Release Notes

v0.15.0

Breaking change: when a query is used in a scalar context, such as an IN expression, make it return the first column only (see #75). Previously, such query would SELECT NULL unless the query ends with an explicit Select().

v0.14.3

Fix MySQL reflection. Thanks to Alexander Plavin.

v0.14.2

Add DuckDB support. Thanks to Alexander Plavin.

v0.14.1

Fix Join incorrectly collapsing an outer branch when it may transform a NULL to a non-NULL value.
Make @funsql macro support operators ≥, ≤, ≢, ≡, ≠, ∉, ∈ as aliases for >=, <=, IS DISTINCT FROM, IS NOT DISTINCT FROM, <>, IN, NOT IN, thanks to Ashlin Harris.

v0.14.0

Define: add parameters before and after for specifying position of the defined columns.
Introduce the SQLColumn type to represent table columns. The type of SQLTable.columns is changed from Vector{Symbol} to OrderedDict{Symbol, SQLColumn}.
Make SQLTable an AbstractDict{Symbol, SQLColumn}.
Add DataAPI-compatible metadata to catalog objects SQLCatalog, SQLTable, and SQLColumn.
Add a field SQLString.columns with an optional Vector{SQLColumn} representing output columns of the SQL query.
Support docstrings in @funsql notation.
Remove support for const and variable assignment syntax from @funsql notation.
Use a simpler and more consistent rule for collapsing JOIN branches (fixes #60).

v0.13.2

Wrap a branch of UNION ALL in a subquery if it contains ORDER BY or LIMIT clause.

v0.13.1

Add support for grouping sets, which are used in SQL to calculate totals and subtotals. The Group() node accepts an optional parameter sets, which is either a grouping mode indicator :cube or :rollup, or a collection of grouping key sets Vector{Vector{Symbol}}. Examples:

From(:person) |> Group(:year_of_birth, sets = :cube)

From(:person) |> Group(:year_of_birth, :month_of_birth, sets = :rollup)

From(:person) |> Group(:year_of_birth, :gender_concept_id,
                       sets = [[:year_of_birth], [:gender_concept_id]])

v0.13.0

This release introduces some backward-incompatible changes. Before upgrading, please review these notes.

Type resolution has been refactored to allow assembling query fragments based on the type information.
Type checking is now more strict. A Define field or an optional Join will be validated even when they are to be elided.
Resolution of ambiguous column names in Join has been changed in favor of the right branch. Previously, an ambiguous name would cause an error.

Node-bound references are no longer supported. The following query will fail:

qₚ = From(:person)
qₗ = From(:location)
q = qₚ |>
    Join(qₗ, on = qₚ.location_id .== qₗ.location_id) |>
    Select(qₚ.person_id, qₗ.state)

Use nested references instead:

q = @funsql begin
    from(person)
    join(
        location => from(location),
        location_id == location.location_id)
    select(person_id, location.state)
end

v0.12.0

This release introduces some backward-incompatible changes. Before upgrading, please review these notes.

The SQLTable constructor drops the schema parameter. Instead, it now accepts an optional vector of qualifiers.
Bare Append(p, q) now produces p UNION ALL q. Previously, bare Append(p, q) would be interpreted as Select() |> Append(p, q).
Add Spark dialect.
Update nodes Group() and Partition() to accept an optional parameter name, which speficies the field that holds the group data.
Add Over() node such that p |> Over(q) is equivalent to q |> With(p).
Add the @funsql macro, described below.

`@funsql`

This release introduces @funsql macro, which provides a new, concise notation for building FunSQL queries. Example:

using FunSQL

q = @funsql from(person).filter(year_of_birth > 1950).select(person_id)

This is equivalent to the following query:

using FunSQL: From, Get, Select, Where

q = From(:person) |> Where(Get.year_of_birth .> 1950) |> Select(Get.person_id)

The @funsql notation reduces syntactic noise, making queries prettier and faster to write. Semantically, any query that can be constructed with @funsql could also be constructed without the macro, and vice versa. Moreover, these two notations could be freely mixed:

@funsql from(person).$(Where(Get.year_of_birth .> 1950)).select(person_id)

From(:person) |>
@funsql(filter(year_of_birth > 1950)) |>
Select(Get.person_id)

In @funsql notation, the chain operator (|>) is replaced either with period (.) or with block syntax:

@funsql begin
    from(person)
    filter(year_of_birth > 1950)
    select(person_id)
end

Names that are not valid identifiers should be wrapped in backticks:

