@gscdump/engine 3.5.0 → 3.6.0

package/dist/adapters/parquet-attach.d.mts

@@ -1,7 +1,7 @@
 import { SnapshotQueryRunner } from "./snapshot-attach.mjs";
-interface AttachParquetIndexOptions {
+export interface AttachParquetIndexOptions {
   /**
-   * Map of table name → list of Parquet URLs. The URL list may mix monthly
+   * Map of table name → list of Parquet URLs or local paths. The list may mix monthly
    * compacted files and per-day files — DuckDB will scan all of them with
    * `union_by_name = true`. Empty lists are skipped (no view created).
    */
@@ -14,10 +14,9 @@ interface AttachParquetIndexOptions {
    */
   forceDownload?: boolean;
 }
-interface AttachParquetIndexResult {
+export interface AttachParquetIndexResult {
   schema: string;
   /** Tables for which a view was created. */
   tables: string[];
 }
-declare function attachParquetIndex(runner: SnapshotQueryRunner, opts: AttachParquetIndexOptions): Promise<AttachParquetIndexResult>;
-export { AttachParquetIndexOptions, AttachParquetIndexResult, attachParquetIndex };
+export declare function attachParquetIndex(runner: SnapshotQueryRunner, opts: AttachParquetIndexOptions): Promise<AttachParquetIndexResult>;

package/dist/adapters/parquet-attach.mjs

@@ -1,11 +1,16 @@
 const IDENT_RE = /^[A-Z_][\w$]*$/i;
+const REMOTE_URL_RE = /^(?:https?|s3):\/\//i;
 async function attachParquetIndex(runner, opts) {
 	const schema = opts.schema ?? "main";
 	const forceDownload = opts.forceDownload !== false;
 	if (!IDENT_RE.test(schema)) throw new TypeError(`attachParquetIndex: invalid schema identifier ${JSON.stringify(schema)}`);
 	for (const table of Object.keys(opts.tables)) if (!IDENT_RE.test(table)) throw new TypeError(`attachParquetIndex: invalid table identifier ${JSON.stringify(table)}`);
-	await runner("LOAD httpfs").catch(() => void 0);
-	if (forceDownload) await runner("SET force_download=true");
+	if (Object.values(opts.tables).some((urls) => urls.some((url) => REMOTE_URL_RE.test(url)))) {
+		await runner("LOAD httpfs").catch((cause) => {
+			throw new Error("Remote files require the DuckDB httpfs extension. Install httpfs before attaching remote files.", { cause });
+		});
+		if (forceDownload) await runner("SET force_download=true");
+	}
 	await runner(`CREATE SCHEMA IF NOT EXISTS ${schema}`);
 	const created = [];
 	for (const [table, urls] of Object.entries(opts.tables)) {

package/dist/adapters/r2-manifest.d.mts

@@ -26,7 +26,7 @@ interface R2ConditionalPutOptions {
  * Minimal Cloudflare R2 binding shape needed for the manifest CAS loop.
  * Structurally compatible with Cloudflare's `R2Bucket` Workers API.
  */
-interface R2ManifestBucketLike {
+export interface R2ManifestBucketLike {
   get: (key: string) => Promise<R2ObjectBody | null>;
   put: (key: string, bytes: string | Uint8Array, options?: R2ConditionalPutOptions) => Promise<R2ObjectMetadata | null>;
   list: (options?: {
@@ -46,7 +46,7 @@ interface R2ManifestBucketLike {
  * into metrics (prom-client, console.table, the contention harness) to
  * measure rejection rate and latency under real R2 load.
  */
-type R2ManifestEvent = {
+export type R2ManifestEvent = {
   kind: 'cas-attempt';
   siteId: string;
   table: TableName;
@@ -62,7 +62,7 @@ type R2ManifestEvent = {
   table: TableName;
   attempts: number;
 };
-interface CreateR2ManifestStoreOptions {
+export interface CreateR2ManifestStoreOptions {
   bucket: R2ManifestBucketLike;
   /** Tenant scope. All shard keys are prefixed `u_<userId>/manifest/...`. */
   userId: string;
@@ -78,5 +78,4 @@ interface CreateR2ManifestStoreOptions {
    */
   onEvent?: (event: R2ManifestEvent) => void;
 }
-declare function createR2ManifestStore(opts: CreateR2ManifestStoreOptions): ManifestStore;
-export { CreateR2ManifestStoreOptions, R2ManifestBucketLike, R2ManifestEvent, createR2ManifestStore };
+export declare function createR2ManifestStore(opts: CreateR2ManifestStoreOptions): ManifestStore;

