What Is Malloy
This guide introduces the basics of querying data and building a semantic model with the Malloy language. By the end of this tutorial, you will understand how to use Malloy to run queries, build re-usable data models, and do analysis on your data that is nearly impossible in SQL.
The easiest way to follow along is by going to the interactive notebook version of this tutorial. The link will launch a browser-based VSCode environment and ask you to install the Malloy extension. Once installed, navigate back to the quickstart notebook file, and dive in.
If you'd like to run Malloy locally on your laptop instead, follow the setup instructions to install the VSCode extension and connect to a database.
A Simple Select Statement
The following query is equivalent to SELECT id, code, city FROM airports LIMIT 10 in SQL:
run: duckdb.table('../data/airports.parquet') -> { select: id code city limit: 10 }
[ { "id": 19783, "code": "1Q9", "city": "MILI ISLAND" }, { "id": 19777, "code": "Q51", "city": "KILI ISLAND" }, { "id": 19787, "code": "3N1", "city": "TAORA IS MALOELAP ATOLL" }, { "id": 19789, "code": "03N", "city": "UTIRIK ISLAND" }, { "id": 19774, "code": "ANG", "city": "ANGAUR ISLAND" } ]
SELECT base."id" as "id", base."code" as "code", base."city" as "city" FROM '../data/airports.parquet' as base LIMIT 10
Let's break down each part of this query.
run:is the opening statement that indicates we're starting to write a queryduckdb.table('../data/airports.parquet')defines the source for the query. Thetable()method creates a source from a table or view in the database.A source is similar to a table or view in SQL, but Malloy sources can include additional information like joins and measures. We'll cover this in depth later on.
The
->operator begins the query. Queries take the formsource -> { ... }, with the query logic specified inside of the curly braces.select:is equivalent toSELECTin SQL. In this clause, we select theid,code, andcitycolumns from the table.limit: 10limits the result set of the query to the first 10 items.
Query Operators
In SQL, the SELECT command does two very different things. A SELECT with a GROUP BY aggregates data according to the GROUP BY clause and produces aggregate calculation against every calculation not in the GROUP BY. In Malloy, the query operator for this is group_by:. Calculations involving data in the group are made using aggregate:.
The second type of SELECT in SQL does not perform any aggregation; All rows in the input table, unless filtered in some way, show up in the output table. In Malloy, this command is select:.
Aggregate
In the query below, the data will be grouped by state and county, and will produce an aggregate calculation for airport_count and average_elevation.
run: duckdb.table('../data/airports.parquet') -> { group_by: state county aggregate: airport_count is count() average_elevation is avg(elevation) }
[ { "state": "CA", "county": "LOS ANGELES", "airport_count": 176, "average_elevation": 689.1647727272727 }, { "state": "TX", "county": "HARRIS", "airport_count": 135, "average_elevation": 106.46666666666667 }, { "state": "AZ", "county": "MARICOPA", "airport_count": 117, "average_elevation": 1395.6666666666667 }, { "state": "CA", "county": "SAN BERNARDINO", "airport_count": 71, "average_elevation": 2376.056338028169 }, { "state": "TX", "county": "TARRANT", "airport_count": 63, "average_elevation": 646.7936507936508 } ]
SELECT base."state" as "state", base."county" as "county", COUNT(1) as "airport_count", AVG(base."elevation") as "average_elevation" FROM '../data/airports.parquet' as base GROUP BY 1,2 ORDER BY 3 desc NULLS LAST
Select
A select: statement produces a list of fields. For every row in the input table, there is a row in the output table. This is similar to a simple SELECT statement in SQL with no aggregations.
run: duckdb.table('../data/airports.parquet') -> { select: code, full_name, city, county where: county = 'SANTA CRUZ' limit: 10 }
[ { "code": "2AZ8", "full_name": "TUBAC ULTRALIGHT FLIGHTPARK", "city": "TUBAC", "county": "SANTA CRUZ" }, { "code": "OLS", "full_name": "NOGALES INTL", "city": "NOGALES", "county": "SANTA CRUZ" }, { "code": "NSI", "full_name": "SAN NICOLAS ISLAND NOLF", "city": "SAN NICOLAS ISLAND", "county": "SANTA CRUZ" }, { "code": "CL77", "full_name": "BONNY DOON VILLAGE", "city": "SANTA CRUZ", "county": "SANTA CRUZ" }, { "code": "CA37", "full_name": "DOMINICAN SANTA CRUZ HOSPITAL", "city": "SANTA CRUZ", "county": "SANTA CRUZ" } ]
SELECT base."code" as "code", base."full_name" as "full_name", base."city" as "city", base."county" as "county" FROM '../data/airports.parquet' as base WHERE base."county"='SANTA CRUZ' LIMIT 10
For the most part, operations can be placed in any order within a query. A where can come before or after a project, and limit can be placed anywhere as well. The above query could also be written:
run: duckdb.table('../data/airports.parquet') -> { limit: 10 where: county = 'SANTA CRUZ' select: code, full_name, city, county }
[ { "code": "2AZ8", "full_name": "TUBAC ULTRALIGHT FLIGHTPARK", "city": "TUBAC", "county": "SANTA CRUZ" }, { "code": "OLS", "full_name": "NOGALES INTL", "city": "NOGALES", "county": "SANTA CRUZ" }, { "code": "NSI", "full_name": "SAN NICOLAS ISLAND NOLF", "city": "SAN NICOLAS ISLAND", "county": "SANTA CRUZ" }, { "code": "CL77", "full_name": "BONNY DOON VILLAGE", "city": "SANTA CRUZ", "county": "SANTA CRUZ" }, { "code": "CA37", "full_name": "DOMINICAN SANTA CRUZ HOSPITAL", "city": "SANTA CRUZ", "county": "SANTA CRUZ" } ]
SELECT base."code" as "code", base."full_name" as "full_name", base."city" as "city", base."county" as "county" FROM '../data/airports.parquet' as base WHERE base."county"='SANTA CRUZ' LIMIT 10
Everything has a Name
In Malloy, all output fields have names. This means that any time a query includes a field with a calculated value, like a scalar or aggregate function, it must be named (unlike SQL, which allows un-named expressions).
run: duckdb.table('../data/airports.parquet') -> { aggregate: max_elevation is max(elevation) }
[ { "max_elevation": 12442 } ]
SELECT max(base."elevation") as "max_elevation" FROM '../data/airports.parquet' as base
Notice that Malloy uses the form name is value instead of SQL's value AS name.
Having the output column name written first makes it easier for someone reading
the code to visualize the resulting query structure.
