassung",
  Rückkehr: "Rückkehr aus Langzeitabwesenheit",
};

export type ReportEvent = {
  employee_id: string;
  first_name: string;
  last_name: string;
  job_title: string;
  /** Die Einheit der Planstelle; null, wenn zum Stichtag keine besetzt war. */
  org_unit_id: string | null;
  location_id: string;
  event_date: string;
  event_type: HistoryEventType;
  description: string;
};

function eventGroupKeyFor(e: ReportEvent, dim: EventGroupDimension, lookups: OrgLookups): string {
  switch (dim) {
    case "event_type":
      return EVENT_TYPE_LABELS[e.event_type] ?? e.event_type;
    case "division":
      return e.org_unit_id ? (lookups.divisionName.get(e.org_unit_id) ?? "Unbekannt") : "–";
    case "department":
      return e.org_unit_id ? (lookups.departmentName.get(e.org_unit_id) ?? "–") : "–";
    case "team":
      return e.org_unit_id ? (lookups.teamName.get(e.org_unit_id) ?? "–") : "–";
    case "location":
      return lookups.locationName.get(e.location_id) ?? "Unbekannt";
    case "event_year":
      return e.event_date.slice(0, 4);
    default:
      return "Unbekannt";
  }
}

export function aggregateEvents(
  events: ReportEvent[],
  group: EventGroupDimension,
  split: EventGroupDimension | null,
  lookups: OrgLookups
): ReportRow[] {
  const byGroup = new Map<string, ReportEvent[]>();
  for (const e of events) {
    const key = eventGroupKeyFor(e, group, lookups);
    if (!byGroup.has(key)) byGroup.set(key, []);
    byGroup.get(key)!.push(e);
  }

  const rows: ReportRow[] = [];
  for (const [key, rowsForGroup] of byGroup) {
    // Repurposes ReportPerson for events: title -> event description,
    // entry_date -> event_date. Keeps the existing drill-down UI/export
    // code working unchanged for both report modes.
    const people: ReportPerson[] = rowsForGroup.map((e) => ({
      id: e.employee_id,
      name: fmtName(e.first_name, e.last_name),
      title: e.description,
      team: e.org_unit_id ? (lookups.teamName.get(e.org_unit_id) ?? "–") : "–",
      entry_date: e.event_date,
    }));
    const row: ReportRow = { key, value: rowsForGroup.length, count: rowsForGroup.length, people };
    if (split) {
      const bySplit = new Map<string, ReportEvent[]>();
      for (const e of rowsForGroup) {
        const sKey = eventGroupKeyFor(e, split, lookups);
        if (!bySplit.has(sKey)) bySplit.set(sKey, []);
        bySplit.get(sKey)!.push(e);
      }
      row.split = Array.from(bySplit.entries()).map(([sKey, sRows]) => ({ key: sKey, value: sRows.length, count: sRows.length }));
    }
    rows.push(row);
  }
  return rows.sort((a, b) => b.value - a.value);
}

export const EVENT_REPORT_PRESETS: { name: string; group: EventGroupDimension; split?: EventGroupDimension; eventType?: HistoryEventType }[] = [
  { name: "Ereignisse nach Typ", group: "event_type" },
  { name: "Eintritte nach Bereich", group: "division", eventType: "Eintritt" },
  { name: "Austritte nach Abteilung", group: "department", eventType: "Austritt" },
  { name: "Beförderungen nach Bereich", group: "division", eventType: "Beförderung" },
];

// ── Query-string parsing ──────────────────────────────────────────
// The Berichte page and every /api/export/* route read the same handful of
// dimension/measure names out of a URL the user fully controls. These used
// to be unchecked `as` casts, which let an unknown value through as a real
// enum value: it reached GROUP_LABELS[group] as undefined (a literal
// "undefined" column header in the export) and was interpolated into the
// download filename, i.e. into a Content-Disposition header. Parsing against
// the label maps — the same objects that define the legal values — keeps the
// two in step by construction.

function parseKeyOf<T extends string>(labels: Record<T, string>, value: string | null | undefined, fallback: T): T {
  return value && Object.hasOwn(labels, value) ? (value as T) : fallback;
}

export function parseMode(value: string | null | undefined): "snapshot" | "events" {
  return value === "events" ? "events" : "snapshot";
}

export function parseMeasure(value: string | null | undefined): Measure {
  return parseKeyOf(MEASURE_LABELS, value, "headcount");
}

export function parseGroupDimension(value: string | null | undefined, fallback: GroupDimension = "division"): GroupDimension {
  return parseKeyOf(GROUP_LABELS, value, fallback);
}

export function parseEventGroupDimension(
  value: string | null | undefined,
  fallback: EventGroupDimension = "event_type"
): EventGroupDimension {
  return parseKeyOf(EVENT_GROUP_LABELS, value, fallback);
}

// Unlike the dimensions above, "no split" and "all event types" are legal —
// hence null rather than a fallback value for an unrecognized input.
export function parseSplitDimension(value: string | null | undefined): GroupDimension | null {
  return value && Object.hasOwn(GROUP_LABELS, value) ? (value as GroupDimension) : null;
}

export function parseEventSplitDimension(value: string | null | undefined): EventGroupDimension | null {
  return value && Object.hasOwn(EVENT_GROUP_LABELS, value) ? (value as EventGroupDimension) : null;
}

export function parseEventType(value: string | null | undefined): HistoryEventType | null {
  return value && Object.hasOwn(EVENT_TYPE_LABELS, value) ? (value as HistoryEventType) : null;
}

// Rejects anything that is not a real calendar date, so a Stichtag from the
// URL can never reach a date comparison (or a column header) as free text.
export function parseIsoDateParam(value: string | null | undefined): string | undefined {
  if (!value || !/^\d{4}-\d{2}-\d{2}$/.test(value)) return undefined;
  const d = new Date(`${value}T00:00:00Z`);
  return Number.isNaN(d.getTime()) || d.toISOString().slice(0, 10) !== value ? undefined : value;
}

// from/to additionally accept the EVENT_DATE_OPEN sentinel ("this side of
// the interval is intentionally unbounded"), which is not a date.
export function parseEventDateParam(value: string | null | undefined): string | undefined {
  return value === EVENT_DATE_OPEN ? EVENT_DATE_OPEN : parseIsoDateParam(value);
}