package/dist/adapters/r2.d.mts

@@ -30,7 +30,7 @@ interface R2BucketLike {
   list: (options?: R2ListOptions) => Promise<R2ListResult>;
   head: (key: string) => Promise<R2HeadObject | null>;
 }
-interface R2DataSourceOptions {
+export interface R2DataSourceOptions {
   /**
    * R2 bucket binding. Structurally typed — pass a Cloudflare `R2Bucket` or
    * any object with the same get/put/delete/list/head surface.
@@ -43,5 +43,4 @@ interface R2DataSourceOptions {
    */
   bucketName?: string;
 }
-declare function createR2DataSource(options: R2DataSourceOptions): DataSource;
-export { R2DataSourceOptions, createR2DataSource };
+export declare function createR2DataSource(options: R2DataSourceOptions): DataSource;

package/dist/adapters/snapshot-attach.d.mts

@@ -4,13 +4,13 @@ import { SnapshotIndex } from "../snapshot.mjs";
  * this so the function works with AsyncDuckDB (browser DuckDB-WASM) or
  * @duckdb/node-api (Node) without coupling to either.
  */
-type SnapshotQueryRunner = (sql: string) => Promise<Array<Record<string, unknown>>>;
-interface AttachSnapshotOptions {
+export type SnapshotQueryRunner = (sql: string) => Promise<Array<Record<string, unknown>>>;
+export interface AttachSnapshotOptions {
   /** Index produced by the builder. */
   index: SnapshotIndex;
   /**
-   * Map from filename (`cold-YYYY-MM.duckdb`, `hot.duckdb`) to an HTTPS
-   * URL (typically a pre-signed R2 URL). Must contain an entry for every
+   * Map from filename (`cold-YYYY-MM.duckdb`, `hot.duckdb`) to a local path
+   * or remote HTTP(S)/S3 URL. Must contain an entry for every
    * cold month in `index.cold` and — if `index.hot` — for `hot.duckdb`.
    */
   attachUrls: Record<string, string>;
@@ -22,7 +22,7 @@ interface AttachSnapshotOptions {
    */
   forceDownload?: boolean;
 }
-interface AttachSnapshotResult {
+export interface AttachSnapshotResult {
   schema: string;
   /** Aliases we ATTACH'd — e.g. ['cold_2024_09', 'cold_2024_10', 'hot']. */
   aliases: string[];
@@ -33,6 +33,5 @@ interface AttachSnapshotResult {
  * Turns a filename like `cold-2024-09.duckdb` into a valid SQL identifier
  * `cold_2024_09`. `hot.duckdb` → `hot`.
  */
-declare function snapshotAlias(fileName: string): string;
-declare function attachSnapshotIndex(runner: SnapshotQueryRunner, opts: AttachSnapshotOptions): Promise<AttachSnapshotResult>;
-export { AttachSnapshotOptions, AttachSnapshotResult, SnapshotQueryRunner, attachSnapshotIndex, snapshotAlias };
+export declare function snapshotAlias(fileName: string): string;
+export declare function attachSnapshotIndex(runner: SnapshotQueryRunner, opts: AttachSnapshotOptions): Promise<AttachSnapshotResult>;

