If you find yourself here wondering "what's an LCBO API?", let me explain. In Ontario, Canada all beverage alcohol sales go through a government owned corporation called the Liquor Control Board of Ontario (LCBO) which handles retail and distribution of alcoholic beverages throughout the province. The LCBO has numerous retail stores and a website that hosts a catalog of every product, store, and even inventory levels. They publish a seasonal catalog with recipes, editorials, and other content called Food & Drink. They also contribute billions of dollars of revenue to our public healthcare system annually. It's a fascinating situation when you think about it, other places have similar systems but to my knowledge none have the breadth and depth of the LCBO. So, now you know what it is, pretty cool eh?
This might be interesting to you even if you don't live in Ontario, Canada, if:
- You want access to a lot of data for learning or testing
- You are curious to learn how some types of web crawlers work
- You are curious to learn what a production Rails application spanning 10 years of development might look like
An important notice
Over the entire course of this project I have struggled massively with the idea of accepting financial support. On one hand LCBO API needed it, on the other I was weary of the complications it would cause. Well, now I have THE PERFECT solution to this problem!
I'm undergoing treatment for blood cancer right now, specifically diffuse large b-cell lymphoma. I'm going to write more about that soon somewhere else but during this past year people from everywhere have supported me in every way they could, and it has changed me. I want us to do something big to show that we care too!
If you've ever wanted to support this project in the past, please, make a donation to Hamilton Health Sciences on LCBO API's behalf, they are saving my life.
I am undergoing treatment at the Juravinski Cancer Center, but really you can choose any option, or leave the default one. It doesn't matter if the amount is small or large, they will notify me when you donate. I will tabulate a list to track the total, let's see how much we can raise!
Finally I'd like to make a special mention to my workplace, Crowdmark. They have been incredibly kind and understanding during all of this, and I quite literally would not have been able to do this without them. We are working tirelessly toward advancing the status quo of assessment in higher education. If you care about education and learning, I urge you to check us out.
In the fall of 2008 I was a freshly minted web developer with a few years of experience under my belt. I was hungry for a challenge, and for some recognition. Apps were becoming a thing at the time and I wanted to build one, badly. I decided I wanted to build one that would require me to first build this API. I never did build that app
If you look into this codebase long enough you are likely to find moments of frustration, dead ends, confusing cruft, etc. I really hope you don't focus on that or on any negativity you might find. I'm not that person anymore, and I don't want you to be that person either. I am an open book on this! Open an issue and ask me a question, I will be as honest and respectful as possible, I only ask you do the same.
I'm releasing this project under GNU GPLv3, I think this is the most fair and responsible option for a project like this. If you feel differently, open an issue and we can have a discussion in the open about it. I only ask, respectfully, that you do not reuse the branding and design. I'm fine with re-use of the documentation, but the styling, identity, and branding must be changed if you want to deploy your own siloed version of this app.
Have big dreams!
What if instead of a monolithic application trying to do everything in one style, in one place, we thought bigger? What if the crawler was a separate project again, responsible for collecting and normalizing data, others could build API nodes in whatever platforms they wished, those nodes would register with the data provider and receive updated data as it was made available and provide that data to users of all different types.
Instead of dozens of similar API servers trying to do the same thing, fighting for LCBO.com's resources we could focus our efforts on building value on this data instead of fighting for ownership over it. We could get other disciplines involved to generate new value beyond the obvious, get the craft beer and wine communities involved, build on this, make it bigger than just Ontario.
I don't know how feasible something like that would be, but I know if others are interested I would love to have these discussions.
Also, maybe we should consider charging corporate users a reasonable fee to access the API nodes, that money could be used to fund the hosting costs to keep things sustainable, it could also be used to fund support programs for people who can't drink, or who don't want to drink, or who want to drink less, to give back to our communities and to actually make a difference.
But I need others to help share the burden of maintaining and managing all of this, my and my family's health and happiness is priority #1, followed by my career, and then my friends and community. This project can't be my #1 time sink anymore, it's not sustainable for me, and it's not healthy for me. But if you are inspired by this message, please reach out to me, ideally in the open, but privately at first is okay too.