Named objects, like columns from a table or fields defined in a source, can be included in field lists without an is:
run: duckdb.table('../data/airports.parquet') -> { select: full_name elevation }
[ { "full_name": "MILI", "elevation": 4 }, { "full_name": "KILI", "elevation": 5 }, { "full_name": "MALOELAP", "elevation": 4 }, { "full_name": "UTIRIK", "elevation": 4 }, { "full_name": "ANGAUR AIRSTRIP", "elevation": 20 } ]
SELECT base."full_name" as "full_name", base."elevation" as "elevation" FROM '../data/airports.parquet' as base
Expressions
Many SQL expressions will work unchanged in Malloy, and many functions available in Standard SQL are usable in Malloy as well. This makes expressions fairly straightforward to understand, given a knowledge of SQL.
run: duckdb.table('../data/airports.parquet') -> { group_by: county_and_state is concat(county, ', ', state) aggregate: airport_count is count() max_elevation is max(elevation) min_elevation is min(elevation) avg_elevation is avg(elevation) }
[ { "county_and_state": "LOS ANGELES, CA", "airport_count": 176, "max_elevation": 3420, "min_elevation": 0, "avg_elevation": 689.1647727272727 }, { "county_and_state": "HARRIS, TX", "airport_count": 135, "max_elevation": 774, "min_elevation": 9, "avg_elevation": 106.46666666666667 }, { "county_and_state": "MARICOPA, AZ", "airport_count": 117, "max_elevation": 3995, "min_elevation": 737, "avg_elevation": 1395.6666666666667 }, { "county_and_state": "SAN BERNARDINO, CA", "airport_count": 71, "max_elevation": 6748, "min_elevation": 631, "avg_elevation": 2376.056338028169 }, { "county_and_state": "TARRANT, TX", "airport_count": 63, "max_elevation": 895, "min_elevation": 472, "avg_elevation": 646.7936507936508 } ]
SELECT CONCAT(base."county",', ',base."state") as "county_and_state", COUNT(1) as "airport_count", max(base."elevation") as "max_elevation", min(base."elevation") as "min_elevation", AVG(base."elevation") as "avg_elevation" FROM '../data/airports.parquet' as base GROUP BY 1 ORDER BY 2 desc NULLS LAST
The basic types of Malloy expressions are string, number, boolean, date, and timestamp.
Sources: the Basic Structure for Modeling and Reuse
One of the main benefits of Malloy is the ability to save common calculations into a data model. The data model is made of sources, which can be thought of as tables or views, but with additional information, such as joins, dimensions and measures.
In the example below, we create a source object named airports and add a dimension calculation for county_and_state and measure calculation for airport_count. Dimensions can be used in group_by, project and where. Measures can be used in aggregate and having.
source: airports is duckdb.table('../data/airports.parquet') extend { dimension: county_and_state is concat(county, ', ', state) measure: airport_count is count() measure: average_elevation is avg(elevation) }
malloy run: airports -> { group_by: county_and_state aggregate: airport_count }
Sources that are defined in one file can be imported into another using import "path/to/some/file.malloy". For example, if the airports source above were defined in a file called flights.malloy, you could create a new file that imports it and immediately start using the airports source:
import "airports.malloy" run: airports -> { group_by: county_and_state aggregate: average_elevation }
Sources can also contain named views. These views are useful for building nested queries (covered later) or for saving a query operation so it can re-used again and again without having to rewrite it.
source: airports_with_named_query is duckdb.table('../data/airports.parquet') extend { dimension: county_and_state is concat(county, ', ', state) measure: airport_count is count() measure: average_elevation is avg(elevation) view: top_county_and_state is { group_by: county_and_state aggregate: airport_count limit:10 } } // The view can now be referenced by name // and run without having to rewrite the logic: run: airports_with_named_query -> top_county_and_state
[ { "county_and_state": "LOS ANGELES, CA", "airport_count": 176 }, { "county_and_state": "HARRIS, TX", "airport_count": 135 }, { "county_and_state": "MARICOPA, AZ", "airport_count": 117 }, { "county_and_state": "SAN BERNARDINO, CA", "airport_count": 71 }, { "county_and_state": "TARRANT, TX", "airport_count": 63 } ]
SELECT CONCAT(base."county",', ',base."state") as "county_and_state", COUNT(1) as "airport_count" FROM '../data/airports.parquet' as base GROUP BY 1 ORDER BY 2 desc NULLS LAST LIMIT 10
Joins
Joins are declared as part of a source. When joining a source to another, it brings with it all child joins.
source: aircraft_models is duckdb.table('../data/aircraft_models.parquet') extend { primary_key: aircraft_model_code } source: aircraft is duckdb.table('../data/aircraft.parquet') extend { primary_key: tail_num join_one: aircraft_models on aircraft_model_code = aircraft_models.aircraft_model_code } source: flights is duckdb.table('../data/flights.parquet') extend { join_one: aircraft on tail_num = aircraft.tail_num } run: flights -> { where: dep_time ? @2003-01 group_by: aircraft.aircraft_models.manufacturer aggregate: flight_count is count() aircraft_count is aircraft.count() average_seats_per_model is aircraft.aircraft_models.seats.avg() }
[ { "manufacturer": "BOEING", "flight_count": 2432, "aircraft_count": 20, "average_seats_per_model": 193.25 }, { "manufacturer": "AIRBUS INDUSTRIE", "flight_count": 877, "aircraft_count": 8, "average_seats_per_model": 177 }, { "manufacturer": "MCDONNELL DOUGLAS", "flight_count": 486, "aircraft_count": 5, "average_seats_per_model": 172 }, { "manufacturer": "EMBRAER", "flight_count": 388, "aircraft_count": 4, "average_seats_per_model": 41.333333333333336 }, { "manufacturer": "AEROSPATIALE/ALENIA", "flight_count": 225, "aircraft_count": 1, "average_seats_per_model": 76 } ]
SELECT aircraft_models_0."manufacturer" as "manufacturer", COUNT(1) as "flight_count", COUNT(DISTINCT aircraft_0."tail_num") as "aircraft_count", ( SELECT AVG(a.val) as value FROM ( SELECT UNNEST(list(distinct {key:aircraft_models_0."aircraft_model_code", val: aircraft_models_0."seats"})) a ) ) as "average_seats_per_model" FROM '../data/flights.parquet' as base LEFT JOIN '../data/aircraft.parquet' AS aircraft_0 ON base."tail_num"=aircraft_0."tail_num" LEFT JOIN '../data/aircraft_models.parquet' AS aircraft_models_0 ON aircraft_0."aircraft_model_code"=aircraft_models_0."aircraft_model_code" WHERE (base."dep_time">=TIMESTAMP '2003-01-01 00:00:00') and (base."dep_time"<TIMESTAMP '2003-02-01 00:00:00') GROUP BY 1 ORDER BY 2 desc NULLS LAST
In this example, the aircraft source is joined to flights, and aircraft_models is joined via aircraft. These examples explicitly name both keys—this same syntax can be used to write more complex joins.
