import { DocumentOptions, Document, DocumentSearchResult } from 'flexsearch';
import { d as IndexedDocument } from './types-DZE2UBNA.js';

/**
 * `@icjia/pdf-search-index/flexsearch` — alternative search-engine
 * adapter for [FlexSearch](https://github.com/nextapps-de/flexsearch).
 *
 * **When to use this instead of `/fuse`.** FlexSearch hits its stride
 * around the corpus size where Fuse starts to slow down. Pick by
 * corpus:
 *
 *   < 2,500 docs       → Fuse.js (better typo tolerance, snippet
 *                                  positions, simpler API).
 *   2,500 – 10,000 docs → **FlexSearch** (sub-millisecond queries,
 *                                  denser index, built-in WorkerIndex).
 *   10,000+ docs       → Pagefind (chunked on-demand index — see the
 *                                  `/pagefind-emit` entry).
 *
 * **Tradeoffs vs Fuse:**
 *
 * - **No native typo tolerance.** FlexSearch's `tolerant: true` is
 *   n-gram-based; it works but needs tuning. If "applicent" → "applicant"
 *   is core to your UX, stay on Fuse.
 * - **No native match-position output.** FlexSearch returns matched
 *   document IDs, not character-range indices. Snippet highlighting
 *   becomes a substring-search problem; we ship
 *   `snippetHTMLForFlexMatch` to handle the common case.
 * - **Encoded index format.** FlexSearch's `index.export()` produces a
 *   denser-than-JSON serialization. We use this directly in the demo's
 *   "Inspect the prebuilt index" panel so you can compare it side-by-
 *   side with the Fuse equivalent.
 *
 * Added in 1.3.
 */

/**
 * Default FlexSearch Document options tuned for the `IndexedDocument`
 * shape. Mirrors the spirit of `DEFAULT_FUSE_OPTIONS` from `/fuse` —
 * sensible defaults that work for the common case (search across title
 * + text, store url/title/format/pages for result rendering).
 *
 * Override per call via the `options.flexOptions` argument.
 */
declare const DEFAULT_FLEX_OPTIONS: {
    readonly document: {
        readonly id: "id";
        readonly index: readonly [{
            readonly field: "title";
            readonly tokenize: "forward";
        }, {
            readonly field: "text";
            readonly tokenize: "forward";
        }];
        readonly store: readonly ["url", "title", "format", "pages"];
    };
    readonly encode: "icase";
};
/**
 * Build a FlexSearch Document index from `IndexedDocument[]` rows.
 *
 * Pure-Node usage:
 *
 * ```ts
 * import { createFlexSearchIndex } from '@icjia/pdf-search-index/flexsearch';
 *
 * const rows = await indexDocuments(urls);
 * const index = await createFlexSearchIndex(rows);
 *
 * // FlexSearch search API
 * const results = await index.search('stigma', { enrich: true });
 * ```
 *
 * Browser usage: build the index in the consumer's bundle the same way.
 * Or use FlexSearch's built-in `WorkerIndex` for off-main-thread search;
 * see https://github.com/nextapps-de/flexsearch/blob/master/doc/worker.md.
 *
 * @param rows  Documents to index.
 * @param options  Optional FlexSearch DocumentOptions override; merged
 *                 onto `DEFAULT_FLEX_OPTIONS`.
 */
declare function createFlexSearchIndex(rows: readonly IndexedDocument[], options?: {
    flexOptions?: DocumentOptions<IndexedDocument, false>;
}): Promise<Document<IndexedDocument, false>>;
/**
 * Render an HTML snippet around the first occurrence of `query` in the
 * matched document's text. FlexSearch returns matched docs but not
 * character positions, so we do our own case-insensitive substring
 * search to locate the highlight.
 *
 * Output shape matches `snippetHTMLFor` from `/snippet` — same
 * `<mark>`-wrapped highlight, same `…` leading/trailing markers, same
 * HTML escaping for source text.
 *
 * @param row           The matched IndexedDocument (from FlexSearch's
 *                      enriched result or your own row lookup).
 * @param query         The user's search query.
 * @param options       contextChars (default 80); matchKey (default 'text');
 *                      collapseWhitespace (default true).
 */
declare function snippetHTMLForFlexMatch(row: IndexedDocument, query: string, options?: {
    contextChars?: number;
    matchKey?: 'title' | 'text';
    collapseWhitespace?: boolean;
}): string;
/**
 * Helper: pull `IndexedDocument` rows out of a FlexSearch enriched
 * search result. FlexSearch's result shape is:
 *
 * ```
 * [ { field: 'text', result: [{ id, doc }, ...] }, ... ]
 * ```
 *
 * — one entry per searched field with a per-field result array. This
 * helper flattens + dedupes by `id` and returns the underlying rows in
 * the order they first appear.
 */
declare function flattenFlexResults<T extends IndexedDocument>(results: DocumentSearchResult<T, false>): T[];

export { DEFAULT_FLEX_OPTIONS, IndexedDocument, createFlexSearchIndex, flattenFlexResults, snippetHTMLForFlexMatch };