I hope this excites you!
I couldn't help myself, I wrote more about my ideas on this: doc/lcboapi-proposed.md
Now, with that out of the way, we can start getting into who I really did this for, and what got me excited and inspired to do this in the first place: the opportunity to learn and grow and to help others do the same. For those of you out there who are curious, let's see where this goes
Running the Rails app
You can probably run the app directly on your host environment, it doesn't require anything too fancy as far as system dependencies are concerned. I develop on Apple hardware, if you do too, you may have success using Postgres.app, and Homebrew for installing Redis. Otherwise, you can use Docker.
If you have experience with another platform, please make a PR or an issue and we can work at adding your platform to the README.
Also, if what follows here makes no sense to you, open an issue, maybe we could do a screencast to demonstrate the process, or maybe someone out there who's good at that would take that on?
What I describe below is only one way to set up a development environment to run LCBO API on your computer. If others have improvements (there's room for many, an entrypoint script to bootstrap the dev database for instance) or even different approaches, like using Vagrant + VirtualBox, open an issue or a PR, I'm happy to add them.
If you want to help, I want to enable you.
First, you'll need to set up some configuration which is not provided in the public repository. The reason this is done is to protect private data such as API keys and secret tokens, but also because some developers may prefer slightly different settings for their personal preferences and things like that.
There are a couple files you'll need to create,
.env. There are template versions in the repo under
.env.example, you can copy those files to get started:
cp config/secrets.yml.example config/secrets.yml cp .env.example .env
If you are just wanting to boot the app and access it locally, you should be good to go at this point. If you want to be able to use the crawler and have it save a snapshot saved to Amazon S3, you'll need to add your AWS credentials and bucket to
The rest of the settings either only really matter in a production environment, are not really used, or only matter if you don't like the default preference. As always, if you need clarification, open and issue and I'm happy to help.
Getting the app running for the first time
First, you'll need to install the Docker client for your system, you can find out about that here. Once you've installed Docker, you can get things started:
Next, you will need to build the containers:
When that is done, you can boot up the whole thing by issuing:
At this point, you don't have any data in the database, so if you load the app, http://localhost:3000, it won't do much, it serves data after all, and there's no data in it. So let's do something about that.
Go ahead and shut down the containers:
That means, press the
C keys simultaneously.
You can download an archive of the latest production database dump from my personal Amazon S3 account here. Please note that there are sensitive tables (emails, users, keys) and that data has been excluded from this file.
Download and extract the archive in the
tmp directory of this project:
cd tmp curl -O https://s3.amazonaws.com/heycarsten/lcboapi-2018-12-17.bz2 tar xzf lcboapi-2018-12-17.bz2 cd ..
The file is about 180MiB, so it might take a while to download depending on your connection speed (this happens on the line that starts with
Once you've downloaded and extracted the database file, you can load the data into the database:
docker-compose run --rm app rake db:create docker-compose run --rm app bash -c 'pv tmp/lcboapi-2018-12-17.sql | psql -q -h db -U $POSTGRES_USER $POSTGRES_DB > /dev/null'
The first line, ending in
rake db:create will create the database schemas in Postgres for development and testing, the second line will load the database dump into the development database. The progress bar indicates how much of the data has been piped into the database, once that completes indexes will be built. This might take some time depending on your machine, it's a fair amount of data. Then you can fire up the app again:
You can also safely delete the archive and extracted SQL file from your
tmp directory too at this point.
If you're finding typing
docker-composeover-and-over tedious, look into shell aliases
You can add an alias line to your shell profile like
alias dc=docker-composeand then you can just type
dcinstead of having to type
Think of other aliases you could create to improve on this even further
Now, navigate to http://localhost:3000/products/438457
Boom. You've got LCBO API running on your computer!
Running the app from now on
When you're done working on the app, just issue
Ctrl+C to shut everything down. The next time you want to work on it again, run
docker-compose up and you're good to go!
