getagun.ca

Map of personally-verified Canadian firearms retailers, with private nearest-retailer search (all matching happens in your browser — no tracking), a fully bilingual EN/FR interface, and a guide to getting your PAL.

Built with React, MapLibre GL (pmtiles basemap), and Cloudflare Workers + D1.

Features

• Category filters (independents, Canadian Tire, Co-op, FCNQ, and other chains), shareable as links via ?categories= — e.g. getagun.ca/?categories=canadian-tire
• Density view: a per-capita choropleth of gun stores per 100,000 residents by 2021 census division, computed live in the browser and responsive to the active category filters; click a division for its rate
• Live "stores in view" counter, pin clustering toggle, five basemap themes, fullscreen mode
• Published data under /sheets/: browsable spreadsheets of all mapped retailers by category and which Canadian Tires stock firearms, with .xlsx downloads and charts — all available in French at the same address plus -fr
• FAQ with rich-text answers and a suggestion form (Turnstile-protected)
• Door-to-door flyer canvassing at /admin/canvass: every residential address in London, Ontario, a five-point reaction scale, and resident contact details, installable to a phone and usable with no signal

Development

npm install
cp .env.example .env       # defaults work for dev
npx wrangler d1 execute getagun --local --file=schema.sql
npx wrangler dev           # API on :8787
npm run dev                # site on :5173, /api proxied

.dev.vars needs TURNSTILE_SECRET (Cloudflare's always-pass test secret 1x0000000000000000000000000000000AA) and DEV_SKIP_ACCESS=1 to bypass Cloudflare Access locally — Access verification otherwise fails closed.

Deploy

npm run deploy

Canvassing

/admin/canvass is a field tool for handing out PAL/RPAL flyers: pick a street or pull the streets around you, walk it, and log each door. It sits behind the same Cloudflare Access app as the rest of /admin, and resident names, phone numbers and emails live only in D1 behind that gate.

The address list is built from two City of London open data layers — 143,124 unit-level address points joined to 6,773 zoning polygons, so each door carries its land use and the list can default to residential. Rebuild it with canvass-prep.py, which writes public/canvass/ (gitignored, ~15 MB).

Add the table before deploying the code. Run both from the repo root, so wrangler picks up wrangler.jsonc and targets the right database:

npx wrangler d1 execute getagun --local  --file=schema.sql
npx wrangler d1 execute getagun --remote --file=../getagun-tools/canvass-migration.sql

--file is resolved against the working directory, not the config, so run these from the repo or pass an absolute path.

Notes:

• Doors are written to the phone first and queued, so nothing is lost in a dead zone; the queue flushes on reconnect. Writes are last-write-wins on a timestamp, so a door queued offline cannot overwrite a newer edit made elsewhere.
• public/admin/canvass/sw.js is scoped to /admin/canvass/ and never touches the public site. It has no precache list, so offline works from the second visit onward, and the first load after a deploy still serves the previous build before updating.
• The basemap is fetched as byte ranges and is not cached: the list, every street and all logging work offline, but the map needs signal.

Requires: D1 database getagun (schema.sql applied --remote), Turnstile site + TURNSTILE_SECRET secret, a Cloudflare Access app covering getagun.ca/admin and getagun.ca/api/admin with ACCESS_TEAM_DOMAIN and ACCESS_AUD vars set, an R2 bucket bound as TILES containing canada.pmtiles, and a www.getagun.ca → apex redirect rule in the zone.

Data sources

Basemap © OpenStreetMap contributors, served as a self-hosted pmtiles extract. Census division boundaries and populations for the density view are adapted from Statistics Canada's 2021 Census (Boundary Files and Population and Dwelling Counts), used under the Statistics Canada Open Licence. Retailer data is my own verification work.