Now, any query that uses the flights source has access to fields in both aircraft and aircraft_models without having to explicitly specify the join condition. The joins are specified once in the source, and usable by any query on flights.
An ad hoc join can also be specified in a query block. In the query below, we join in the airports table using the destination column as a join key, then compute the top 5 destination airports by flight count.
source: airports2 is duckdb.table('../data/airports.parquet') source: flights2 is duckdb.table('../data/flights.parquet') extend { join_one: airports2 on destination = airports2.code } run: flights2 -> { group_by: airports2.full_name aggregate: flight_count is count() limit: 5 }
[ { "full_name": "THE WILLIAM B HARTSFIELD ATLANTA INTL", "flight_count": 17832 }, { "full_name": "DALLAS/FORT WORTH INTERNATIONAL", "flight_count": 17776 }, { "full_name": "CHICAGO O'HARE INTL", "flight_count": 14213 }, { "full_name": "PHOENIX SKY HARBOR INTL", "flight_count": 12477 }, { "full_name": "MC CARRAN INTL", "flight_count": 11092 } ]
SELECT airports2_0."full_name" as "full_name", COUNT(1) as "flight_count" FROM '../data/flights.parquet' as base LEFT JOIN '../data/airports.parquet' AS airports2_0 ON base."destination"=airports2_0."code" GROUP BY 1 ORDER BY 2 desc NULLS LAST LIMIT 5
Filtering
When working with data, filtering is something you do in almost every query. Malloy provides consistent syntax for filtering everywhere within a query. The most basic type of filter is applied using a where: clause, very similar to a WHERE clause in SQL.
The following query grabs the top 5 counties in California with the highest airport count:
run: duckdb.table('../data/airports.parquet') -> { where: state = 'CA' limit: 5 group_by: county aggregate: airport_count is count() }
[ { "county": "LOS ANGELES", "airport_count": 176 }, { "county": "SAN BERNARDINO", "airport_count": 71 }, { "county": "ORANGE", "airport_count": 53 }, { "county": "KERN", "airport_count": 49 }, { "county": "SAN DIEGO", "airport_count": 49 } ]
SELECT base."county" as "county", COUNT(1) as "airport_count" FROM '../data/airports.parquet' as base WHERE base."state"='CA' GROUP BY 1 ORDER BY 2 desc NULLS LAST LIMIT 5
Filters can also be applied to sources:
source: airports_in_california is duckdb.table('../data/airports.parquet') extend { where: state = 'CA' } run: airports_in_california -> { limit: 5 group_by: county aggregate: airport_count is count() }
[ { "county": "LOS ANGELES", "airport_count": 176 }, { "county": "SAN BERNARDINO", "airport_count": 71 }, { "county": "ORANGE", "airport_count": 53 }, { "county": "SAN DIEGO", "airport_count": 49 }, { "county": "KERN", "airport_count": 49 } ]
SELECT base."county" as "county", COUNT(1) as "airport_count" FROM '../data/airports.parquet' as base WHERE base."state"='CA' GROUP BY 1 ORDER BY 2 desc NULLS LAST LIMIT 5
Any query run on the airports_in_california source will run against the airports table, and always include the filter in state = 'CA'.
Filtering Measures
A filter on an aggregate calculation (a measure) narrows down the data used in that specific calculation. In the example below, the calculations for airports and heliports are filtered separately.
run: duckdb.table('../data/airports.parquet') -> { group_by: state aggregate: airports is count() { where: fac_type = 'AIRPORT' } heliports is count() { where: fac_type = 'HELIPORT' } total is count() }
[ { "state": "TX", "airports": 1389, "heliports": 435, "total": 1845 }, { "state": "IL", "airports": 625, "heliports": 245, "total": 890 }, { "state": "CA", "airports": 569, "heliports": 396, "total": 984 }, { "state": "OH", "airports": 537, "heliports": 201, "total": 749 }, { "state": "FL", "airports": 511, "heliports": 280, "total": 856 } ]
SELECT base."state" as "state", COUNT(CASE WHEN base."fac_type"='AIRPORT' THEN 1 END) as "airports", COUNT(CASE WHEN base."fac_type"='HELIPORT' THEN 1 END) as "heliports", COUNT(1) as "total" FROM '../data/airports.parquet' as base GROUP BY 1 ORDER BY 2 desc NULLS LAST
In SQL, this same calculation is often done using CASE statements inside of the aggregates, which is verbose and difficult to read. A query like the above would look like:
SELECT state , SUM(CASE WHEN fac_type = 'AIRPORT' THEN 1 ELSE 0 END) AS airports , SUM(CASE WHEN fac_type = 'HELIPORT' THEN 1 ELSE 0 END) AS heliports , COUNT(*) AS total FROM `malloy-data.faa.airports` GROUP BY state
Nested Queries
The next several examples will use this simple source definition:
source: airports is duckdb.table('../data/airports.parquet') extend { measure: airport_count is count() };
Nested Views
In Malloy, views can be nested to produce a nested query with subtables on each output row.