@funsql begin
    from(person)
    define(`Patient's age` => 2024 - year_of_birth)
    filter(`Patient's age` >= 16)
end

Many SQL functions and operators are available out of the box:

@funsql from(location).define(city_state => concat(city, ", ", state))

@funsql from(person).filter(year_of_birth < 1900 || year_of_birth > 2024)

Comparison chaining is supported:

@funsql from(person).filter(1950 < year_of_birth < 2000)

The if statement and the ternary ? : operator are converted to a CASE expression:

@funsql begin
    from(person)
    define(
        generation =>
            year_of_birth <= 1964 ? "Baby Boomer" :
            year_of_birth <= 1980 ? "Generation X" :
            "Millenial")
end

A @funsql query can invoke any SQL function or even an arbitrary scalar SQL expression:

@funsql from(location).select(fun(`SUBSTRING(? FROM ? FOR ?)`, zip, 1, 3))

Aggregate and window functions are supported:

@funsql begin
    from(person)
    group()
    select(
        count(),
        min(year_of_birth),
        max(year_of_birth, filter = gender_concept_id == 8507),
        median =>
            agg(`(percentile_cont(0.5) WITHIN GROUP (ORDER BY ?))`, year_of_birth))
end

@funsql begin
    from(visit_occurrence)
    partition(person_id, order_by = [visit_start_date])
    filter(row_number() <= 1)
end

Custom scalar and aggregate functions can be integrated to @funsql notation:

const funsql_substring = FunSQL.FunClosure("SUBSTRING(? FROM ? FOR ?)")

@funsql from(location).select(substring(zip, 1, 3))

const funsql_median = FunSQL.AggClosure("(percentile_cont(0.5) WITHIN GROUP (ORDER BY ?))")

@funsql from(person).group().select(median(year_of_birth))

In general, any Julia function with a name funsql_f() can be invoked as f() within @funsql macro. For example:

funsql_concept(v, cs...) =
    @funsql from(concept).filter(vocabulary_id == $v && in(concept_id, $cs...))

funsql_ICD10CM(cs...) =
    @funsql concept("ICD10CM", $cs...)

For convenience, @funsql macro can wrap such function definitions:

@funsql concept(v, cs...) = begin
    from(concept)
    filter(vocabulary_id == $v && in(concept_id, $cs...))
end

@funsql ICD10CM(cs...) =
    concept("ICD10CM", $cs...)

Or even:

@funsql begin

concept(v, cs...) = begin
    from(concept)
    filter(vocabulary_id == $v && in(concept_id, $cs...))
end

ICD10CM(cs...) =
    concept("ICD10CM", $cs...)

end

Here are some other SQL features expressed in @funsql notation. JOIN and GROUP BY:

@funsql begin
    from(person)
    filter(between(year_of_birth, 1930, 1940))
    join(
        location => from(location).filter(state == "IL"),
        on = location_id == location.location_id)
    left_join(
        visit_group => begin
            from(visit_occurrence)
            group(person_id)
        end,
        on = person_id == visit_group.person_id)
    select(
        person_id,
        latest_visit_date => visit_group.max(visit_start_date))
end

ORDER BY and LIMIT:

@funsql from(person).order(year_of_birth.desc()).limit(10)

UNION ALL:

@funsql begin
    append(
        from(measurement).define(date => measurement_date),
        from(observation).define(date => observation_date))
end

Recursive queries:

@funsql begin
    define(n => 1, f => 1)
    iterate(define(n => n + 1, f => f * (n + 1)).filter(n <= 10))
end

Query parameters:

@funsql from(person).filter(year_of_birth >= :THRESHOLD)

v0.11.2

Fix a number of problems with serializing Order() and Limit() for MS SQL Server.
Add a column alias to the dummy NULL when generating zero-column output.

v0.11.1

No changes since the last release, enable Zenodo integration.

v0.11.0

This release introduces some backward-incompatible changes. Before upgrading, please review these notes.

The Iterate() node, which is used for assembling recursive queries, has been simplified. In the previous version, the argument of Iterate() must explicitly refer to the output of the previous iteration:
```
Base() |>
Iterate(From(:iteration) |>
        IterationStep() |>
        As(:iteration))
```
In v0.11.0, this query is written as:
```
Base() |>
Iterate(IterationStep())
```
Alternatively, the output of the previous iteration can be fetched with the From(^) node:
```
Base() |>
Iterate(From(^) |>
        IterationStep())
```
Function nodes with names Fun."&&", Fun."||", and Fun."!" are now serialized as logical operators AND, OR, and NOT. Broadcasting notation p .&& q, p .|| q, .!p is also supported.