package/dist/adapters/snapshot-attach.mjs

@@ -3,6 +3,7 @@ import { err, ok, unwrapResult } from "gscdump/result";
 const YEAR_MONTH_RE = /^\d{4}-\d{2}$/;
 const SCHEMA_IDENT_RE = /^[A-Z_][\w$]*$/i;
 const COLD_FILENAME_RE = /^cold-(\d{4}-\d{2})\.duckdb$/;
+const REMOTE_URL_RE = /^(?:https?|s3):\/\//i;
 function snapshotAlias(fileName) {
 	if (fileName === "hot.duckdb") return "hot";
 	const m = fileName.match(COLD_FILENAME_RE);
@@ -27,8 +28,6 @@ async function attachSnapshotIndexResult(runner, opts) {
 	if (index?.version !== 1) return err(engineErrors.unsupportedSnapshotIndexVersion(index?.version));
 	if (!SCHEMA_IDENT_RE.test(schema)) return err(engineErrors.invalidSchemaIdentifier(schema));
 	for (const ym of index.cold) if (!YEAR_MONTH_RE.test(ym)) return err(engineErrors.invalidYearMonth(ym));
-	await runner("LOAD httpfs").catch(() => void 0);
-	if (forceDownload) await runner("SET force_download=true");
 	const plan = [];
 	for (const ym of index.cold) {
 		const fileName = `cold-${ym}.duckdb`;
@@ -50,6 +49,12 @@ async function attachSnapshotIndexResult(runner, opts) {
 			url
 		});
 	}
+	if (plan.some(({ url }) => REMOTE_URL_RE.test(url))) {
+		await runner("LOAD httpfs").catch((cause) => {
+			throw new Error("Remote files require the DuckDB httpfs extension. Install httpfs before attaching remote files.", { cause });
+		});
+		if (forceDownload) await runner("SET force_download=true");
+	}
 	const aliases = [];
 	for (const { alias, url } of plan) {
 		await runner(`ATTACH '${url.replace(/'/g, "''")}' AS ${alias} (READ_ONLY)`);

package/dist/analysis-range.d.mts

@@ -1,7 +1,6 @@
 import { AnalysisParams } from "./analysis-types.mjs";
-interface AnalysisCoverageRange {
+export interface AnalysisCoverageRange {
   start: string;
   end: string;
 }
-declare function analysisRequestedRange(params: AnalysisParams, onInvalidFilter?: (error: unknown) => void): AnalysisCoverageRange | null;
-export { AnalysisCoverageRange, analysisRequestedRange };
+export declare function analysisRequestedRange(params: AnalysisParams, onInvalidFilter?: (error: unknown) => void): AnalysisCoverageRange | null;

package/dist/analysis-types.d.mts

@@ -1,6 +1,6 @@
 import { BuilderState, SearchType } from "gscdump/query";
-type AnalysisTool = 'striking-distance' | 'opportunity' | 'movers' | 'decay' | 'zero-click' | 'brand' | 'cannibalization' | 'clustering' | 'concentration' | 'seasonality' | 'trends' | 'ctr-anomaly' | 'position-volatility' | 'long-tail' | 'intent-atlas' | 'query-migration' | 'bayesian-ctr' | 'stl-decompose' | 'change-point' | 'bipartite-pagerank' | 'survival' | 'position-distribution' | 'ctr-curve' | 'dark-traffic' | 'content-velocity' | 'keyword-breadth' | 'device-gap' | 'data-query' | 'data-detail';
-interface AnalysisParams {
+export type AnalysisTool = 'striking-distance' | 'opportunity' | 'movers' | 'decay' | 'zero-click' | 'brand' | 'cannibalization' | 'clustering' | 'concentration' | 'seasonality' | 'trends' | 'ctr-anomaly' | 'position-volatility' | 'long-tail' | 'intent-atlas' | 'query-migration' | 'bayesian-ctr' | 'stl-decompose' | 'change-point' | 'bipartite-pagerank' | 'survival' | 'position-distribution' | 'ctr-curve' | 'dark-traffic' | 'content-velocity' | 'keyword-breadth' | 'device-gap' | 'data-query' | 'data-detail';
+export interface AnalysisParams {
   type: AnalysisTool;
   startDate?: string;
   endDate?: string;
@@ -48,10 +48,9 @@ interface AnalysisParams {
   /** GSC slice the analysis is scoped to. Undefined = analyzer runs cross-type (today's behaviour for web-only sites). */
   searchType?: SearchType;
 }
-interface AnalysisResult {
+export interface AnalysisResult {
   results: Record<string, unknown>[];
   meta: Record<string, unknown>;
 }
 /** Coerce arbitrary value (number, bigint, string, null) to number, defaulting to 0. */
-declare function num(v: unknown): number;
-export { AnalysisParams, AnalysisResult, AnalysisTool, num };
+export declare function num(v: unknown): number;