run: airports -> { group_by: state aggregate: airport_count nest: by_facility is { group_by: fac_type aggregate: airport_count limit: 3 } }
[ { "state": "TX", "airport_count": 1845, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 1389 }, { "fac_type": "HELIPORT", "airport_count": 435 }, { "fac_type": "ULTRALIGHT", "airport_count": 8 } ] }, { "state": "CA", "airport_count": 984, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 569 }, { "fac_type": "HELIPORT", "airport_count": 396 }, { "fac_type": "SEAPLANE BASE", "airport_count": 12 } ] }, { "state": "IL", "airport_count": 890, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 625 }, { "fac_type": "HELIPORT", "airport_count": 245 }, { "fac_type": "SEAPLANE BASE", "airport_count": 8 } ] }, { "state": "FL", "airport_count": 856, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 511 }, { "fac_type": "HELIPORT", "airport_count": 280 }, { "fac_type": "SEAPLANE BASE", "airport_count": 43 } ] }, { "state": "PA", "airport_count": 804, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 468 }, { "fac_type": "HELIPORT", "airport_count": 307 }, { "fac_type": "ULTRALIGHT", "airport_count": 13 } ] } ]
WITH __stage0 AS ( SELECT group_set, base."state" as "state__0", CASE WHEN group_set=0 THEN COUNT(1) END as "airport_count__0", CASE WHEN group_set=1 THEN base."fac_type" END as "fac_type__1", CASE WHEN group_set=1 THEN COUNT(1) END as "airport_count__1" FROM '../data/airports.parquet' as base CROSS JOIN (SELECT UNNEST(GENERATE_SERIES(0,1,1)) as group_set ) as group_set GROUP BY 1,2,4 ) SELECT "state__0" as "state", MAX(CASE WHEN group_set=0 THEN "airport_count__0" END) as "airport_count", COALESCE(LIST({ "fac_type": "fac_type__1", "airport_count": "airport_count__1"} ORDER BY "airport_count__1" desc NULLS LAST) FILTER (WHERE group_set=1)[1:3],[]) as "by_facility" FROM __stage0 GROUP BY 1 ORDER BY 2 desc NULLS LAST
Here we can see that the by_facility column of the output table contains a nested subtable on each row. by_facility contains the counts for the top 3 facility types for each state, i.e., the number of airports, heliports, and stolports in Texas, the number of airports, heliports, and seaplane bases in California, etc.
When a view is nested inside another view, each output row of the outer view will have a nested table for the inner view which only includes data limited to that row.
Views can be nested infinitely, allowing for rich, complex output structures. A view may always include another nested view, regardless of depth.
run: airports -> { group_by: state aggregate: airport_count nest: top_5_counties is { limit: 5 group_by: county aggregate: airport_count nest: by_facility is { group_by: fac_type aggregate: airport_count } } }
[ { "state": "TX", "airport_count": 1845, "top_5_counties": [ { "county": "HARRIS", "airport_count": 135, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 110 }, { "fac_type": "AIRPORT", "airport_count": 25 } ] }, { "county": "TARRANT", "airport_count": 63, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 35 }, { "fac_type": "AIRPORT", "airport_count": 26 }, { "fac_type": "ULTRALIGHT", "airport_count": 1 }, { "fac_type": "STOLPORT", "airport_count": 1 } ] }, { "county": "DENTON", "airport_count": 53, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 47 }, { "fac_type": "HELIPORT", "airport_count": 6 } ] }, { "county": "DALLAS", "airport_count": 42, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 32 }, { "fac_type": "AIRPORT", "airport_count": 9 }, { "fac_type": "STOLPORT", "airport_count": 1 } ] }, { "county": "BEXAR", "airport_count": 40, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 24 }, { "fac_type": "HELIPORT", "airport_count": 16 } ] } ] }, { "state": "CA", "airport_count": 984, "top_5_counties": [ { "county": "LOS ANGELES", "airport_count": 176, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 151 }, { "fac_type": "AIRPORT", "airport_count": 23 }, { "fac_type": "SEAPLANE BASE", "airport_count": 2 } ] }, { "county": "SAN BERNARDINO", "airport_count": 71, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 47 }, { "fac_type": "HELIPORT", "airport_count": 24 } ] }, { "county": "ORANGE", "airport_count": 53, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 47 }, { "fac_type": "AIRPORT", "airport_count": 6 } ] }, { "county": "SAN DIEGO", "airport_count": 49, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 30 }, { "fac_type": "HELIPORT", "airport_count": 17 }, { "fac_type": "GLIDERPORT", "airport_count": 2 } ] }, { "county": "KERN", "airport_count": 49, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 41 }, { "fac_type": "HELIPORT", "airport_count": 7 }, { "fac_type": "ULTRALIGHT", "airport_count": 1 } ] } ] }, { "state": "IL", "airport_count": 890, "top_5_counties": [ { "county": "COOK", "airport_count": 51, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 44 }, { "fac_type": "AIRPORT", "airport_count": 7 } ] }, { "county": "LA SALLE", "airport_count": 39, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 35 }, { "fac_type": "HELIPORT", "airport_count": 4 } ] }, { "county": "MC HENRY", "airport_count": 29, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 20 }, { "fac_type": "HELIPORT", "airport_count": 7 }, { "fac_type": "SEAPLANE BASE", "airport_count": 2 } ] }, { "county": "DE KALB", "airport_count": 27, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 24 }, { "fac_type": "HELIPORT", "airport_count": 3 } ] }, { "county": "LEE", "airport_count": 24, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 18 }, { "fac_type": "HELIPORT", "airport_count": 5 }, { "fac_type": "ULTRALIGHT", "airport_count": 1 } ] } ] }, { "state": "FL", "airport_count": 856, "top_5_counties": [ { "county": "PALM BEACH", "airport_count": 45, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 30 }, { "fac_type": "AIRPORT", "airport_count": 14 }, { "fac_type": "GLIDERPORT", "airport_count": 1 } ] }, { "county": "DADE", "airport_count": 44, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 27 }, { "fac_type": "AIRPORT", "airport_count": 12 }, { "fac_type": "SEAPLANE BASE", "airport_count": 2 }, { "fac_type": "GLIDERPORT", "airport_count": 2 }, { "fac_type": "STOLPORT", "airport_count": 1 } ] }, { "county": "POLK", "airport_count": 43, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 18 }, { "fac_type": "HELIPORT", "airport_count": 16 }, { "fac_type": "SEAPLANE BASE", "airport_count": 9 } ] }, { "county": "MARION", "airport_count": 37, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 27 }, { "fac_type": "HELIPORT", "airport_count": 7 }, { "fac_type": "SEAPLANE BASE", "airport_count": 2 }, { "fac_type": "STOLPORT", "airport_count": 1 } ] }, { "county": "ORANGE", "airport_count": 36, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 24 }, { "fac_type": "AIRPORT", "airport_count": 8 }, { "fac_type": "SEAPLANE BASE", "airport_count": 4 } ] } ] }, { "state": "PA", "airport_count": 804, "top_5_counties": [ { "county": "BUCKS", "airport_count": 55, "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 32 }, { "fac_type": "HELIPORT", "airport_count": 19 }, { "fac_type": "ULTRALIGHT", "airport_count": 2 }, { "fac_type": "STOLPORT", "airport_count": 1 }, { "fac_type": "GLIDERPORT", "airport_count": 1 } ] }, { "county": "MONTGOMERY", "airport_count": 44, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 29 }, { "fac_type": "AIRPORT", "airport_count": 14 }, { "fac_type": "SEAPLANE BASE", "airport_count": 1 } ] }, { "county": "ALLEGHENY", "airport_count": 31, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 22 }, { "fac_type": "AIRPORT", "airport_count": 8 }, { "fac_type": "SEAPLANE BASE", "airport_count": 1 } ] }, { "county": "CHESTER", "airport_count": 27, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 16 }, { "fac_type": "AIRPORT", "airport_count": 11 } ] }, { "county": "PHILADELPHIA", "airport_count": 26, "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 22 }, { "fac_type": "AIRPORT", "airport_count": 4 } ] } ] } ]
WITH __stage0 AS ( SELECT group_set, base."state" as "state__0", CASE WHEN group_set=0 THEN COUNT(1) END as "airport_count__0", CASE WHEN group_set IN (1,2) THEN base."county" END as "county__1", CASE WHEN group_set=1 THEN COUNT(1) END as "airport_count__1", CASE WHEN group_set=2 THEN base."fac_type" END as "fac_type__2", CASE WHEN group_set=2 THEN COUNT(1) END as "airport_count__2" FROM '../data/airports.parquet' as base CROSS JOIN (SELECT UNNEST(GENERATE_SERIES(0,2,1)) as group_set ) as group_set GROUP BY 1,2,4,6 ) , __stage1 AS ( SELECT CASE WHEN group_set=2 THEN 1 ELSE group_set END as group_set, "state__0" as "state__0", FIRST("airport_count__0") FILTER (WHERE "airport_count__0" IS NOT NULL) as "airport_count__0", CASE WHEN group_set IN (1,2) THEN "county__1" END as "county__1", FIRST("airport_count__1") FILTER (WHERE "airport_count__1" IS NOT NULL) as "airport_count__1", COALESCE(LIST({ "fac_type": "fac_type__2", "airport_count": "airport_count__2"} ORDER BY "airport_count__2" desc NULLS LAST) FILTER (WHERE group_set=2),[]) as "by_facility__1" FROM __stage0 GROUP BY 1,2,4 ) SELECT "state__0" as "state", MAX(CASE WHEN group_set=0 THEN "airport_count__0" END) as "airport_count", COALESCE(LIST({ "county": "county__1", "airport_count": "airport_count__1", "by_facility": "by_facility__1"} ORDER BY "airport_count__1" desc NULLS LAST) FILTER (WHERE group_set=1)[1:5],[]) as "top_5_counties" FROM __stage1 GROUP BY 1 ORDER BY 2 desc NULLS LAST
Filtering Nested Views
Filters can be isolated to any level of nesting. In the following example, we limit the major_facilities view to only airports where major is 'Y'. This particular filter applies only to major_facilities, and not to other parts of the outer query.
run: airports -> { where: state = 'CA' group_by: county aggregate: airport_count nest: major_facilities is { where: major = 'Y' group_by: name is concat(code, ' (', full_name, ')') } nest: by_facility is { group_by: fac_type aggregate: airport_count } }
[ { "county": "LOS ANGELES", "airport_count": 176, "major_facilities": [ { "name": "BUR (BURBANK-GLENDALE-PASADENA)" }, { "name": "LAX (LOS ANGELES INTL)" }, { "name": "LGB (LONG BEACH /DAUGHERTY FIELD/)" } ], "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 151 }, { "fac_type": "AIRPORT", "airport_count": 23 }, { "fac_type": "SEAPLANE BASE", "airport_count": 2 } ] }, { "county": "SAN BERNARDINO", "airport_count": 71, "major_facilities": [ { "name": "ONT (ONTARIO INTL)" } ], "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 47 }, { "fac_type": "HELIPORT", "airport_count": 24 } ] }, { "county": "ORANGE", "airport_count": 53, "major_facilities": [ { "name": "SNA (JOHN WAYNE AIRPORT-ORANGE COUNTY)" } ], "by_facility": [ { "fac_type": "HELIPORT", "airport_count": 47 }, { "fac_type": "AIRPORT", "airport_count": 6 } ] }, { "county": "KERN", "airport_count": 49, "major_facilities": [ { "name": "BFL (MEADOWS FIELD)" }, { "name": "IYK (INYOKERN)" } ], "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 41 }, { "fac_type": "HELIPORT", "airport_count": 7 }, { "fac_type": "ULTRALIGHT", "airport_count": 1 } ] }, { "county": "SAN DIEGO", "airport_count": 49, "major_facilities": [ { "name": "SAN (SAN DIEGO INTL-LINDBERGH FLD)" } ], "by_facility": [ { "fac_type": "AIRPORT", "airport_count": 30 }, { "fac_type": "HELIPORT", "airport_count": 17 }, { "fac_type": "GLIDERPORT", "airport_count": 2 } ] } ]
WITH __stage0 AS ( SELECT group_set, base."county" as "county__0", CASE WHEN group_set=0 THEN COUNT(1) END as "airport_count__0", CASE WHEN group_set=1 THEN CONCAT(base."code",' (',base."full_name",')') END as "name__1", CASE WHEN group_set=2 THEN base."fac_type" END as "fac_type__2", CASE WHEN group_set=2 THEN COUNT(1) END as "airport_count__2" FROM '../data/airports.parquet' as base CROSS JOIN (SELECT UNNEST(GENERATE_SERIES(0,2,1)) as group_set ) as group_set WHERE (base."state"='CA') AND ((group_set NOT IN (1) OR (group_set IN (1) AND base."major"='Y'))) GROUP BY 1,2,4,5 ) SELECT "county__0" as "county", MAX(CASE WHEN group_set=0 THEN "airport_count__0" END) as "airport_count", COALESCE(LIST({ "name": "name__1"} ORDER BY "name__1" asc NULLS LAST) FILTER (WHERE group_set=1),[]) as "major_facilities", COALESCE(LIST({ "fac_type": "fac_type__2", "airport_count": "airport_count__2"} ORDER BY "airport_count__2" desc NULLS LAST) FILTER (WHERE group_set=2),[]) as "by_facility" FROM __stage0 GROUP BY 1 ORDER BY 2 desc NULLS LAST
Dates and Timestamps
Working with time in data is often needlessly complex; Malloy has built in constructs to simplify many time-related operations. This section gives a brief introduction to some of these tools, but for more details see the Time Ranges section.
Time Literals
Literals of type date and timestamp are notated with an @, e.g. @2003-03-29 or @1994-07-14 10:23:59. Similarly, years (@2021), quarters (@2020-Q1), months (@2019-03), weeks (@WK2021-08-01), and minutes (@2017-01-01 10:53) can be expressed.
