import { SearchType } from "../storage.mjs";
import { IcebergColumn, IcebergPartitionField, IcebergS3Config, IcebergTableSpec, PartitionKeyEncoding, PartitionKeyEncoding as PartitionKeyEncoding$1 } from "@gscdump/lakehouse/schema";
import { TableName } from "@gscdump/contracts";
/** The 9 fact tables that exist as global Iceberg tables. */
type IcebergTableName = Extract<TableName, 'pages' | 'queries' | 'countries' | 'page_queries' | 'dates' | 'search_appearance' | 'search_appearance_pages' | 'search_appearance_queries' | 'search_appearance_page_queries'>;
/** The 9 Iceberg table names, in canonical order. */
declare const ICEBERG_TABLES: readonly IcebergTableName[];
/**
 * Stable `search_type` enum → int map for `'int'`-encoded catalogs. Engine-owned
 * and FROZEN: never renumber or reuse an id (it's the on-disk partition value).
 */
declare const SEARCH_TYPE_INT: Record<SearchType, number>;
/** Reverse of {@link SEARCH_TYPE_INT} — int → `search_type`, for read-result mapping. */
declare const INT_SEARCH_TYPE: Record<number, SearchType>;
/**
 * The two partition-identity columns Iceberg rows carry that the legacy
 * parquet path encoded in the object-key prefix instead. Field ids 1–2 are
 * the first two columns; per-table metric/dimension columns follow
 * contiguously from id 3 (see `ICEBERG_FIELD_ID_BASE`).
 */
declare const ICEBERG_PARTITION_COLUMNS: readonly IcebergColumn[];
/**
 * The two partition-identity columns for a given {@link PartitionKeyEncoding}.
 * `'string'` returns {@link ICEBERG_PARTITION_COLUMNS} verbatim; `'int'` swaps
 * BOTH to INT — `site_id` (the app's small `user_sites.int_id`; ≪ 2.1B sites, so
 * INT is ample) and `search_type` (its fixed enum code). Integer identity columns
 * avoid R2 SQL's truncated-string-stats equality undercount and restore pruning.
 * Field ids are unchanged (1, 2) — only the column types differ.
 */
declare function icebergPartitionColumns(encoding?: PartitionKeyEncoding): readonly IcebergColumn[];
/**
 * First field id used for per-table (non-partition) columns — immediately
 * after the two partition-identity columns (`site_id`=1, `search_type`=2).
 * No longer read by `defineDataset` below (field ids are now assigned by
 * `defineIcebergDataset` itself, identity-then-dims-then-columns) — kept as a
 * documented, byte-identity-tested constant for existing external readers.
 */
declare const ICEBERG_FIELD_ID_BASE = 3;
/** Shared partition spec — identical across every table. */
declare const ICEBERG_PARTITION_SPEC: readonly IcebergPartitionField[];
/**
 * Derive the full Iceberg table spec for a table + encoding. Thin wrapper
 * over {@link gscDataset}'s `tableSpec` (ADR-0021 R2-FIXES C5) — kept as a
 * named GSC-specific export. Generic schema authoring types are owned by
 * `@gscdump/lakehouse`.
 *
 * CONTRACT NOTE: implementation agents must treat the RETURNED VALUE as the
 * source of truth — do not hand-list columns elsewhere.
 */
declare function icebergTableSpec(table: IcebergTableName, encoding?: PartitionKeyEncoding): IcebergTableSpec;
/** All Iceberg table specs in legacy `'string'` encoding, keyed by table name. */
declare const ICEBERG_SCHEMAS_STRING: Record<IcebergTableName, IcebergTableSpec>;
/** All Iceberg table specs in `'int'` encoding (INT site_id + INT search_type). */
declare const ICEBERG_SCHEMAS_INT: Record<IcebergTableName, IcebergTableSpec>;
/** All Iceberg table specs for the default new-catalog encoding. */
declare const ICEBERG_SCHEMAS: Record<IcebergTableName, IcebergTableSpec>;
/** Table specs for the given encoding (`'int'` default). */
declare function icebergSchemasFor(encoding?: PartitionKeyEncoding): Record<IcebergTableName, IcebergTableSpec>;
/** True when `table` is one of the canonical {@link ICEBERG_TABLES}. */
declare function isIcebergTable(table: string): table is IcebergTableName;
/**
 * Narrow an arbitrary table name to a canonical {@link IcebergTableName},
 * throwing a clear error otherwise. Guards write paths that index
 * `ICEBERG_SCHEMAS` (a `Record<IcebergTableName, …>`) — a non-canonical name
 * silently yields `undefined` there, propagating a corrupt/empty spec into the
 * Iceberg job instead of failing loudly.
 */
declare function assertIcebergTable(table: string): IcebergTableName;
export { ICEBERG_FIELD_ID_BASE, ICEBERG_PARTITION_COLUMNS, ICEBERG_PARTITION_SPEC, ICEBERG_SCHEMAS, ICEBERG_SCHEMAS_INT, ICEBERG_SCHEMAS_STRING, ICEBERG_TABLES, INT_SEARCH_TYPE, type IcebergS3Config, IcebergTableName, type PartitionKeyEncoding$1 as PartitionKeyEncoding, SEARCH_TYPE_INT, type SearchType, assertIcebergTable, icebergPartitionColumns, icebergSchemasFor, icebergTableSpec, isIcebergTable };