package/dist/analyzer/define.d.mts

@@ -2,7 +2,7 @@ import { AnalysisParams } from "../analysis-types.mjs";
 import { FileSet } from "../source/source-types.mjs";
 import { Analyzer, BuildContext, RequiredCapability, SqlExtraQuery } from "./types.mjs";
 import { BuilderState } from "gscdump/query";
-interface SqlPlanSpec {
+export interface SqlPlanSpec {
   sql: string;
   params: unknown[];
   current: FileSet;
@@ -10,15 +10,15 @@ interface SqlPlanSpec {
   extraFiles?: Record<string, FileSet>;
   extraQueries?: SqlExtraQuery[];
 }
-interface ReduceCtx<InputRow> {
+export interface ReduceCtx<InputRow> {
   /** Extra SQL-query results keyed by `SqlExtraQuery.name` (SQL path only). */
   extras?: Record<string, InputRow[]>;
 }
-type Reducer<Params, InputRow, Result> = (rows: InputRow[] | Record<string, InputRow[]>, params: Params, ctx: ReduceCtx<InputRow>) => {
+export type Reducer<Params, InputRow, Result> = (rows: InputRow[] | Record<string, InputRow[]>, params: Params, ctx: ReduceCtx<InputRow>) => {
   results: Result;
   meta?: Record<string, unknown>;
 };
-interface DefineAnalyzerOptions<Params extends AnalysisParams, InputRow, Result> {
+export interface DefineAnalyzerOptions<Params extends AnalysisParams, InputRow, Result> {
   id: string;
   /**
    * Shared reducer used by both SQL and row paths. Use this when the
@@ -39,10 +39,9 @@ interface DefineAnalyzerOptions<Params extends AnalysisParams, InputRow, Result>
   /** Capabilities required by the row plan. Defaults to `[]`. */
   rowsRequires?: readonly RequiredCapability[];
 }
-interface DefinedAnalyzer {
+export interface DefinedAnalyzer {
   id: string;
   sql?: Analyzer;
   rows?: Analyzer;
 }
-declare function defineAnalyzer<Params extends AnalysisParams, InputRow, Result>(opts: DefineAnalyzerOptions<Params, InputRow, Result>): DefinedAnalyzer;
-export { DefineAnalyzerOptions, DefinedAnalyzer, ReduceCtx, Reducer, SqlPlanSpec, defineAnalyzer };
+export declare function defineAnalyzer<Params extends AnalysisParams, InputRow, Result>(opts: DefineAnalyzerOptions<Params, InputRow, Result>): DefinedAnalyzer;

package/dist/analyzer/dispatch.d.mts

@@ -3,7 +3,7 @@ import { AnalysisQuerySource } from "../source/source-types.mjs";
 import { RequiredCapability } from "./types.mjs";
 import { EngineError } from "../errors.mjs";
 import { AnalyzerRegistry } from "./registry.mjs";
-declare class AnalyzerCapabilityError extends Error {
+export declare class AnalyzerCapabilityError extends Error {
   readonly tool: string;
   readonly missing: readonly RequiredCapability[];
   readonly engineError: EngineError;
@@ -13,5 +13,4 @@ declare class AnalyzerCapabilityError extends Error {
  * Run an analyzer against a generic `AnalysisQuerySource`. The registry is
  * an explicit parameter — callers build one via `createAnalyzerRegistry`.
  */
-declare function runAnalyzerFromSource(source: AnalysisQuerySource, params: AnalysisParams, registry: AnalyzerRegistry): Promise<AnalysisResult>;
-export { AnalyzerCapabilityError, runAnalyzerFromSource };
+export declare function runAnalyzerFromSource(source: AnalysisQuerySource, params: AnalysisParams, registry: AnalyzerRegistry): Promise<AnalysisResult>;

