PRQL is a modern language for transforming data — a simple, powerful, pipelined SQL replacement
pq
allows you to use PRQL to easily query and
transform your data. It is powered by Apache Arrow
DataFusion and
DuckDB and is written in Rust (so it's "blazingly fast"
™)!
$ pq --from albums.csv "take 5"
+----------+---------------------------------------+-----------+
| album_id | title | artist_id |
+----------+---------------------------------------+-----------+
| 1 | For Those About To Rock We Salute You | 1 |
| 2 | Balls to the Wall | 2 |
| 3 | Restless and Wild | 2 |
| 4 | Let There Be Rock | 1 |
| 5 | Big Ones | 3 |
+----------+---------------------------------------+-----------+
$ pq -f i=invoices.csv -f c=customers.csv --to invoices_with_names.parquet \
'from i | join c [customer_id] | derive [name = f"{first_name} {last_name}"]'
$ pq -f invoices_with_names.parquet --format json \
'group name (aggregate [spend = sum total]) | sort [-spend] | take 10'
{"name":"Helena Holý","spend":49.620000000000005}
{"name":"Richard Cunningham","spend":47.620000000000005}
{"name":"Luis Rojas","spend":46.62}
{"name":"Hugh O'Reilly","spend":45.62}
{"name":"Ladislav Kovács","spend":45.62}
{"name":"Julia Barnett","spend":43.620000000000005}
{"name":"Fynn Zimmermann","spend":43.62}
{"name":"Frank Ralston","spend":43.62}
{"name":"Astrid Gruber","spend":42.62}
{"name":"Victor Stevens","spend":42.62}
Binaries are built for Windows, macOS and Linux for every release and can be dowloaded from Releases (latest).
docker pull ghcr.io/prql/prql-query:v0.0.9
alias pq="docker run --rm -it -v $(pwd):/data -w /data -u $(id -u):$(id -g) ghcr.io.prql/prql-query:v0.0.9"
pq --help
cargo install prql-query
brew tap prql/homebrew-prql-query
brew install prql-query
At its simplest pq
takes PRQL queries and transpiles them to SQL queries:
$ pq "from a | select b"
SELECT
b
FROM
a
Input can also come from stdin:
$ cat examples/queries/invoice_totals.prql | pq
For convenience, queries ending in ".prql" are assumed to be paths to PRQL query files and will be read in so this produces the same as above:
$ pq examples/queries/invoice_totals.prql
Both of these produce the output:
SELECT
STRFTIME('%Y-%m', i.invoice_date) AS month,
STRFTIME('%Y-%m-%d', i.invoice_date) AS day,
COUNT(DISTINCT i.invoice_id) AS num_orders,
SUM(ii.quantity) AS num_tracks,
SUM(ii.unit_price * ii.quantity) AS total_price,
SUM(SUM(ii.quantity)) OVER (
PARTITION BY STRFTIME('%Y-%m', i.invoice_date)
ORDER BY
STRFTIME('%Y-%m-%d', i.invoice_date) ROWS BETWEEN UNBOUNDED PRECEDING
AND CURRENT ROW
) AS running_total_num_tracks,
LAG(SUM(ii.quantity), 7) OVER (
ORDER BY
STRFTIME('%Y-%m-%d', i.invoice_date) ROWS BETWEEN UNBOUNDED PRECEDING
AND UNBOUNDED FOLLOWING
) AS num_tracks_last_week
FROM
invoices AS i
JOIN invoice_items AS ii USING(invoice_id)
GROUP BY
STRFTIME('%Y-%m', i.invoice_date),
STRFTIME('%Y-%m-%d', i.invoice_date)
ORDER BY
day
With the functionality described above, you should be able to query your favourite SQL RDBMS using your favourite CLI client and pq
. For example with the psql
client for PostgreSQL:
$ pq "from my_table | take 5" | psql postgresql:https://username:password@host:port/database
Or using the mysql
client for MySQL with a PRQL query stored in a file:
$ pq my_query.prql | mysql -h myhost -d mydb -u myuser -p mypassword
Similarly for MS SQL Server and other databases.
For querying and transforming data stored on the local filesystem, pq
comes in with a number of built-in backend query processing engines. The default backend is Apache Arrow DataFusion. However DuckDB and SQLite (planned) are also supported.
