In 1970 a paper came out that changed the world of business computing – A Relational Model of Data for Large Shared Data Banks by E. F. Codd. It laid the ground work for modern databases.
So I just finished reading the OG relational model paper from 1970 and wow you have no idea how much it’s done for us.— Swizec Teller writing a book you’ll wanna read (@Swizec) September 9, 2022
The state of the art back then was wild 🤯 some highlights 👇 pic.twitter.com/VF6Kl9xpqg
Reading from 2022, the world described in Codd's paper is almost unimaginable. He talks about the biggest challenges in data management systems:
- data coupled to its program
- programs coupled to how data is stored
- zero flexibility when reading data
The wild part is that before 1970 it was common, even expected, that your data store and your code were inseparable.— Swizec Teller writing a book you’ll wanna read (@Swizec) September 9, 2022
Wanna optimize storage? Gotta rewrite all your code
Need more space? Rewrite your code
Add networking? Rewrite
New type of data? Rewrite
New query? Change data
You had to write complex data transformation programs for any analysis that needed data in a new shape. Like seeing how many comments a user made on all articles, for example.
And forget about slapping more memory or disk space in your server when data grows too large. You'd have to re-optimize the entire data system to take advantage.
The only people who can imagine such a world are NoSQL users 😛
Codd suggests his relational model can support a system that gets around those common flaws. The paper does not go into implementation detail.
His core thesis is that:
- Data retrieval should be independent of its representation
- Data semantics shouldn't depend on storage concerns
- Different programs should be able to read the same data
The major insight from Codd's paper is that how you think about data – its abstract model – should be separate from how it's stored. Worry less about arrays and trees, think about what the data means.
The paper’s core thesis is that:— Swizec Teller writing a book you’ll wanna read (@Swizec) September 9, 2022
- data retrieval should be independent of its representation
- data semantics shouldn’t depend on how it’s stored
- different programs should be able to read the same data
You know, things we take for granted now :D
Once you've separated data semantics from data representation, you unlock a wonderful possibility: An intermediate language can be constructed that asks the data system for data based on its properties, and the system figures out the details.
Instead of "Read 50 bytes from position A" you could say "Get article with id=2". We now call this a database.
Based on all this an “experienced systems programmer” could implement a system that— Swizec Teller writing a book you’ll wanna read (@Swizec) September 9, 2022
- manages data for you
- uses a simple language for data retrieval
- guarantees consistency at expense of speed
- hides data storage details from the user
We now call this a database
You've likely learned the relational model by osmosis if nothing else. It's the default way we think about data these days.
I found that Codd's original paper explained his model more clearly and concisely than all my college professors and years of experience. Maybe it's that I now know what to look for and 10 years ago it would fly right over my head.
Note that many things you do to optimize a relational database or make SQL easier to write do not fit Codd's original model. Denormalization, for example, goes strictly against everything Codd stood for.
However, SQL databases deviate from the relational model in many details, and Codd fiercely argued against deviations that compromise the original principles.
Codd describes data relations as named sets
R of columns
A where each column describes a nameable property of the relation. Each row holds a distinct tuple of column values.
Columns belong to a table, which means different tables can share columns of the same name. You can leverage this to represent relationships between domains. An identity column in
R1 can be reused in
R2 to mean "belongs to".
R1:user name1 Swiz2 KiwiR2:article title user1 sdf 12 dfs 1
user field in one table matches the
user field in another. Now you can store user info about each article without multiplying user data.
This is known as the normal form:
- Has a primary key (no duplicates)
- Columns cannot be turned into a table (atomic)
Because every relation is a set, standard set operations apply:
- permutations can shuffle the ordering of columns without changing their meaning. Useful for optimizing storage and reading patterns
- projections let you select partial data out of a table, conceptually this is a new relation
- joins bring the relational model's true power – you can create new relations by joining stored relations
For example if you join
R2 from before, you get a new relation
R':article title user name1 sdf 1 Swiz2 dfs 1 Swiz
Because this is a relation, you can continue to do further joins, projections, permutations, and the rest.
- composition composes two relations much like a join, but unlike with a join, two relations are composable iff only one possible join exists. The example above was composable.
- restriction lets you get a subset of relation
Rbased on a relation
S– like saying "Give me all rows that have these 2 column values
My favorite part made me think of @davidkpiano and his gripes with redundant state in React components— Swizec Teller writing a book you’ll wanna read (@Swizec) September 9, 2022
Strong redundancy – state for data deriveable from named relations, makes code worse
Weak redundancy – state for data deriveable from joins, may be helpfully convenient
Codd distinguishes between storage redundancy for performance reasons and redundancy in your data model.
It can be useful, for example, to store 2 copies of a table next to data that's often read together (like joining users on articles and on comments), but you don't want to replicate that in your data model. Let "the system" deal with those details.
He defines the concept of strong redundancy as a set of relations that can always be derived from other data. Like storing a count of comments with every article.
This can be useful for convenience, but is never required.
Weak redundancy means you can get the same information in multiple ways. This is unavoidable as your data grows in relationships.
The relational model on its own does not guarantee data consistency or anything like that. The system has to implement those at the cost of performance.
The model enables guarantees by creating followable rules and separating implementation details from data modeling concepts.
And that's bloody cool.
PS: a common problem with NoSQL, despite its many performance benefits, is that it breaks the separation between implementation details and data modeling. Best added to your stack once data usage patterns are known and speed matters.
I write articles with real insight into the career and skills of a modern software engineer. "Raw and honest from the heart!" as one reader described them. Fueled by lessons learned over 20 years of building production code for side-projects, small businesses, and hyper growth startups. Both successful and not.
Subscribe below 👇
Join Swizec's Newsletter and get insightful emails 💌 on mindsets, tactics, and technical skills for your career. Real lessons from building production software. No bullshit.
"Man, love your simple writing! Yours is the only newsletter I open and only blog that I give a fuck to read & scroll till the end. And wow always take away lessons with me. Inspiring! And very relatable. 👌"
Senior Mindset Book
Get promoted, earn a bigger salary, work for top companiesLearn more
Have a burning question that you think I can answer? Hit me up on twitter and I'll do my best.
Who am I and who do I help? I'm Swizec Teller and I turn coders into engineers with "Raw and honest from the heart!" writing. No bullshit. Real insights into the career and skills of a modern software engineer.
Want to become a true senior engineer? Take ownership, have autonomy, and be a force multiplier on your team. The Senior Engineer Mindset ebook can help 👉 swizec.com/senior-mindset. These are the shifts in mindset that unlocked my career.
Curious about Serverless and the modern backend? Check out Serverless Handbook, for frontend engineers 👉 ServerlessHandbook.dev
Want to Stop copy pasting D3 examples and create data visualizations of your own? Learn how to build scalable dataviz React components your whole team can understand with React for Data Visualization
Did someone amazing share this letter with you? Wonderful! You can sign up for my weekly letters for software engineers on their path to greatness, here: swizec.com/blog
By the way, just in case no one has told you it yet today: I love and appreciate you for who you are ❤️