Time literals can be used as values, but are more often useful in filters. For example, the following query shows the number of flights in 2003.
run: duckdb.table('../data/flights.parquet') -> { where: dep_time ? @2003 aggregate: flight_count is count() }
[ { "flight_count": 58676 } ]
SELECT COUNT(1) as "flight_count" FROM '../data/flights.parquet' as base WHERE (base."dep_time">=TIMESTAMP '2003-01-01 00:00:00') and (base."dep_time"<TIMESTAMP '2004-01-01 00:00:00')
There is a special time literal now, referring to the current timestamp, which allows for relative time filters.
run: duckdb.table('../data/flights.parquet') -> { where: dep_time > now - 6 hours aggregate: flights_last_6_hours is count() }
[ { "flights_last_6_hours": 0 } ]
SELECT COUNT(1) as "flights_last_6_hours" FROM '../data/flights.parquet' as base WHERE base."dep_time">(LOCALTIMESTAMP - INTERVAL (6) hour)
Truncation
Time values can be truncated to a given timeframe, which can be second, minute, hour, day, week, month, quarter, or year.
run: duckdb.table('../data/flights.parquet') -> { group_by: flight_year is dep_time.year flight_month is dep_time.month aggregate: flight_count is count() }
[ { "flight_year": "2005-01-01T00:00:00.000Z", "flight_month": "2005-05-01T00:00:00.000Z", "flight_count": 6064 }, { "flight_year": "2005-01-01T00:00:00.000Z", "flight_month": "2005-07-01T00:00:00.000Z", "flight_count": 6227 }, { "flight_year": "2005-01-01T00:00:00.000Z", "flight_month": "2005-01-01T00:00:00.000Z", "flight_count": 5997 }, { "flight_year": "2005-01-01T00:00:00.000Z", "flight_month": "2005-03-01T00:00:00.000Z", "flight_count": 5732 }, { "flight_year": "2005-01-01T00:00:00.000Z", "flight_month": "2005-11-01T00:00:00.000Z", "flight_count": 5533 } ]
SELECT DATE_TRUNC('year', base."dep_time") as "flight_year", DATE_TRUNC('month', base."dep_time") as "flight_month", COUNT(1) as "flight_count" FROM '../data/flights.parquet' as base GROUP BY 1,2 ORDER BY 1 desc NULLS LAST
Extraction
Numeric values can be extracted from time values, e.g. day_of_year(some_date) or minute(some_time). See the full list of extraction functions here.
run: duckdb.table('../data/flights.parquet') -> { order_by: 1 group_by: day_of_week is day(dep_time) aggregate: flight_count is count() }
[ { "day_of_week": 1, "flight_count": 11265 }, { "day_of_week": 2, "flight_count": 11389 }, { "day_of_week": 3, "flight_count": 11060 }, { "day_of_week": 4, "flight_count": 11264 }, { "day_of_week": 5, "flight_count": 11201 }, { "day_of_week": 6, "flight_count": 11401 }, { "day_of_week": 7, "flight_count": 11512 } ]
SELECT EXTRACT(day FROM base."dep_time") as "day_of_week", COUNT(1) as "flight_count" FROM '../data/flights.parquet' as base GROUP BY 1 ORDER BY 1 ASC NULLS LAST
Time Ranges
Two kinds of time ranges are given special syntax: the range between two times and the range starting at some time for some duration. These are represented like @2003 to @2005 and @2004-Q1 for 6 quarters respectively. These ranges can be used in filters just like time literals.
run: duckdb.table('../data/flights.parquet') -> { where: dep_time ? @2003 to @2005 aggregate: flight_count is count() }
[ { "flight_count": 127073 } ]
SELECT COUNT(1) as "flight_count" FROM '../data/flights.parquet' as base WHERE (base."dep_time">=TIMESTAMP '2003-01-01 00:00:00') and (base."dep_time"<TIMESTAMP '2005-01-01 00:00:00')
Time literals and truncations can also behave like time ranges. Each kind of time literal has an implied duration that takes effect when it is used in a comparison, e.g. @2003 represents the whole of the year 2003, and @2004-Q1 lasts the whole 3 months of the quarter. Similarly, when a time value is truncated, it takes on the
timeframe from the truncation, e.g. now.month means the entirety of the current month.
When a time range is used in a comparison, = checks for "is in the range", > "is after", and < "is before." So some_time > @2003 filters dates starting on January 1, 2004, while some_time = @2003 filters to dates in the year 2003.
run: duckdb.table('../data/flights.parquet') -> { where: dep_time > @2003 limit: 3 order_by: departure_date asc group_by: departure_date is dep_time.day aggregate: flight_count is count() }
[ { "departure_date": "2004-01-01T00:00:00.000Z", "flight_count": 144 }, { "departure_date": "2004-01-02T00:00:00.000Z", "flight_count": 155 }, { "departure_date": "2004-01-03T00:00:00.000Z", "flight_count": 134 } ]
SELECT DATE_TRUNC('day', base."dep_time") as "departure_date", COUNT(1) as "flight_count" FROM '../data/flights.parquet' as base WHERE base."dep_time">=TIMESTAMP '2004-01-01 00:00:00' GROUP BY 1 ORDER BY 1 asc NULLS LAST LIMIT 3
Pipelines and Multi-stage Queries
The output from one stage of a query can be passed into another stage using ->. For example, we'll start with this query which outputs, for California and New York, the total number of airports, as well as the number of airports in each county.
