mirror of
https://github.com/lancedb/lancedb.git
synced 2026-10-12 23:02:28 +00:00
A view over a view could only be brought fully up to date by refreshing each view in the lineage by hand, in order. `cascade` on the refresh builder, Python `refresh`/`refresh_async` and Node `refresh` does that in one call: every upstream view is refreshed furthest first, each reading the version the one before it left, then the named view. `full` applies to the named view only. Locally the lineage is walked through each view's source, and each hop is refreshed against the incarnation read in that walk. Remote tables send `cascade` in the refresh request, only when set, so older servers that reject unknown fields are unaffected by plain refreshes.
200 lines
6.4 KiB
TypeScript
200 lines
6.4 KiB
TypeScript
// SPDX-License-Identifier: Apache-2.0
|
|
// SPDX-FileCopyrightText: Copyright The LanceDB Authors
|
|
|
|
import * as tmp from "tmp";
|
|
|
|
import { Connection, connect } from "../lancedb";
|
|
import {
|
|
DEFINITION_META_KEY,
|
|
definitionFromMetadata,
|
|
} from "../lancedb/materialized_view";
|
|
|
|
describe("materialized views", () => {
|
|
let tmpDir: tmp.DirResult;
|
|
let db: Connection;
|
|
|
|
beforeEach(async () => {
|
|
tmpDir = tmp.dirSync({ unsafeCleanup: true });
|
|
db = await connect(tmpDir.name);
|
|
await db.createTable(
|
|
"people",
|
|
[
|
|
{ name: "ada", age: 36 },
|
|
{ name: "kid", age: 7 },
|
|
{ name: "grace", age: 85 },
|
|
],
|
|
{ storageOptions: { newTableEnableStableRowIds: "true" } },
|
|
);
|
|
});
|
|
afterEach(() => tmpDir.removeCallback());
|
|
|
|
it("reads stored queries and legacy layouts", () => {
|
|
const read = (stored: string) =>
|
|
definitionFromMetadata(new Map([[DEFINITION_META_KEY, stored]]), "v");
|
|
const query =
|
|
"SELECT id, c.chunk FROM ns.docs, UNNEST(chunks) AS c WHERE id > 1";
|
|
expect(read(`{"format":1,"query":${JSON.stringify(query)}}`).query).toBe(
|
|
query,
|
|
);
|
|
|
|
// The structured layout written before the format number reads as the
|
|
// query it described, under either of its kind tags.
|
|
expect(
|
|
read(
|
|
'{"kind":"namespaced_select","source_table":"people","source_namespace":["ns"],' +
|
|
'"projections":[{"output":"name","expression":"`name`"},' +
|
|
'{"output":"Shout","expression":"upper(name)"}],"filter":"age >= 18","limit":42}',
|
|
).query,
|
|
).toBe(
|
|
"SELECT `name`, upper(name) AS `Shout` FROM ns.people WHERE age >= 18 LIMIT 42",
|
|
);
|
|
expect(read('{"kind":"select","source_table":"people"}').query).toBe(
|
|
"SELECT * FROM people",
|
|
);
|
|
|
|
// A newer writer's layout is reported, never guessed at.
