NYC Eats 🍴

A static-site dashboard mapping every NYC restaurant and bar, built from two open data sources.

Data sources

Source	Records	API
DOHMH — NYC Dept. of Health Restaurant Inspection Results	~30 k	Socrata 43nn-pn8j
SLA — NYS Liquor Authority Active Licenses (NYC only)	~22 k	ArcGIS FeatureServer

Sources are pluggable — drop a new DataSource subclass into sources/ and it's automatically discovered at build time.

Merge pipeline

Venues appearing in both datasets are merged into a single marker with source="both". The pipeline runs three passes:

1. Exact address + borough — normalised addresses (suffix canonicalization, ordinal stripping, AKA/unit removal) are compared by borough.
2. Address range containment — SLA range addresses like "77 79 HUDSON ST" match a DOHMH entry at 77 HUDSON STREET if the number falls within the range. Queens block-lot addresses (30-12 20TH AVE) are excluded.
3. Geo-proximity ≤ 30 m — remaining unmatched venues within 30 metres are merged using a grid-based spatial index. Catches typos, AKA addresses, and other text variations.

Before merging, DOHMH venues are deduplicated:

• First by camis (unique restaurant ID — the dataset has one row per inspection).
• Then by normalised name + BIN (Building Identification Number) to collapse duplicate registrations in the same building.

Typical result: ~51 k raw → ~37 k post-merge, with ~14 k merged pairs. Exact counts are shown in the sidebar's Pipeline table.

Map features

• Leaflet 1.9 with CARTO light base tiles and MarkerCluster
• Grade-letter markers (A/B/C/P/Z) for DOHMH & merged venues
• Martini-glass markers for SLA-only venues
• Purple merged markers showing both names in the popup
• Spiderfying on every zoom level for overlapping markers
• Filter by source, cuisine tag, and search by name
• Cache-busted venues.js via SHA-256 content hash

Project structure

nyc-eats/
├── build.py              # Static site generator (fetch → merge → render)
├── Makefile              # Build, deploy, cron targets
├── requirements.txt      # Python deps (requests, jinja2)
├── nginx.conf.in         # nginx config template (values from .env)
├── .env.example          # Server-specific settings template
├── cron/
│   ├── nyc-eats-refresh            # Refresh script (runs on server)
│   ├── nyc-eats-refresh.service.in # systemd service template
│   ├── nyc-eats-refresh.timer      # systemd timer unit
│   └── README.md                   # Cron/timer setup instructions
├── sources/
│   ├── base.py           # Venue dataclass + DataSource ABC + auto-discovery
│   ├── dohmh.py          # DOHMH Socrata source
│   └── sla.py            # SLA ArcGIS FeatureServer source
├── templates/
│   └── index.html        # Jinja2 template (Leaflet map + sidebar)
├── static/
│   ├── style.css         # Dashboard styles
│   └── favicon.svg       # 🍴 favicon
└── dist/                 # Generated site (git-ignored)

Quick start

# Clone & set up
git clone https://github.com/YOUR_USER/nyc-eats.git
cd nyc-eats
cp .env.example .env     # edit with your server details
make install              # creates .venv, installs deps

# Build (fetches live data, ~30s)
make build

# Or use cached data if already fetched in the last 24h
make build-cached

# Local dev server
make serve            # http://localhost:8000

Deployment

The site can be deployed to any VPS with nginx. Server-specific settings (hostname, paths) live in .env (see .env.example).

make deploy           # build + rsync + reload nginx
make deploy-only      # rsync + reload (skip rebuild)

Server setup (one-time)

# On the VPS — create the serving directory:
mkdir -p ~/your-domain.com/dist

# Symlink nginx config (done automatically by `make deploy`)
sudo ln -sf ~/your-domain.com/nginx.conf \
  /etc/nginx/sites-enabled/your-domain.com.conf
sudo nginx -t && sudo systemctl reload nginx

# HTTPS
sudo certbot --nginx -d your-domain.com

Automated refresh

The data is refreshed every Sunday at 3:00 AM ET via a systemd timer on the server. See cron/README.md for full details.

make timer-install    # uploads units + enables timer on VPS

The refresh script lives at cron/nyc-eats-refresh — it rebuilds the site and rsyncs the output to the serving directory. Logs go to $VPS_PATH/refresh.log.

Adding a data source

1. Create sources/my_source.py
2. Subclass DataSource (from sources.base)
3. Implement name, description properties and fetch() → list[Venue]
4. Run make build — it's auto-discovered

License

Unlicensed