/**
 * HybridSearchService — combines FTS5 keyword search with vector similarity
 * search using Reciprocal Rank Fusion (RRF) to produce a hybrid ranking.
 *
 * When no embedding provider is configured (or alpha = 0), the service
 * transparently falls back to FTS5-only mode with zero overhead.
 *
 * Configuration model:
 *   alpha = 0.0  → FTS5 only
 *   alpha = 0.5  → balanced hybrid (default)
 *   alpha = 1.0  → vector only
 */

import type Database from 'better-sqlite3';
import type { EmbeddingProvider } from '../embeddings/provider.js';
import { SnippetSearchResult, SnippetRepositoryRef } from '$lib/server/models/search-result.js';
import { SnippetEntity } from '$lib/server/models/snippet.js';
import { SearchResultMapper } from '$lib/server/mappers/search-result.mapper.js';
import { SearchService } from './search.service.js';
import { VectorSearch } from './vector.search.js';
import { reciprocalRankFusion } from './rrf.js';

// ---------------------------------------------------------------------------
// Public interfaces
// ---------------------------------------------------------------------------

export interface HybridSearchOptions {
	repositoryId: string;
	versionId?: string;
	type?: 'code' | 'info';
	/** Maximum number of results to return. Default: 20. */
	limit?: number;
	/**
	 * Blend weight between FTS5 and vector search.
	 * 0.0 = FTS5 only, 1.0 = vector only, 0.5 = balanced.
	 * Default: 0.5.
	 */
	alpha?: number;
}

/**
 * Global search configuration stored in the settings table under
 * `search_config`.
 */
export interface SearchConfig {
	/** Blend weight (0.0–1.0). Default: 0.5. */
	alpha: number;
	/** Maximum results per search. Default: 20. */
	maxResults: number;
	/** True when an embedding provider is configured. */
	enableHybrid: boolean;
}

// ---------------------------------------------------------------------------
// Raw DB row used when re-fetching snippets by ID
// ---------------------------------------------------------------------------

interface RawSnippetById extends SnippetEntity {
	repo_id: string;
	repo_title: string;
}

// ---------------------------------------------------------------------------
// HybridSearchService
// ---------------------------------------------------------------------------

export class HybridSearchService {
	private readonly vectorSearch: VectorSearch;

	constructor(
		private readonly db: Database.Database,
		private readonly searchService: SearchService,
		private readonly embeddingProvider: EmbeddingProvider | null
	) {
		this.vectorSearch = new VectorSearch(db);
	}

	/**
	 * Execute a hybrid search combining FTS5 and (optionally) vector search.
	 *
	 * When `embeddingProvider` is null or `alpha` is 0, the method returns
	 * FTS5 results directly without embedding the query.
	 *
	 * @param query   - Raw search string (preprocessing handled by SearchService).
	 * @param options - Search parameters including repositoryId and alpha blend.
	 * @returns Ranked array of SnippetSearchResult, deduplicated by snippet ID.
	 */
	async search(
		query: string,
		options: HybridSearchOptions
	): Promise<SnippetSearchResult[]> {
		const limit = options.limit ?? 20;
		const alpha = options.alpha ?? 0.5;

		// Always run FTS5 — it is synchronous and fast.
		const ftsResults = this.searchService.searchSnippets(query, {
			repositoryId: options.repositoryId,
			versionId: options.versionId,
			type: options.type,
			limit: limit * 3 // wider candidate pool for fusion
		});

		// Degenerate cases: no provider or pure FTS5 mode.
		if (!this.embeddingProvider || alpha === 0) {
			return ftsResults.slice(0, limit);
		}

		// Embed query and run vector search.
		const embeddings = await this.embeddingProvider.embed([query]);

		// Provider may be a Noop (returns empty array) — fall back gracefully.
		if (embeddings.length === 0) {
			return ftsResults.slice(0, limit);
		}

		const queryEmbedding = embeddings[0].values;

		const vectorResults = this.vectorSearch.vectorSearch(
			queryEmbedding,
			options.repositoryId,
			limit * 3
		);

		// Pure vector mode: skip RRF and return vector results directly.
		if (alpha === 1) {
			const topIds = vectorResults.slice(0, limit).map((r) => r.snippetId);
			return this.fetchSnippetsByIds(topIds, options.repositoryId, options.type);
		}

		// Build ranked lists for RRF.  Score field is unused by RRF — only
		// the array index (rank) matters.
		const ftsRanked = ftsResults.map((r, i) => ({ id: r.snippet.id, score: i }));
		const vecRanked = vectorResults.map((r, i) => ({ id: r.snippetId, score: i }));

		const fused = reciprocalRankFusion(ftsRanked, vecRanked);

		const topIds = fused.slice(0, limit).map((r) => r.id);
		return this.fetchSnippetsByIds(topIds, options.repositoryId, options.type);
	}

	// -------------------------------------------------------------------------
	// Private helpers
	// -------------------------------------------------------------------------

	/**
	 * Load full snippet + repository data for the given ordered snippet IDs.
	 *
	 * Results are returned in the same order as `ids` so callers receive the
	 * RRF-ranked list intact. Snippets not found in the database (or filtered
	 * out by optional type constraint) are silently omitted.
	 */
	private fetchSnippetsByIds(
		ids: string[],
		repositoryId: string,
		type?: 'code' | 'info'
	): SnippetSearchResult[] {
		if (ids.length === 0) return [];

		const placeholders = ids.map(() => '?').join(', ');
		const params: unknown[] = [...ids, repositoryId];
		let typeClause = '';
		if (type !== undefined) {
			typeClause = ' AND s.type = ?';
			params.push(type);
		}

		const rows = this.db
			.prepare<unknown[], RawSnippetById>(
				`SELECT
					s.id, s.document_id, s.repository_id, s.version_id, s.type,
					s.title, s.content, s.language, s.breadcrumb, s.token_count,
					s.created_at,
					r.id   AS repo_id,
					r.title AS repo_title
				FROM snippets s
				JOIN repositories r ON r.id = s.repository_id
				WHERE s.id IN (${placeholders})
				  AND s.repository_id = ?${typeClause}`
			)
			.all(...params) as RawSnippetById[];

		// Build a map for O(1) lookup, then reconstruct in rank order.
		const byId = new Map<string, RawSnippetById>();
		for (const row of rows) {
			byId.set(row.id, row);
		}

		const results: SnippetSearchResult[] = [];
		for (const id of ids) {
			const row = byId.get(id);
			if (!row) continue;

			results.push(
				new SnippetSearchResult({
					snippet: SearchResultMapper.snippetFromEntity(
						new SnippetEntity(row),
						new SnippetRepositoryRef({ id: row.repo_id, title: row.repo_title }),
						0
					).snippet,
					score: 0,
					repository: new SnippetRepositoryRef({ id: row.repo_id, title: row.repo_title })
				})
			);
		}

		return results;
	}
}