Rebuilding (bundle install)
If you add a new gem to the
Gemfile, you will need to re-install the packages and update the dependencies. Docker is pretty good at doing this, it can tell when
Gemfile changes, and it knows to rebuild the
app container for you.
Opening a Rails console
To fire up a Rails console and inspect objects inside the application:
docker-compose exec app rails c
Once that's running you can do stuff like:
Fetch the first product in the database:
Find LCBO retail store #25 (my local store):
If you change code in the app, you'll need to issue the
reload! command in the console to refresh the changes.
Inside of the
spec folder, you'll find the test suite for LCBO API. It is regrettably not comprehensive, but it's not too bad either. I have struggled my entire career to maintain test suites that I am satisfied with. We should improve these tests!
To run the test suite:
docker-compose exec app rspec
You'll see a bunch of green dots
., each one of those represents a passed test case. That's good. If a failure occurs you'll see a red
F, that's bad... JUST KIDDING! Actually it's good! Tests give you the power to change things in an existing codebase and see if you cause any regressions to existing functionality. Of course nothing is perfect, but I can tell you without a doubt, from experience, tests are good.
As applications get bigger and bigger and more and more complex not having tests becomes a literal nightmare, it makes changing your application and adding features an extremely brittle process. Things like using languages with type systems and other various different programming paradimgs can go a long way to help with this too, but I am of the opinion that there really is no replacement for at least a solid acceptance test suite.
This is the part of LCBO API that kind of makes the whole thing possible. Crawlers for complicated websites are hard to build and maintain. The first version of LCBO API had a full test suite for the crawler, when everything changed many years ago I gave up on that codebase and just build something as fast as I could in this one.
The crawler logic is located in
lib/crawler.rb, from there you'll see all of the various tasks that happen in succession to encompass a complete crawl of the LCBO websites.
The parser logic is located in
lib/lcbo.rb and all of the various files within
lib/lcbo/*, this includes all of the various requests that need to happen, and the code responsible with turning the data from those requests into structured data that can ultimately go into the database.
I designed the crawler to perform requests in serial, this is a very good approach to take when you are crawling one website. It's simple for one which is always the best route to take if you can, and it's polite. We could fire off n AWS Lambda jobs and crawl every page on LCBO.com in a few seconds, but that would be rude, and we'd probably DDoS their website momentarily, not good.
Manager app (
This encompasses an Ember application, when you sign up/in to LCBO API and generate API keys, this is what you're interacting with. It's quite outdated, I haven't tried building it. I've been using Ember since day zero so if you have any questions about this, make issues. I'd actually quite enjoy discussing this part of LCBO API and working with you all to improve it.
Static site (
This contains a Middleman site, when you visit lcboapi.com this is what you're looking at. It also contains a very small (also outdated) React app which is the "Give it a try" thingy on the right of the homepage. It has it's own Gemfile and build script
static/generate, when that is run it builds the site and syncs the changes into the
public folder. In Rails apps the
public folder is served as static content.
There are A LOT of dead ends in this codebase, branches that went 40-60-80% of the way to a feature and then stagnated, experiments, etc. As always if you find something that's got you like
It would also be cool if we could tie up some of the dead ends in here, I'd be super interested in finally getting JSON:API and GraphQL added in. The current API response design is from 2008!!! In a way I'm kind of surprised that nobody ever complains about it.
The #1 by far most requested feature I never implemented was categories, it's sort of in here, I forgot why I never shipped it, I don't remember what final stuff had to fall into place, but maybe that would be a good first thing to tackle? There's also Producers, and Origins which never quite got wrapped up.
I also have a whole bunch of other repos and little experiments I made over the years, I was always fascinated by the idea of inventory level prediction, somewhere there's a dataset dump analysis tool written in Go for analyzing the CSV dumps for a particular product inventory over a set of time. I'd be happy to release that stuff too if there's interest.
WIP (Work in progress)
I'm going to leave it here for now and I'm going to wait to hear back from you. I would love to keep adding to this base of knowledge in whatever way people want to see (screencasts, interviews, inline documentation, etc.) I also want you to do the same, if you're not sure, ask.