package/dist/analyzer/registry.d.mts

@@ -1,10 +1,10 @@
 import { Analyzer } from "./types.mjs";
 import { DefinedAnalyzer } from "./define.mjs";
-interface AnalyzerVariants {
+export interface AnalyzerVariants {
   sql?: Analyzer;
   rows?: Analyzer;
 }
-interface AnalyzerRegistryInit {
+export interface AnalyzerRegistryInit {
   /**
    * Preferred for in-tree composition: pass `DefinedAnalyzer[]` directly so
    * SQL/row variants can never drift apart from their `defineAnalyzer` site.
@@ -15,7 +15,7 @@ interface AnalyzerRegistryInit {
   /** Flat-array path retained for narrow tree-shaken registry composition. */
   sql?: readonly Analyzer[];
 }
-interface AnalyzerRegistry {
+export interface AnalyzerRegistry {
   listAnalyzerIds: () => readonly string[];
   getAnalyzerVariants: (id: string) => AnalyzerVariants | undefined;
   resolveAnalyzer: (id: string, sourceSupportsSql: boolean) => Analyzer | undefined;
@@ -29,5 +29,4 @@ interface AnalyzerRegistry {
  * No global state; call this once per logical use (typically at startup
  * or per-request in a worker).
  */
-declare function createAnalyzerRegistry(init?: AnalyzerRegistryInit): AnalyzerRegistry;
-export { AnalyzerRegistry, AnalyzerRegistryInit, AnalyzerVariants, createAnalyzerRegistry };
+export declare function createAnalyzerRegistry(init?: AnalyzerRegistryInit): AnalyzerRegistry;

package/dist/analyzer/types.d.mts

@@ -12,8 +12,8 @@ import { BuilderState } from "gscdump/query";
  * on `SourceCapabilities`. Single source of truth — adding a new capability
  * is one line in `SourceCapabilities`.
  */
-type RequiredCapability = 'executeSql' | keyof SourceCapabilities;
-interface SqlExtraQuery {
+export type RequiredCapability = 'executeSql' | keyof SourceCapabilities;
+export interface SqlExtraQuery {
   name: string;
   sql: string;
   params: unknown[];
@@ -22,7 +22,7 @@ interface SqlExtraQuery {
  * SQL-native plan: SQL string + placeholders, with optional extra file sets
  * and follow-up queries.
  */
-interface SqlPlan {
+export interface SqlPlan {
   kind: 'sql';
   sql: string;
   params: unknown[];
@@ -31,7 +31,7 @@ interface SqlPlan {
   extraFiles?: Record<string, FileSet>;
   extraQueries?: SqlExtraQuery[];
 }
-interface TypedRowQuery<T extends Row = Row> {
+export interface TypedRowQuery<T extends Row = Row> {
   state: BuilderState;
   /** Optional type tag for downstream narrowing. */
   rowType?: (row: Row) => T;
@@ -41,11 +41,11 @@ interface TypedRowQuery<T extends Row = Row> {
  * dispatcher runs each against a source's `queryRows` and hands the row
  * collection to `reduce`.
  */
-interface RowQueriesPlan {
+export interface RowQueriesPlan {
   kind: 'rows';
   queries: Record<string, TypedRowQuery>;
 }
-type Plan = SqlPlan | RowQueriesPlan;
+export type Plan = SqlPlan | RowQueriesPlan;
 /**
  * Plan-build context. Surfaced from the source at dispatch time so analyzers
  * that compose SQL from a typed `BuilderState` can pick up the right dialect
@@ -58,7 +58,7 @@ type Plan = SqlPlan | RowQueriesPlan;
  * declaration (`'adapter'` in `requires`) is the runtime guarantee; the
  * helper makes the failure mode loud if the contract is broken.
  */
-interface BuildContext {
+export interface BuildContext {
   adapter?: ResolverAdapter<any>;
   siteId?: string | number;
 }
@@ -67,8 +67,8 @@ interface BuildContext {
  * but the dispatcher handed it a context without one. Centralizes the assert
  * so analyzers don't repeat `ctx.adapter!` with explanatory comments.
  */
-declare function requireAdapter(ctx: BuildContext, analyzerId: string): ResolverAdapter<any>;
-interface ReduceContext<TRow extends Row = Row> {
+export declare function requireAdapter(ctx: BuildContext, analyzerId: string): ResolverAdapter<any>;
+export interface ReduceContext<TRow extends Row = Row> {
   params: AnalysisParams;
   /** Extra SQL-query results keyed by `SqlExtraQuery.name`. */
   extras?: Record<string, TRow[]>;
@@ -79,7 +79,7 @@ interface ReduceContext<TRow extends Row = Row> {
  * when their reducer assumes specific columns exist — catches drift between
  * `build` (SELECT list) and `reduce` (column access) at compile time.
  */
-interface Analyzer<P extends AnalysisParams = AnalysisParams, R = unknown, TRow extends Row = Row> {
+export interface Analyzer<P extends AnalysisParams = AnalysisParams, R = unknown, TRow extends Row = Row> {
   /** Stable tool id (e.g. `striking-distance`, `opportunity`). */
   id: string;
   /** Capabilities a host source must provide. */
@@ -91,5 +91,4 @@ interface Analyzer<P extends AnalysisParams = AnalysisParams, R = unknown, TRow
     results: R;
     meta?: Record<string, unknown>;
   };
-}
-export { Analyzer, BuildContext, Plan, ReduceContext, RequiredCapability, RowQueriesPlan, SqlExtraQuery, SqlPlan, TypedRowQuery, requireAdapter };
+}

