@filter-enginez/mock-data

Three mock datasets at ~10,000 records each, plus MSW handlers that serve both the example apps in development and the test suites. Isomorphic.

| Dataset | Default tier | Full tier (opt-in) | | ----------- | ------------ | ------------------------------------ | | Real estate | ~10,000 | ~50,000 | | Wine | 10,000 | 150,000, generated on demand in ~1 s | | Books | ~10,000 | ~100,000 |

Every record this package ships is synthetic, produced by seeded procedural generators shaped by public reference distributions. No third-party data is included. Generation scripts live in scripts/. Run pnpm regen to validate the default tier and pnpm regen --full to write the real-estate and books full tiers (gitignored). The wine full tier needs no file: it is generated where it is used, including inside a browser Web Worker.

Real estate

Each Listing carries its position twice, with the same values: coordinates as { lat, lon }, and coords as { lat, lng }, the shape GeoProximityRule and the engine's R-tree read. Name coords as the rule's field and the filter round-trips through exportConfig → JSON → loadConfig, a URL, a worker or a server; a function extractor over coordinates cannot, because a function field serializes as a placeholder.

import { createFilterEngine, GeoProximityRule } from '@filter-enginez/core';
import { type Listing, listings } from '@filter-enginez/mock-data/real-estate';

const engine = createFilterEngine<Listing>({ sourceData: listings });

engine.addRule('near-sf', new GeoProximityRule('coords', { center: { lat: 37.7749, lng: -122.4194 }, radiusKm: 25 }));

locationId (state/metro/neighborhood) is stored for the same reason: a hierarchical location filter can name it by string.

Wine

The wine generator samples aggregate statistics measured from the Kaggle "Wine Reviews" dataset (129,907 reviews), never its records: no review text, wine title, winery name or taster name is included. What it reproduces:

• Origin and varietal, jointly. All 1,619 country → province → region → subregion leaves, all 701 varietals and the 9,663 pairings between them, at their measured shares: Napa Valley is mostly Cabernet Sauvignon, Barolo is Nebbiolo, Champagne is sparkling. At full-tier size every leaf, varietal and pairing appears, the rarest at least once.
• Scores and prices. The measured 80–100 point histogram and price-by-score curve, with each country's and varietal's lean (Austrian wines score higher, Chilean wines cost less) and the long tail of expensive bottles.
• Vintages. The measured age profile per color, anchored so the youngest vintage is 2024; non-vintage wines (most Champagne and Port) carry 2025.
• Tasting notes. Each wine's descriptors are chosen at the frequency measured for its varietal, so a text search for "cassis" finds Cabernet and "petrol" finds Riesling about as often as in the real reviews. Higher-scoring wines get longer notes.
• Alcohol, body and sweetness, which the dataset does not record, follow hand-curated domain tables. Red and rosé wines are never sweet or dessert.
• Wineries and designations are invented, in each country's language.

Generation is deterministic, and record i depends only on the seed and i. The default tier is therefore exactly the first 10,000 records of the full tier, and any range can be generated on its own. Only correctly rounded arithmetic is used on the way from seed to record, so Node and every browser engine produce identical records.

The full tier serializes to about 130 MB of JSON, so generate it where the engine lives rather than shipping it. The ./wine/generator sub-export has no loaders, no MSW and no top-level await, which makes it safe to import in a Web Worker:

import { FULL_WINE_TIER_SIZE, generateWineRange, type Wine } from '@filter-enginez/mock-data/wine/generator';

declare function reportProgress(loaded: number, total: number): void; // e.g. a postMessage to the page

const CHUNK = 10_000;
const wines: Wine[] = [];

for (let start = 0; start < FULL_WINE_TIER_SIZE; start += CHUNK) {
  wines.push(...generateWineRange(start, Math.min(start + CHUNK, FULL_WINE_TIER_SIZE), undefined, 'full'));
  reportProgress(wines.length, FULL_WINE_TIER_SIZE);
}

generateWines(FULL_WINE_TIER_SIZE, undefined, 'full') does the same in one call. In Node, loadFullWines() from ./wine returns the full tier too, preferring the local real corpus when one has been imported.

Real wine data (optional, local only)

When you work in this repository, you can swap the synthetic wines for the real Kaggle "Wine Reviews" dataset (zackthoutt, scraped from WineEnthusiast). It is published under its own license, whose terms rule out shipping it in an Apache-2.0 package (the NOTICE.md at the root of the engine repository gives them), and its review text is WineEnthusiast's. That is why it is never committed, packaged, or deployed:

• pnpm regen:wine:real --accept-data-license downloads it to your machine. Without the flag the script prints the terms and stops.
• The output (src/wine/data-real-*.json) is gitignored, and the published package contains compiled code only.
• The example apps' production builds strip it, so anything deployed runs on synthetic data.

Install

pnpm add @filter-enginez/mock-data

Peer dependencies: @filter-enginez/core (same release), msw@^2.0.0 — supplied by your app.

Sub-exports: @filter-enginez/mock-data/real-estate, @filter-enginez/mock-data/wine, @filter-enginez/mock-data/wine/generator (worker-safe, generator only), @filter-enginez/mock-data/books, @filter-enginez/mock-data/handlers.

Documentation

Consumer guides live in the engine repository under docs/guides/:

• extending_rules.md — writing a custom FilterRule that survives serialization
• json_config.md — declaring filters as data, and why the format is safe to accept from a tenant
• acceleration_tuning.md — what the five acceleration toggles are worth, measured
• worker_offload.md — running the engine in a Web Worker

Project home, with six live demos: https://filterenginez.com. Documentation site: https://docs.filterenginez.com.

Apache-2.0. See LICENSE.md and NOTICE.md, both included in this package.