FunSQL interpretation of || conflicts with SQL dialects that use || to represent string concatenation. Other dialects use function concat. In FunSQL, always use Fun.concat, which will pick correct serialization depending on the target dialect.
Rendering of Fun nodes can now be customized by overriding the method serialize!(), which is dispatched on the node name. The following names have customized serialization: and, between, case, cast, concat, current_date, current_timestamp, exists, extract, in, is_not_null, is_null, like, not, not_between, not_exists, not_in, not_like, or.
Introduce template notation for representing irregular SQL syntax. For example, Fun."SUBSTRING(? FROM ? FOR ?)"(Get.zip, 1, 3) generates SUBSTRING("location_1"."zip" FROM 1 to 3).

In general, the name of a Fun node is interpreted as a function name, an operator name, or a template string:
1. If the name has a custom serialize!() method, it is used for rendering the node. Example: Fun.in(Get.zip, "60614", "60615").
2. If the name contains a placeholder character ?, it is interpreted as a template. Placeholders are replaced with the arguments of the Fun node. Use ?? to represent a literal ? character. Wrap the template in parentheses if this is necessary to make the syntax unambiguous.
3. If the name contains only symbol characters, or if the name starts or ends with a space, it is interpreted as an operator name. Examples: Fun."-"(2020, Get.year_of_birth), Fun." ILIKE "(Get.city, "Ch%"), Fun." COLLATE \"C\""(Get.state), Fun."CURRENT_TIME "().
4. Otherwise, the name is interpreted as a function name. The function arguments are separated by a comma and wrapped in parentheses. Examples: Fun.now(), Fun.coalesce(Get.city, "N/A").
Agg nodes also support serialize!() and template notation. Custom serialization is provided for Agg.count() and Agg.count_distinct(arg).
Remove distinct flag from Agg nodes. To represent COUNT(DISTINCT …), use Agg.count_distinct(…). Otherwise, use template notation, e.g., Agg."string_agg(DISTINCT ?, ? ORDER BY ?)"(Get.state, ",", Get.state).
Function names are no longer automatically converted to upper case.
Remove ability to customize translation of Fun nodes to clause tree by overriding the translate() method.
Remove clauses OP, KW, CASE, which were previously used for expressing operators and irregular syntax.

Add support for table-valued functions. Such functions return tabular data and must appear in a FROM clause. Example:

From(Fun.regexp_matches("2,3,5,7,11", "(\\d+)", "g"),
     columns = [:captures]) |>
Select(Fun."CAST(?[1] AS INTEGER)"(Get.captures))

SELECT CAST("regexp_matches_1"."captures"[1] AS INTEGER) AS "_"
FROM regexp_matches('2,3,5,7,11', '(\d+)', 'g') AS "regexp_matches_1" ("captures")

To prevent duplicating SQL expressions, Define() and Group() may create an additional subquery (#11). Take, for example:

From(:person) |>
Define(:age => 2020 .- Get.year_of_birth) |>
Where(Get.age .>= 16) |>
Select(Get.person_id, Get.age)

In the previous version, the definition of age is replicated in both SELECT and WHERE clauses:

SELECT
  "person_1"."person_id",
  (2020 - "person_1"."year_of_birth") AS "age"
FROM "person" AS "person_1"
WHERE ((2020 - "person_1"."year_of_birth") >= 16)

In v0.11.0, age is evaluated once in a nested subquery:

SELECT
  "person_2"."person_id",
  "person_2"."age"
FROM (
  SELECT
    "person_1"."person_id",
    (2020 - "person_1"."year_of_birth") AS "age"
  FROM "person" AS "person_1"
) AS "person_2"
WHERE ("person_2"."age" >= 16)

Similarly, aggregate expressions used in Bind() are evaluated in a separate subquery (#12).
Fix a bug where FunSQL fails to generate a correct SELECT DISTINCT query when key columns of Group() are not used by the following nodes. For example, the following query pipeline would fail to render SQL:
```
From(:person) |>
Group(Get.year_of_birth) |>
Group() |>
Select(Agg.count())
```
Now it correctly renders:
```
SELECT count(*) AS "count"
FROM (
  SELECT DISTINCT "person_1"."year_of_birth"
  FROM "person" AS "person_1"
) AS "person_2"
```
Article Two Kinds of SQL Query Builders is added to documentation.
Where possible, pretty-printing replaces Lit nodes with their values.

v0.10.2

Generate aliases for CTE tables (fixes #33).

