import CoreML
import SQLiteVec
import Testing

@testable import EmbeddingsBenchmarkLib

func createDatabase(_ data: [[Float]]) async throws -> Database {
    try SQLiteVec.initialize()
    let db = try Database(.inMemory)
    try await db.execute("CREATE VIRTUAL TABLE embeddings USING vec0(embedding float[3])")
    for (index, row) in data.enumerated() {
        try await db.execute(
            """
                INSERT INTO embeddings(rowid, embedding)
                VALUES (?, ?)
            """,
            params: [index, row]
        )
    }
    return db
}

@Test func testEmbeddingMethods() async throws {
    let data: [[Float]] = [
        [1.0, 2.0, 3.0],
        [4.0, 5.0, 6.0],
        [7.0, 8.0, 9.0]
    ]
    let embeddings = MLTensor(shape: [3, 3], scalars: data.flatMap { $0 })
    let coreMLResult = await queryEmbeddings(embeddings: embeddings, tokenIds: [0, 2])

    let db = try await createDatabase(data)
    let sqliteResult = try await queryEmbeddings(
        db: db,
        query: "(?, ?)",
        tokenIds: [0, 2],
        vectorSize: 3)

    #expect(coreMLResult == sqliteResult)
}