run: airports -> { where: state = 'CA' | 'NY' group_by: state aggregate: airport_count nest: by_county is { group_by: county aggregate: airport_count } }
[ { "state": "CA", "airport_count": 984, "by_county": [ { "county": "LOS ANGELES", "airport_count": 176 }, { "county": "SAN BERNARDINO", "airport_count": 71 }, { "county": "ORANGE", "airport_count": 53 }, { "county": "SAN DIEGO", "airport_count": 49 }, { "county": "KERN", "airport_count": 49 }, { "county": "RIVERSIDE", "airport_count": 46 }, { "county": "FRESNO", "airport_count": 40 }, { "county": "SACRAMENTO", "airport_count": 24 }, { "county": "SANTA BARBARA", "airport_count": 22 }, { "county": "SAN JOAQUIN", "airport_count": 20 }, { "county": "VENTURA", "airport_count": 19 }, { "county": "TULARE", "airport_count": 16 }, { "county": "SUTTER", "airport_count": 16 }, { "county": "KINGS", "airport_count": 15 }, { "county": "MERCED", "airport_count": 15 }, { "county": "EL DORADO", "airport_count": 15 }, { "county": "ALAMEDA", "airport_count": 14 }, { "county": "SAN LUIS OBISPO", "airport_count": 14 }, { "county": "SISKIYOU", "airport_count": 13 }, { "county": "MONTEREY", "airport_count": 13 }, { "county": "IMPERIAL", "airport_count": 12 }, { "county": "MADERA", "airport_count": 12 }, { "county": "SONOMA", "airport_count": 12 }, { "county": "SHASTA", "airport_count": 11 }, { "county": "MODOC", "airport_count": 11 }, { "county": "BUTTE", "airport_count": 11 }, { "county": "SANTA CLARA", "airport_count": 10 }, { "county": "NEVADA", "airport_count": 10 }, { "county": "STANISLAUS", "airport_count": 10 }, { "county": "CONTRA COSTA", "airport_count": 9 }, { "county": "TEHAMA", "airport_count": 9 }, { "county": "COLUSA", "airport_count": 9 }, { "county": "YOLO", "airport_count": 9 }, { "county": "INYO", "airport_count": 9 }, { "county": "HUMBOLDT", "airport_count": 9 }, { "county": "MENDOCINO", "airport_count": 8 }, { "county": "SOLANO", "airport_count": 8 }, { "county": "LAKE", "airport_count": 8 }, { "county": "NAPA", "airport_count": 8 }, { "county": "PLUMAS", "airport_count": 8 }, { "county": "SANTA CRUZ", "airport_count": 8 }, { "county": "PLACER", "airport_count": 8 }, { "county": "TUOLUMNE", "airport_count": 7 }, { "county": "TRINITY", "airport_count": 7 }, { "county": "AMADOR", "airport_count": 7 }, { "county": "MARIN", "airport_count": 6 }, { "county": "MONO", "airport_count": 6 }, { "county": "LASSEN", "airport_count": 6 }, { "county": "CALAVERAS", "airport_count": 5 }, { "county": "YUBA", "airport_count": 5 }, { "county": "SAN BENITO", "airport_count": 4 }, { "county": "SAN MATEO", "airport_count": 4 }, { "county": "GLENN", "airport_count": 4 }, { "county": "DEL NORTE", "airport_count": 3 }, { "county": "MARIPOSA", "airport_count": 3 }, { "county": "SAN FRANCISCO", "airport_count": 3 }, { "county": "ALPINE", "airport_count": 2 }, { "county": "SIERRA", "airport_count": 2 }, { "county": "ESMERALDA", "airport_count": 1 } ] }, { "state": "NY", "airport_count": 576, "by_county": [ { "county": "SUFFOLK", "airport_count": 34 }, { "county": "ERIE", "airport_count": 26 }, { "county": "NIAGARA", "airport_count": 20 }, { "county": "DUTCHESS", "airport_count": 20 }, { "county": "ONONDAGA", "airport_count": 18 }, { "county": "ONEIDA", "airport_count": 18 }, { "county": "ULSTER", "airport_count": 17 }, { "county": "ORANGE", "airport_count": 16 }, { "county": "ONTARIO", "airport_count": 15 }, { "county": "NASSAU", "airport_count": 15 }, { "county": "CHAUTAUQUA", "airport_count": 14 }, { "county": "DELAWARE", "airport_count": 13 }, { "county": "ALBANY", "airport_count": 13 }, { "county": "MONROE", "airport_count": 13 }, { "county": "SARATOGA", "airport_count": 12 }, { "county": "MONTGOMERY", "airport_count": 12 }, { "county": "OSWEGO", "airport_count": 12 }, { "county": "QUEENS", "airport_count": 11 }, { "county": "WESTCHESTER", "airport_count": 11 }, { "county": "CATTARAUGUS", "airport_count": 11 }, { "county": "WAYNE", "airport_count": 10 }, { "county": "SULLIVAN", "airport_count": 10 }, { "county": "ESSEX", "airport_count": 10 }, { "county": "WYOMING", "airport_count": 9 }, { "county": "STEUBEN", "airport_count": 9 }, { "county": "SCHOHARIE", "airport_count": 9 }, { "county": "ORLEANS", "airport_count": 9 }, { "county": "JEFFERSON", "airport_count": 9 }, { "county": "CLINTON", "airport_count": 8 }, { "county": "COLUMBIA", "airport_count": 8 }, { "county": "FULTON", "airport_count": 8 }, { "county": "LIVINGSTON", "airport_count": 8 }, { "county": "CHEMUNG", "airport_count": 8 }, { "county": "CAYUGA", "airport_count": 8 }, { "county": "ST LAWRENCE", "airport_count": 8 }, { "county": "GREENE", "airport_count": 8 }, { "county": "TOMPKINS", "airport_count": 7 }, { "county": "RENSSELAER", "airport_count": 7 }, { "county": "BROOME", "airport_count": 7 }, { "county": "MADISON", "airport_count": 7 }, { "county": "SCHUYLER", "airport_count": 7 }, { "county": "SCHENECTADY", "airport_count": 7 }, { "county": "HERKIMER", "airport_count": 7 }, { "county": "HAMILTON", "airport_count": 7 }, { "county": "NEW YORK", "airport_count": 7 }, { "county": "OTSEGO", "airport_count": 6 }, { "county": "GENESEE", "airport_count": 6 }, { "county": "CHENANGO", "airport_count": 5 }, { "county": "WASHINGTON", "airport_count": 4 }, { "county": "ALLEGANY", "airport_count": 4 }, { "county": "SENECA", "airport_count": 4 }, { "county": "WARREN", "airport_count": 4 }, { "county": "TIOGA", "airport_count": 4 }, { "county": "ROCKLAND", "airport_count": 4 }, { "county": "YATES", "airport_count": 4 }, { "county": "CORTLAND", "airport_count": 2 }, { "county": "FRANKLIN", "airport_count": 2 }, { "county": "LEWIS", "airport_count": 1 }, { "county": "BRONX", "airport_count": 1 }, { "county": "KINGS", "airport_count": 1 }, { "county": "PUTNAM", "airport_count": 1 } ] } ]
WITH __stage0 AS ( SELECT group_set, base."state" as "state__0", CASE WHEN group_set=0 THEN COUNT(1) END as "airport_count__0", CASE WHEN group_set=1 THEN base."county" END as "county__1", CASE WHEN group_set=1 THEN COUNT(1) END as "airport_count__1" FROM '../data/airports.parquet' as base CROSS JOIN (SELECT UNNEST(GENERATE_SERIES(0,1,1)) as group_set ) as group_set WHERE (base."state"='CA') or (base."state"='NY') GROUP BY 1,2,4 ) SELECT "state__0" as "state", MAX(CASE WHEN group_set=0 THEN "airport_count__0" END) as "airport_count", COALESCE(LIST({ "county": "county__1", "airport_count": "airport_count__1"} ORDER BY "airport_count__1" desc NULLS LAST) FILTER (WHERE group_set=1),[]) as "by_county" FROM __stage0 GROUP BY 1 ORDER BY 2 desc NULLS LAST
Next, we'll use the output of that query as the input to another, where we determine which counties have the highest percentage of airports compared to the whole state, taking advantage of the nested structure of the data to to so.