package/dist/arrow-utils.d.mts

@@ -2,13 +2,12 @@
  * Arrow → row conversion shared by every engine adapter.
  *
  * DuckDB returns Arrow tables; drizzle / consumers want plain row objects.
- * `toArray()` yields Arrow row proxies; `.toJSON()` materializes them, but
- * some Arrow versions skip `.toJSON` and yield plain objects already.
+ * Read Arrow vectors directly to avoid creating a proxy for every row.
+ * Simpler table wrappers can supply row objects through `toArray()`.
  */
-interface ArrowTableLike {
+export interface ArrowTableLike {
   toArray: () => Array<Record<string, unknown> & {
     toJSON?: () => Record<string, unknown>;
   }>;
 }
-declare function arrowToRows(result: unknown): Record<string, unknown>[];
-export { ArrowTableLike, arrowToRows };
+export declare function arrowToRows(result: unknown): Record<string, unknown>[];

package/dist/arrow-utils.mjs

@@ -1,4 +1,14 @@
 function arrowToRows(result) {
+	const table = result;
+	if (!Array.isArray(table) && typeof table?.getChildAt === "function" && Array.isArray(table.schema?.fields) && Number.isInteger(table.numRows) && table.numRows >= 0) {
+		const rows = Array.from({ length: table.numRows }, () => ({}));
+		for (let columnIndex = 0; columnIndex < table.schema.fields.length; columnIndex++) {
+			const name = table.schema.fields[columnIndex].name;
+			const column = table.getChildAt(columnIndex);
+			for (let rowIndex = 0; rowIndex < rows.length; rowIndex++) rows[rowIndex][name] = column.get(rowIndex);
+		}
+		return rows;
+	}
 	const r = result;
 	const arr = Array.isArray(r) ? r : typeof r?.toArray === "function" ? r.toArray() : [];
 	if (!arr || arr.length === 0) return [];

package/dist/coerce.d.mts

@@ -1,3 +1,2 @@
 import { Row } from "@gscdump/contracts";
-declare function coerceRows(rows: readonly Row[]): Row[];
-export { coerceRows };
+export declare function coerceRows(rows: readonly Row[]): Row[];

package/dist/compaction.d.mts

@@ -5,10 +5,9 @@ import "./storage.mjs";
  * - d7 → d30 once the entire weekly bucket sits behind `d7` days (default 30).
  * - d30 → d90 once the entire monthly bucket sits behind `d30` days (default 90).
  */
-interface CompactionThresholds {
+export interface CompactionThresholds {
   raw?: number;
   d7?: number;
   d30?: number;
 }
-declare function enumeratePartitions(startDate: string, endDate: string): string[];
-export { CompactionThresholds, enumeratePartitions };
+export declare function enumeratePartitions(startDate: string, endDate: string): string[];

