One year of PRQL
A year ago, we posted a proposal for a modern SQL replacement to Hacker News. It immediately sparked the interest of many people dealing with data. The project grew a community of people who are now developing the language, tooling, and the idea of a modern relational language. Since then we’ve opened 1577 issues & PRs on our main repo, submitted 4211 comments, and made 1176 commits.
The number of stars is skyrocketing every time the project appears on Hacker News, which we believe to be an indicator that people are eager to adopt the language if the tooling is made accessible enough.
Where are we?
Language design & development in the last year have been focused on these areas:
PRQL is now in a state where it can greatly improve the developer experience for writing complex analytical queries, but it does require a bit of fiddling to set up in your environment.
In the coming year, we are aiming to improve that by providing a dbt plugin and integrations for tools like Rill-developer, Metabase, and DataGrip. Read more about our plans and ambitions in the roadmap.
How are people using it?
For data analytics at SuperSimple:
We’ve been using PRQL under the hood to power complex analytics workflows at Supersimple for more than 6 months now. The speed of iteration and response to user feedback has been amazing during this period!
PRQL is what I think SQL should have been like from day 1 and adopting it has likely literally saved us months.
For making pretty reports with SQLite.
For quick, readable scripts in the command line.
What have you missed?
In the past year, the community has created many things, some of which have not been noticed as much as they should have been. In the summary of the year, we want to fix that and put a spotlight on the amazing work that was done.
We have a Playground that can compile and execute PRQL queries in-browser. It’s using prql-compiler and DuckDB, both compiled to WASM modules.
Currently, this is the best way to see how the relation is manipulated instantly as you type the query.
VS Code extension
Extension that provides syntax highlighting and compiled SQL within your editor.
CLI tool that uses DuckDB and DataFusion to execute PRQL queries from your terminal. Useful for wrangling your CSV and parquet files on the go.
Define functional semantics
Given the initial proposal of the language, we constructed consistent semantics of how functions in PRQL work, what they can express and how can they be abused.
To keep this post brief, we’ll expose a single snippet of what’s possible and invite you to read more in a recent post.
let take_oldest = n rel -> ( rel sort [-invoice_date] take n ) from invoices take_oldest 3
take_oldestis used in the main pipeline, just any other regular transform.
Its goal is the ability to express any operation possible in SQL while containing as few constructs as possible. This makes it easy to implement backends that compile RQ to SQL or any other language or library dealing with relations or dataframes.
Note how prql-compiler inferred the structure of the table we are selecting from. It knows that it must contain columns
total, but also notes that there may be many other columns.