Files

T

nngo c2064b84b4 06/15 Phase 1 + 2 codes

2026-06-15 11:23:05 -04:00

6.1 KiB

Raw Blame History

Classifieds — Phase 1 (Foundation)

App factory, config, extensions, users/plans schema, full auth (register / login / logout / email-verify / password-reset), RBAC, i18n scaffold (EN/VI/ES) with accent-insensitive normalizer, Turnstile hook.

Stack

Flask · MySQL 8.0 · Redis · Gunicorn · systemd · Nginx · Ubuntu 22.04

Local setup

python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env          # edit secrets

Quick smoke test (no MySQL/Redis needed)

A committed test exercises the whole Phase-1 surface against a temp SQLite file with the limiter in-memory:

python -m tests.test_smoke      # or: pytest -q

It checks: plan seeding, accent normalizer (phở→pho, ñandú→nandu), register → email-verify (+5 trust) → login/logout, password reset round-trip, duplicate-email and bad-password rejection, home/healthz/404.

Note: it runs over https://localhost with a Referer header on purpose — production sets PREFERRED_URL_SCHEME=https, so Flask-WTF enforces the secure referrer CSRF check. Real browsers send Referer; the test mirrors that.

Manual run with the dev server

Flask-SQLAlchemy resolves a relative sqlite:///dev.db against the instance/ folder, not the CWD. To reset, delete instance/dev.db (or use an absolute DATABASE_URL=sqlite:////abs/path/dev.db).

Set in .env: DATABASE_URL=sqlite:////tmp/classifieds_dev.db (absolute), then:

export FLASK_APP=wsgi:app
flask db init
flask db migrate -m "phase1 users+plans+trust"
flask db upgrade
python seed.py --admin admin@example.com 'StrongPass123'
flask run

Open http://127.0.0.1:5000

Real MySQL 8.0

CREATE DATABASE classifieds CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
CREATE USER 'classifieds'@'127.0.0.1' IDENTIFIED BY 'change-me';
GRANT ALL PRIVILEGES ON classifieds.* TO 'classifieds'@'127.0.0.1';
FLUSH PRIVILEGES;

Leave DATABASE_URL blank in .env (URI is built from DB_*). Then run the same flask db ... + seed.py sequence.

i18n (translations)

pybabel extract -F babel.cfg -o messages.pot .
pybabel init -i messages.pot -d app/translations -l vi
pybabel init -i messages.pot -d app/translations -l es
# ...translate the .po files...
pybabel compile -d app/translations

Re-extract + pybabel update after adding new _() strings.

What works now

Register → email verify link (dev: printed to console) → verify grants +5 trust.
Login / logout, remember-me, rate-limited, suspended-account block.
Password reset (enumeration-safe), signed expiring tokens.
Language switch persists to session + profile.
Argon2 hashing, CSRF on all forms, RBAC decorators ready.
Free plan auto-assigned on registration.

Notes / Phase-1 limits

Turnstile bypassed when keys blank (dev). Set keys in .env for prod.
Email prints to console until SMTP (Brevo) configured.
requirements.txt uses PyMySQL. For the SQLite smoke test no driver needed.

Production deploy (outline)

Code → /opt/classifieds, venv, pip install -r requirements.txt.
.env with prod secrets, FLASK_CONFIG=prod, real SECRET_KEY.
MySQL DB + user (above). flask db upgrade && python seed.py.
deploy/classifieds.service → /etc/systemd/system/, systemctl enable --now classifieds.
deploy/nginx.conf.sample → adapt, enable site, add TLS (certbot), reload Nginx.
Redis running for sessions + rate-limit + (later) queue.

Phase 2 — Listings core (built)

New models: Category (self-referential subcategories + field_schema JSON), Listing (+ title_norm, denormalized hot columns, lat/lng, status, expiry), ListingImage, ZipGeo, Metro.

Features:

Listing CRUD with owner/moderator authorization; mark-sold; my-listings.
Category-specific fields driven by field_schema (text/number/select/bool), validated server-side; dynamic form toggles fields by selected category.
Tier enforcement: active-listing cap, images-per-listing cap, listing life (expires_at) — all read from plans.config.
Image pipeline (Pillow): validate, re-encode to JPEG (strips EXIF/metadata), thumbnail, randomized filenames, per-listing media folder.
Browse + filters (category, state, price range, condition/job_type hot fields).
Accent-insensitive keyword search via title_norm (phở matches "pho").
Radius search: ZIP geocode → bounding-box SQL prefilter → exact haversine refine.
Expiry sweep: flask expire-listings (wire the systemd timer below).

Expiry sweep (systemd timer)

cp deploy/classifieds-expire.{service,timer} /etc/systemd/system/
systemctl enable --now classifieds-expire.timer

Seed

python seed.py now also seeds the 6 categories + subcategories and a small sample of zip_geo rows. Replace the ZIP sample with the full dataset (SimpleMaps US ZIP or Census ZCTA, ~42k rows) before production.

MySQL upgrades (optional, post-Phase-2)

The Phase-2 code is portable (runs on SQLite for tests, MySQL in prod). Three spots can be upgraded to native MySQL features when you want them:

Spatial radius — replace lat/lng bounding-box + haversine with a generated POINT column + SPATIAL INDEX and ST_Distance_Sphere:

ALTER TABLE listings ADD COLUMN geo POINT
  GENERATED ALWAYS AS (ST_SRID(POINT(lng, lat), 4326)) STORED,
  ADD SPATIAL INDEX spx_listings_geo (geo);

Generated hot columns — instead of app-maintained attr_* columns, derive them from JSON:

ALTER TABLE listings ADD COLUMN attr_condition VARCHAR(40)
  GENERATED ALWAYS AS (JSON_UNQUOTE(JSON_EXTRACT(attributes,'$.condition'))) STORED,
  ADD INDEX ix_attr_condition (attr_condition);

Full-text search — add a FULLTEXT index and switch the keyword filter to MATCH ... AGAINST for relevance + speed:
```
ALTER TABLE listings ADD FULLTEXT INDEX ftx_listings (title, body);
```
(Keep title_norm for accent-insensitive matching; combine as needed.)

Next: Phase 3 — Messaging + favorites

conversations, user-to-user messages, saved listings, contact masking gated by trust tier.

6.1 KiB Raw Blame History