package/dist/drizzle-schema.d.mts

@@ -1,5 +1,5 @@
 import { TableName } from "@gscdump/contracts";
-declare const pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "pages";
   schema: undefined;
   columns: {
@@ -81,7 +81,7 @@ declare const pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   };
   dialect: "pg";
 }>;
-declare const queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "queries";
   schema: undefined;
   columns: {
@@ -163,7 +163,7 @@ declare const queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   };
   dialect: "pg";
 }>;
-declare const countries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const countries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "countries";
   schema: undefined;
   columns: {
@@ -245,7 +245,7 @@ declare const countries: import("drizzle-orm/pg-core/table").PgTableWithColumns<
   };
   dialect: "pg";
 }>;
-declare const page_queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const page_queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "page_queries";
   schema: undefined;
   columns: {
@@ -358,7 +358,7 @@ declare const page_queries: import("drizzle-orm/pg-core/table").PgTableWithColum
  * - `clicks_{device}` / `impressions_{device}` / `sum_position_{device}`:
  *   device breakdown pivoted from a GSC `['date','device']` query.
  */
-declare const dates: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const dates: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "dates";
   schema: undefined;
   columns: {
@@ -575,7 +575,7 @@ declare const dates: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   };
   dialect: "pg";
 }>;
-declare const search_appearance: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const search_appearance: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "search_appearance";
   schema: undefined;
   columns: {
@@ -657,7 +657,7 @@ declare const search_appearance: import("drizzle-orm/pg-core/table").PgTableWith
   };
   dialect: "pg";
 }>;
-declare const search_appearance_pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const search_appearance_pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "search_appearance_pages";
   schema: undefined;
   columns: {
@@ -754,7 +754,7 @@ declare const search_appearance_pages: import("drizzle-orm/pg-core/table").PgTab
   };
   dialect: "pg";
 }>;
-declare const search_appearance_queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const search_appearance_queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "search_appearance_queries";
   schema: undefined;
   columns: {
@@ -851,7 +851,7 @@ declare const search_appearance_queries: import("drizzle-orm/pg-core/table").PgT
   };
   dialect: "pg";
 }>;
-declare const search_appearance_page_queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const search_appearance_page_queries: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "search_appearance_page_queries";
   schema: undefined;
   columns: {
@@ -963,7 +963,7 @@ declare const search_appearance_page_queries: import("drizzle-orm/pg-core/table"
   };
   dialect: "pg";
 }>;
-declare const hourly_pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
+export declare const hourly_pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
   name: "hourly_pages";
   schema: undefined;
   columns: {
@@ -1060,7 +1060,7 @@ declare const hourly_pages: import("drizzle-orm/pg-core/table").PgTableWithColum
   };
   dialect: "pg";
 }>;
-declare const drizzleSchema: {
+export declare const drizzleSchema: {
   pages: import("drizzle-orm/pg-core/table").PgTableWithColumns<{
     name: "pages";
     schema: undefined;
@@ -2107,10 +2107,9 @@ declare const drizzleSchema: {
     dialect: "pg";
   }>;
 };
-type DrizzleSchema = typeof drizzleSchema;
-declare const TABLE_METADATA: Record<TableName, {
+export type DrizzleSchema = typeof drizzleSchema;
+export declare const TABLE_METADATA: Record<TableName, {
   sortKey: string[];
   clusterKey: string[];
   version: number;
-}>;
-export { DrizzleSchema, TABLE_METADATA, countries, dates, drizzleSchema, hourly_pages, page_queries, pages, queries, search_appearance, search_appearance_page_queries, search_appearance_pages, search_appearance_queries };
+}>;

package/dist/duckdb.d.mts

@@ -1,5 +1,5 @@
 import { ParquetCodec, QueryExecutor, Row } from "./storage.mjs";
-interface DuckDBHandle {
+export interface DuckDBHandle {
   query: (sql: string, params?: unknown[]) => Promise<Row[]>;
   registerFileBuffer: (name: string, bytes: Uint8Array) => Promise<void>;
   copyFileToBuffer: (name: string) => Promise<Uint8Array>;
@@ -11,10 +11,10 @@ interface DuckDBHandle {
    */
   makeTempPath: (ext: string) => string;
 }
-interface DuckDBFactory {
+export interface DuckDBFactory {
   getDuckDB: () => Promise<DuckDBHandle>;
 }
-interface DuckDBReadOptions {
+export interface DuckDBReadOptions {
   /**
    * Maximum non-URI parquet reads held in flight while registering DuckDB
    * buffers. Defaults to 16; callers handling unusually large files can lower
@@ -22,6 +22,5 @@ interface DuckDBReadOptions {
    */
   bufferReadConcurrency?: number;
 }
-declare function createDuckDBCodec(factory: DuckDBFactory, options?: DuckDBReadOptions): ParquetCodec;
-declare function createDuckDBExecutor(factory: DuckDBFactory, options?: DuckDBReadOptions): QueryExecutor;
-export { DuckDBFactory, DuckDBHandle, DuckDBReadOptions, createDuckDBCodec, createDuckDBExecutor };
+export declare function createDuckDBCodec(factory: DuckDBFactory, options?: DuckDBReadOptions): ParquetCodec;
+export declare function createDuckDBExecutor(factory: DuckDBFactory, options?: DuckDBReadOptions): QueryExecutor;