|
|
for (const newer of [
|
|
`{"format":3,"query":${JSON.stringify(query)}}`,
|
|
'{"kind":"select_v3","source_table":"people"}',
|
|
]) {
|
|
expect(() => read(newer)).toThrow(/cannot refresh/);
|
|
}
|
|
});
|
|
|
|
it("rejects a stored limit a number cannot carry", () => {
|
|
const big = new Map([
|
|
[
|
|
DEFINITION_META_KEY,
|
|
'{"kind":"select","source_table":"people","limit":9007199254740993}',
|
|
],
|
|
]);
|
|
expect(() => definitionFromMetadata(big, "v")).toThrow(
|
|
/too large to represent exactly/,
|
|
);
|
|
});
|
|
|
|
it("creates, refreshes and queries a view", async () => {
|
|
const view = await db.createMaterializedView("adults", "people", {
|
|
select: ["name", ["shout", "upper(name)"]],
|
|
where: "age >= 18",
|
|
});
|
|
expect(view.name).toBe("adults");
|
|
expect(await view.table().countRows()).toBe(2);
|
|
|
|
const rows = await view.table().query().toArray();
|
|
expect(rows.map((r) => r.shout).sort()).toEqual(["ADA", "GRACE"]);
|
|
});
|
|
|
|
it("round-trips the definition", async () => {
|
|
await db.createMaterializedView("adults", "people", {
|
|
where: "age >= 18",
|
|
});
|
|
const view = await db.openMaterializedView("adults");
|
|
const definition = await view.definition();
|
|
expect(definition.query).toBe(
|
|
"SELECT name, age FROM people WHERE age >= 18",
|
|
);
|
|
});
|
|
|
|
it("refreshes incrementally after an append", async () => {
|
|
const view = await db.createMaterializedView("copy", "people", {
|
|
withNoData: true,
|
|
});
|
|
await view.refresh();
|
|
|
|
const people = await db.openTable("people");
|
|
await people.add([{ name: "alan", age: 41 }]);
|
|
const result = await view.refresh();
|
|
expect(result.mode).toBe("incremental");
|
|
expect(Number(result.rowsWritten)).toBe(1);
|
|
expect(await view.table().countRows()).toBe(4);
|
|
|
|
expect((await view.refresh()).mode).toBe("no_op");
|
|
});
|
|
|
|
it("cascades a refresh through upstream views", async () => {
|
|
await db.createMaterializedView("adults", "people", {
|
|
where: "age >= 18",
|
|
withNoData: true,
|
|
});
|
|
const seniors = await db.createMaterializedView("seniors", "adults", {
|
|
where: "age >= 65",
|
|
withNoData: true,
|
|
});
|
|
expect(Number((await seniors.refresh()).rowsWritten)).toBe(0);
|
|
|
|
expect(Number((await seniors.refresh({ cascade: true })).rowsWritten)).toBe(
|
|
1,
|
|
);
|
|
expect(await (await db.openTable("adults")).countRows()).toBe(2);
|
|
});
|
|
|
|
it("lists views and rejects non-views", async () => {
|
|
await db.createMaterializedView("adults", "people", {
|
|
where: "age >= 18",
|
|
});
|
|
expect(await db.listMaterializedViews()).toEqual(["adults"]);
|
|
await expect(db.openMaterializedView("people")).rejects.toThrow(
|
|
"not a materialized view",
|
|
);
|
|
await expect(db.dropMaterializedView("people")).rejects.toThrow(
|
|
"not a materialized view",
|
|
);
|
|
|
|
await db.dropMaterializedView("adults");
|
|
expect(await db.listMaterializedViews()).toEqual([]);
|
|
});
|
|
|
|
it("returns a job when dropping a view asynchronously", async () => {
|
|
await db.createMaterializedView("adults", "people");
|
|
|
|
const job = await db.dropMaterializedViewAsync("adults");
|
|
expect(job.id).toBeNull();
|
|
await job.wait();
|
|
expect(await db.listMaterializedViews()).toEqual([]);
|
|
});
|
|
|
|
it("rejects an invalid expression at create time", async () => {
|
|
await expect(
|
|
db.createMaterializedView("bad", "people", {
|
|
select: [["x", "missing + 1"]],
|
|
}),
|
|
).rejects.toThrow("missing");
|
|
});
|
|
|
|
it("rejects invalid numeric options before creating anything", async () => {
|
|
for (const limit of [-5, 1.5, Infinity, NaN]) {
|
|
await expect(
|
|
db.createMaterializedView("bad", "people", { limit }),
|
|
).rejects.toThrow("non-negative integer");
|
|
}
|
|
expect(await db.listMaterializedViews()).toEqual([]);
|
|
|
|
const view = await db.createMaterializedView("copy", "people");
|
|
for (const sourceVersion of [-1, 1.5, Infinity, NaN]) {
|
|
await expect(view.refresh({ sourceVersion })).rejects.toThrow(
|
|
"non-negative integer",
|
|
);
|
|
}
|
|
});
|
|
|
|
it("quotes bare select names", async () => {
|
|
await db.createTable("odd_names", [{ "order item": "widget" }], {
|
|
storageOptions: { newTableEnableStableRowIds: "true" },
|
|
});
|
|
const view = await db.createMaterializedView("quoted", "odd_names", {
|
|
select: ["order item"],
|
|
withNoData: true,
|
|
});
|
|
const result = await view.refresh();
|
|
expect(Number(result.rowsWritten)).toBe(1);
|
|
});
|
|
|
|
it("requires stable row ids on the source", async () => {
|
|
await db.createTable("plain", [{ x: 1 }]);
|
|
await expect(db.createMaterializedView("v", "plain")).rejects.toThrow(
|
|
"stable row ids",
|
|
);
|
|
});
|
|
});
|