run: airports -> { where: state = 'CA' | 'NY' group_by: state aggregate: airport_count nest: by_county is { group_by: county aggregate: airport_count } } -> { limit: 10; order_by: 4 desc select: by_county.county airports_in_county is by_county.airport_count airports_in_state is airport_count # percent percent_in_county is by_county.airport_count / airport_count }
[ { "county": "LOS ANGELES", "airports_in_county": 176, "airports_in_state": 984, "percent_in_county": 0.17886178861788618 }, { "county": "SAN BERNARDINO", "airports_in_county": 71, "airports_in_state": 984, "percent_in_county": 0.07215447154471545 }, { "county": "SUFFOLK", "airports_in_county": 34, "airports_in_state": 576, "percent_in_county": 0.059027777777777776 }, { "county": "ORANGE", "airports_in_county": 53, "airports_in_state": 984, "percent_in_county": 0.05386178861788618 }, { "county": "SAN DIEGO", "airports_in_county": 49, "airports_in_state": 984, "percent_in_county": 0.049796747967479675 } ]
WITH __stage0 AS ( SELECT group_set, base."state" as "state__0", CASE WHEN group_set=0 THEN COUNT(1) END as "airport_count__0", CASE WHEN group_set=1 THEN base."county" END as "county__1", CASE WHEN group_set=1 THEN COUNT(1) END as "airport_count__1" FROM '../data/airports.parquet' as base CROSS JOIN (SELECT UNNEST(GENERATE_SERIES(0,1,1)) as group_set ) as group_set WHERE (base."state"='CA') or (base."state"='NY') GROUP BY 1,2,4 ) , __stage1 AS ( SELECT "state__0" as "state", MAX(CASE WHEN group_set=0 THEN "airport_count__0" END) as "airport_count", COALESCE(LIST({ "county": "county__1", "airport_count": "airport_count__1"} ORDER BY "airport_count__1" desc NULLS LAST) FILTER (WHERE group_set=1),[]) as "by_county" FROM __stage0 GROUP BY 1 ) SELECT by_county_0."county" as "county", by_county_0."airport_count" as "airports_in_county", base."airport_count" as "airports_in_state", by_county_0."airport_count"*1.0/base."airport_count" as "percent_in_county" FROM __stage1 as base LEFT JOIN LATERAL (SELECT UNNEST(base."by_county"), 1 as ignoreme) as by_county_0_outer(by_county_0,ignoreme) ON by_county_0_outer.ignoreme=1 ORDER BY 4 desc NULLS LAST LIMIT 10
Aggregate Locality
When computing sum, avg, and count on fields in joined sources with one-to-many relationships, Malloy will automatically handle the duplication of rows that occurs in the join, and compute accurate aggregations on the fanned-out table. See the Aggregate Locality section for more information.
run: aircraft -> { aggregate: // The average number of seats on models of registered aircraft models_avg_seats is aircraft_models.seats.avg() // The average number of seats on registered aircraft aircraft_avg_seats is avg(aircraft_models.seats) }
[ { "models_avg_seats": 12.919491525423728, "aircraft_avg_seats": 8.671575437621561 } ]
SELECT ( SELECT AVG(a.val) as value FROM ( SELECT UNNEST(list(distinct {key:aircraft_models_0."aircraft_model_code", val: aircraft_models_0."seats"})) a ) ) as "models_avg_seats", AVG(aircraft_models_0."seats") as "aircraft_avg_seats" FROM '../data/aircraft.parquet' as base LEFT JOIN '../data/aircraft_models.parquet' AS aircraft_models_0 ON base."aircraft_model_code"=aircraft_models_0."aircraft_model_code"
Comments
Malloy code can include both line and block comments. Line comments, which begin with -- or //,
may appear anywhere within a line, and cause all subsequent characters on that line to be ignored.
Block comments, which are enclosed between /* and */, cause all enclosed characters to be ignored
and may span multiple lines.
-- The total number of flight entries run: flights -> { aggregate: flight_count // Defined simply as `count()` } /* * A comparison of the total number of flights * for each of the tracked carriers. */ run: flights -> { group_by: carrier aggregate: flight_count /* , total_distance */ }
Ordering and Limiting
In Malloy, ordering and limiting work pretty much the same way they do in SQL, though Malloy introduces some reasonable defaults.
The limit: statement limits the number of rows returned. Results below are sorted by the first measure descending—in this case, airport_count.
run: duckdb.table('../data/airports.parquet') -> { limit: 2 group_by: state aggregate: airport_count is count() }
[ { "state": "TX", "airport_count": 1845 }, { "state": "CA", "airport_count": 984 } ]
SELECT base."state" as "state", COUNT(1) as "airport_count" FROM '../data/airports.parquet' as base GROUP BY 1 ORDER BY 2 desc NULLS LAST LIMIT 2
Default ordering can be overridden with order_by:, as in the following query, which shows the states in alphabetical order. order_by: can take a field index number or the name of a field.
run: duckdb.table('../data/airports.parquet') -> { order_by: state group_by: state aggregate: airport_count is count() }
[ { "state": "AK", "airport_count": 608 }, { "state": "AL", "airport_count": 260 }, { "state": "AR", "airport_count": 299 }, { "state": "AS", "airport_count": 4 }, { "state": "AZ", "airport_count": 319 } ]
SELECT base."state" as "state", COUNT(1) as "airport_count" FROM '../data/airports.parquet' as base GROUP BY 1 ORDER BY 1 ASC NULLS LAST
Next Steps
This was a whirlwind tour of the syntax and features of Malloy. To continue on your Malloy journey:
Explore sample analyses and data models built in Malloy in our Patterns Github repo.
Learn how to connect Malloy to your own database.
Take a look at our guide for translating SQL to Malloy.
Join the Malloy community Slack channel!