package/dist/duckdb.mjs

@@ -156,11 +156,15 @@ function createDuckDBExecutor(factory, options = {}) {
 					const uri = dataSource.uri?.(key);
 					if (uri !== void 0) resolved.push(uri);
 					else {
-						if (!bufferedByName.has(key)) bufferedByName.set(key, {
-							key,
-							name: key
-						});
-						resolved.push(key);
+						let buffered = bufferedByName.get(key);
+						if (!buffered) {
+							buffered = {
+								key,
+								name: db.makeTempPath("parquet")
+							};
+							bufferedByName.set(key, buffered);
+						}
+						resolved.push(buffered.name);
 					}
 				}
 				placeholders[name] = resolved;

package/dist/engine.d.mts

@@ -1,3 +1,2 @@
 import { EngineOptions, StorageEngine } from "./storage.mjs";
-declare function createStorageEngine(opts: EngineOptions): StorageEngine;
-export { createStorageEngine };
+export declare function createStorageEngine(opts: EngineOptions): StorageEngine;

package/dist/entities/empty-types.d.mts

@@ -1,22 +1,21 @@
 import { DataSource } from "../storage.mjs";
 import { TenantCtx } from "@gscdump/contracts";
-interface EmptyTypesDoc {
+export interface EmptyTypesDoc {
   version: 1;
   /** SearchType strings detected as empty for this (user, site). */
   emptyTypes: string[];
   /** When each type was last marked empty (unix ms). Helps debug stale skips. */
   markedAt: Record<string, number>;
 }
-interface EmptyTypesStore {
+export interface EmptyTypesStore {
   load: (ctx: TenantCtx) => Promise<EmptyTypesDoc>;
   /** Add types to the empty set, preserving existing markers. No-op if all already present. */
   mark: (ctx: TenantCtx, types: readonly string[], now?: number) => Promise<EmptyTypesDoc>;
   /** Remove types from the empty set. Returns the updated doc. */
   clear: (ctx: TenantCtx, types: readonly string[]) => Promise<EmptyTypesDoc>;
 }
-interface CreateEmptyTypesStoreOptions {
+export interface CreateEmptyTypesStoreOptions {
   dataSource: DataSource;
   now?: () => number;
 }
-declare function createEmptyTypesStore(opts: CreateEmptyTypesStoreOptions): EmptyTypesStore;
-export { CreateEmptyTypesStoreOptions, EmptyTypesDoc, EmptyTypesStore, createEmptyTypesStore };
+export declare function createEmptyTypesStore(opts: CreateEmptyTypesStoreOptions): EmptyTypesStore;

package/dist/entities/indexing-metadata.d.mts

@@ -1,6 +1,6 @@
 import { DataSource } from "../storage.mjs";
 import { TenantCtx } from "@gscdump/contracts";
-interface IndexingMetadataRecord {
+export interface IndexingMetadataRecord {
   url: string;
   capturedAt: string;
   /** ISO-8601 notifyTime of the latest `URL_UPDATED` notification we've seen. */
@@ -9,18 +9,17 @@ interface IndexingMetadataRecord {
   latestRemoveAt?: string;
   raw?: unknown;
 }
-interface IndexingMetadataIndex {
+export interface IndexingMetadataIndex {
   version: 1;
   records: Record<string, IndexingMetadataRecord>;
 }
-interface IndexingMetadataStore {
+export interface IndexingMetadataStore {
   writeBatch: (ctx: TenantCtx, records: readonly IndexingMetadataRecord[]) => Promise<void>;
   loadIndex: (ctx: TenantCtx) => Promise<IndexingMetadataIndex>;
   getLatest: (ctx: TenantCtx, url: string) => Promise<IndexingMetadataRecord | undefined>;
 }
-interface CreateIndexingMetadataStoreOptions {
+export interface CreateIndexingMetadataStoreOptions {
   dataSource: DataSource;
   hash?: (url: string) => string;
 }
-declare function createIndexingMetadataStore(opts: CreateIndexingMetadataStoreOptions): IndexingMetadataStore;
-export { CreateIndexingMetadataStoreOptions, IndexingMetadataIndex, IndexingMetadataRecord, IndexingMetadataStore, createIndexingMetadataStore };
+export declare function createIndexingMetadataStore(opts: CreateIndexingMetadataStoreOptions): IndexingMetadataStore;