v0.10.1

From can take a DataFrame or any Tables.jl-compatible source.
More examples.

v0.10.0

Add SQLCatalog type that encapsulates information about database tables, the target SQL dialect, and a cache of rendered queries.
Add function reflect to retrieve information about available database tables.
From(::Symbol) can refer to a table in the database catalog.
Add a dependency on DBInterface package.
Add type SQLConnection <: DBInterface.Connection that combines a raw database connection and the database catalog. Also add SQLStatement <: DBInterface.Statement.
Implement DBInterface.connect to create a database connection and call reflect to retrieve the database catalog.
Implement DBInterface.prepare and DBInterface.execute to render a query node to SQL and immediately compile or execute it.
Allow render to take a SQLConnection or a SQLCatalog argument.
Rename SQLStatement to SQLString, drop the dialect field.
Update the guide and the examples to use the DBInterface API and improve many docstrings.

v0.9.2

Compatibility with PrettyPrinting 0.4.

v0.9.1

Join: add optional flag to omit the JOIN clause in the case when the data from the right branch is not used by the query.
With: add materialized flag to make CTEs with MATERIALIZED and NOT MATERIALIZED annotations.
Add WithExternal node that can be used to prepare the definition for a CREATE TABLE AS or a SELECT INTO statement.
Rearranging clause types: drop CTE clause; add columns to AS clause; add NOTE clause.

v0.9.0

Add Iterate node for making recursive queries.
Add With and From(::Symbol) nodes for assigning a name to an intermediate dataset.
From(nothing) for making a unit dataset.
Rename Select.list, Append.list, etc to args.
More documentation updates.

v0.8.2

Require Julia ≥ 1.6.
Render each argument on a separate line for SELECT, GROUP BY, ORDER BY, as well as for a top-level AND in WHERE and HAVING.
Improve SQLDialect interface.
Add Jacob's CASE example.

v0.8.1

Update documentation and examples.
Fix quoting of MySQL identifiers.

v0.8.0

Refactor the SQL translator to make it faster and easier to maintain.
Improve error messages.
Include columns added by Define to the output.
Report an error when Agg is used without Group.
Deduplicate identical aggregates in a Group subquery.
Support for WITH clause.
Update the Tutorial/Usage Guide.

v0.7.0

Add Order, Asc, Desc.
Add Limit.
Fun.in with a list, Fun.current_timestamp.

v0.6.0

Add Append for creating UNION ALL queries.

v0.5.1

Support for specifying the window frame.

v0.5.0

Add Define for calculated columns.
Correlated subqueries and JOIN LATERAL with Bind.
Support for query parameters with Var.
Support for subqueries.
Add Partition and window functions.

v0.4.0

Add Group.
Add aggregate expressions.
Collapse WHERE that follows GROUP to HAVING.
Add LeftJoin alias for Join(..., left = true).

v0.3.0

Add Join.

v0.2.0

Flatten nested SQL subqueries.
Broadcasting syntax for SQL functions.
Special Fun.<name> notation for SQL functions.
Support for irregular syntax for functions and operators.
Outline of the documentation.

v0.1.1

Use let notation for displaying SQLNode objects.
Add Highlight node.
Highlighting of problematic nodes in error messages.

v0.1.0

Initial release.
From, Select, and Where are implemented.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

NEWS.md

NEWS.md

Release Notes

v0.15.0

v0.14.3

v0.14.2

v0.14.1

v0.14.0

v0.13.2

v0.13.1

v0.13.0

v0.12.0

`@funsql`

v0.11.2

v0.11.1

v0.11.0

v0.10.2

v0.10.1

v0.10.0

v0.9.2

v0.9.1

v0.9.0

v0.8.2

v0.8.1

v0.8.0

v0.7.0

v0.6.0

v0.5.1

v0.5.0

v0.4.0

v0.3.0

v0.2.0

v0.1.1

v0.1.0

Files

NEWS.md

Latest commit

History

NEWS.md

File metadata and controls

Release Notes

v0.15.0

v0.14.3

v0.14.2

v0.14.1

v0.14.0

v0.13.2

v0.13.1

v0.13.0

v0.12.0

@funsql

v0.11.2

v0.11.1

v0.11.0

v0.10.2

v0.10.1

v0.10.0

v0.9.2

v0.9.1

v0.9.0

v0.8.2

v0.8.1

v0.8.0

v0.7.0

v0.6.0

v0.5.1

v0.5.0

v0.4.0

v0.3.0

v0.2.0

v0.1.1

v0.1.0

`@funsql`