When --from
arguments are supplied which specify data files, the PRQL query will be applied to those files. The files can be referenced in the queries by the filenames without the extensions, e.g. customers.csv can be referenced as the table customers
. For convenience, unless a query already begins with a from ...
step, a from <table>
pipeline step will automatically be inserted at the beginning of the query referring to the last --from
argument encountered, i.e. the following two are equivalent:
$ pq --from examples/data/chinook/csv/invoices.csv "from invoices|take 5"
$ pq --from examples/data/chinook/csv/invoices.csv "take 5"
+------------+-------------+-------------------------------+-------------------------+--------------+---------------+-----------------+---------------------+-------+
| invoice_id | customer_id | invoice_date | billing_address | billing_city | billing_state | billing_country | billing_postal_code | total |
+------------+-------------+-------------------------------+-------------------------+--------------+---------------+-----------------+---------------------+-------+
| 1 | 2 | 2009-01-01T00:00:00.000000000 | Theodor-Heuss-Straße 34 | Stuttgart | | Germany | 70174 | 1.98 |
| 2 | 4 | 2009-01-02T00:00:00.000000000 | Ullevålsveien 14 | Oslo | | Norway | 0171 | 3.96 |
| 3 | 8 | 2009-01-03T00:00:00.000000000 | Grétrystraat 63 | Brussels | | Belgium | 1000 | 5.94 |
| 4 | 14 | 2009-01-06T00:00:00.000000000 | 8210 111 ST NW | Edmonton | AB | Canada | T6G 2C7 | 8.91 |
| 5 | 23 | 2009-01-11T00:00:00.000000000 | 69 Salem Street | Boston | MA | USA | 2113 | 13.86 |
+------------+-------------+-------------------------------+-------------------------+--------------+---------------+-----------------+---------------------+-------+
You can also assign an alias for source file with the following form --from <alias>=<filepath>
and then refer to it by that alias in your queries. So the following is another equivalent form of the queries above:
$ pq --from i=examples/data/chinook/csv/invoices.csv "from i|take 5"
This works with multiple files which means that the extended example above can be run as follows:
$ pq -b duckdb -f examples/data/chinook/csv/invoices.csv -f examples/data/chinook/csv/invoice_items.csv examples/queries/invoice_totals.prql
When a --to
argument is supplied, the output will be written there in the appropriate file format instead of stdout (the "" query is equivalent to select *
and is required because select *
currently does not work):
$ pq --from examples/data/chinook/csv/invoices.csv --to invoices.parquet ""
Currently csv, parquet and json file formats are supported for both readers and writers:
$ cat examples/queries/customer_totals.prql
group [customer_id] (
aggregate [
customer_total = sum total,
])
$ pq -f invoices.parquet -t customer_totals.json examples/queries/customer_totals.prql
$ pq -f customer_totals.json "sort [-customer_total] | take 10"
+-------------+--------------------+
| customer_id | customer_total |
+-------------+--------------------+
| 6 | 49.620000000000005 |
| 26 | 47.620000000000005 |
| 57 | 46.62 |
| 46 | 45.62 |
| 45 | 45.62 |
| 28 | 43.620000000000005 |
| 37 | 43.62 |
| 24 | 43.62 |
| 7 | 42.62 |
| 25 | 42.62 |
+-------------+--------------------+
DuckDB is natively supported and can be queried by supplying a database URI beginning with "duckdb:https://".
$ pq --database duckdb:https://examples/chinook/duckdb/chinook.duckdb \
'from albums | join artists [artist_id] | group name (aggregate [num_albums = count]) | sort [-num_albums] | take 10'
Sqlite is currently supported through the sqlite_scanner DuckDB extension. This appears to work for some queries but not for others so this is currently EXPERIMENTAL.
$ pq --database sqlite:https://examples/chinook/sqlite/chinook.sqlite \
'from albums | select title | take 10'
PostgreSQL is currently supported through the postgres-scanner DuckDB extension. (See the announcement blog post for a good introduction.)
$ pq -d postgresql:https://username:password@host:port/database \
'from table | take 10'
One noteworthy limitation of this approach is that you can only query tables in the postgres database and not views.
By default you will be connected to the "public" schema and can reference tables
there within your query. You can specify a different schema to connect to using
the "?currentSchema=schema" paramter. If you want to query tables from another schema
outside of that then you currently have to reference these through aliased
--from
parameters like so:
$ pq -d postgresql:https://username:password@host:port/database?currentSchema=schema \
--from alias=other_schema.table 'from alias | take 10'
If you plan to work with the same database repeatedly, then specifying the
details each time quickly becomes tedious. pq
allows you to supply all
command line arguments from environment variables with a PQ_
prefix. So for
example the same query from above could be achieved with:
$ export PQ_DATABASE="postgresql:https://username:password@host:port/database"
$ pq --from alias=schema.table 'take 10'
Environment variables can also be read from a .env
files. Since you probably
don't want to expose your database credentials at the shell, it makes sense to
put these in a .env
file. This also allows you to set up directories with
configuration for common environments together with common queries for that
environment, for example:
$ echo 'PQ_DATABASE="postgresql:https://username:password@host:port/database"' > .env
$ pq 'from my_schema.my_table | take 5'
Or say that you have a status_query.prql
that you need to run for a number of environments with .env files set up in subdirectories:
$ for e in prod uat dev; do cd $e && pq ../status_query.prql; done
- Tests
- Publish to crates.io
- Support for object stores
- Support for other databases through
connectorx