Skip to content

columns selector type #1274

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
+71 −46
Open

columns selector type #1274

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
61806b4
columns selector type
AndreiKingsley Jun 25, 2025
7977ff5
ColumnResolver docs fixes
AndreiKingsley Jun 26, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
117 changes: 71 additions & 46 deletions docs/StardustDocs/topics/ColumnSelectors.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,33 +45,34 @@ df.move { name.firstName and name.lastName }.after { city }
`first {}`, `firstCol()`, `last {}`, `lastCol()`, `single {}`, `singleCol()`

Returns the first, last, or single column from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet` that adheres to the optional given condition. If no column adheres to the given condition,
or [`ColumnSet`](#column-resolvers) that adheres to the optional given condition. If no column adheres to the given condition,
`NoSuchElementException` is thrown.

##### Col {collapsible="true"}
`col(name)`, `col(5)`

Creates a [ColumnAccessor](DataColumn.md) (or `SingleColumn`) for a column with the given
Creates a [`ColumnAccessor`](#column-resolvers) (or [`SingleColumn`](#column-resolvers)) for a column with the given
argument from the top-level or specified [column group](DataColumn.md#columngroup). The argument can be either an
index (`Int`) or a reference to a column (`String`, `ColumnPath`, `KProperty`, or `ColumnAccessor`;
index (`Int`) or a reference to a column (`String`, [`ColumnPath`](#column-resolvers), or
[`ColumnAccessor`](#column-resolvers);
any [AccessApi](apiLevels.md)).

##### Value Col, Frame Col, Col Group {collapsible="true"}
`valueCol(name)`, `valueCol(5)`, `frameCol(name)`, `frameCol(5)`, `colGroup(name)`, `colGroup(5)`

Creates a [ColumnAccessor](DataColumn.md) (or `SingleColumn`) for a
Creates a [`ColumnAccessor`](DataColumn.md) (or `SingleColumn`) for a
[value column](DataColumn.md#valuecolumn) / [frame column](DataColumn.md#framecolumn) /
[column group](DataColumn.md#columngroup) with the given argument from the top-level or
specified [column group](DataColumn.md#columngroup). The argument can be either an index (`Int`) or a reference
to a column (`String`, `ColumnPath`, `KProperty`, or `ColumnAccessor`; any [AccessApi](apiLevels.md)).
The functions can be both typed and untyped (in case you're supplying a column name, -path, or index).
to a column (`String`, [`ColumnPath`](#column-resolvers), or [`ColumnAccessor`](#column-resolvers); any [AccessApi](apiLevels.md)).
The functions can be both typed and untyped (in case you're supplying a column name, path, or index).
These functions throw an `IllegalArgumentException` if the column found is not the right kind.

##### Cols {collapsible="true"}
`cols {}`, `cols()`, `cols(colA, colB)`, `cols(1, 5)`, `cols(1..5)`, `[{}]`, `colSet[1, 3]`

Creates a subset of columns (`ColumnSet`) from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet`.
Creates a subset of columns ([`ColumnSet`](#column-resolvers)) from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers).
You can use either a `ColumnFilter`, or any of the `vararg` overloads for any [AccessApi](apiLevels.md).
The function can be both typed and untyped (in case you're supplying a column name, -path, or index (range)).

Expand All @@ -80,36 +81,36 @@ Note that you can also use the `[]` operator for most overloads of `cols` to ach
##### Range of Columns {collapsible="true"}
`colA.."colB"`

Creates a `ColumnSet` containing all columns from `colA` to `colB` (inclusive) from the top-level.
Creates a [`ColumnSet`](#column-resolvers) containing all columns from `colA` to `colB` (inclusive) from the top-level.
Columns inside [column groups](DataColumn.md#columngroup) are also supported
(as long as they share the same direct parent), as well as any combination of [AccessApi](apiLevels.md).

##### Value Columns, Frame Columns, Column Groups {collapsible="true"}
`valueCols {}`, `valueCols()`, `frameCols {}`, `frameCols()`, `colGroups {}`, `colGroups()`

Creates a subset of columns (`ColumnSet`) from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet` containing only [value columns](DataColumn.md#valuecolumn) / [frame columns](DataColumn.md#framecolumn) /
Creates a subset of columns ([`ColumnSet`](#column-resolvers)) from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers) containing only [value columns](DataColumn.md#valuecolumn) / [frame columns](DataColumn.md#framecolumn) /
[column groups](DataColumn.md#columngroup) that adhere to the optional condition.

##### Cols of Kind {collapsible="true"}
`colsOfKind(Value, Frame) {}`, `colsOfKind(Group, Frame)`

Creates a subset of columns (`ColumnSet`) from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet` containing only columns of the specified kind(s) that adhere to the optional condition.
Creates a subset of columns ([`ColumnSet`](#column-resolvers)) from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers) containing only columns of the specified kind(s) that adhere to the optional condition.

##### All (Cols) {collapsible="true"}
`all()`, `allCols()`

Creates a `ColumnSet` containing all columns from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet`. This is the opposite of [`none()`](ColumnSelectors.md#none) and equivalent to
Creates a [`ColumnSet`](#column-resolvers) containing all columns from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers). This is the opposite of [`none()`](ColumnSelectors.md#none) and equivalent to
[`cols()`](ColumnSelectors.md#cols) without filter.
Note, on [column groups](DataColumn.md#columngroup), `all` is named `allCols` instead to avoid confusion.

##### All (Cols) After, -Before, -From, -Up To {collapsible="true"}
`allAfter(colA)`, `allBefore(colA)`, `allColsFrom(colA)`, `allColsUpTo(colA)`

Creates a `ColumnSet` containing a subset of columns from the top-level,
specified [column group](DataColumn.md#columngroup), or `ColumnSet`.
Creates a [`ColumnSet`](#column-resolvers) containing a subset of columns from the top-level,
specified [column group](DataColumn.md#columngroup), or [`ColumnSet`](#column-resolvers).
The subset includes:
- `all(Cols)Before(colA)`: All columns before the specified column, excluding that column.
- `all(Cols)After(colA)`: All columns after the specified column, excluding that column.
Expand All @@ -123,10 +124,10 @@ On `ColumnSets` they are a `ColumnFilter` instead.
##### Cols at any Depth {collapsible="true"}
`colsAtAnyDepth {}`, `colsAtAnyDepth()`

Creates a `ColumnSet` containing all columns from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet` at any depth if they satisfy the optional given predicate. This means that columns (of all three kinds!)
Creates a [`ColumnSet`](#column-resolvers) containing all columns from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers) at any depth if they satisfy the optional given predicate. This means that columns (of all three kinds!)
nested inside [column groups](DataColumn.md#columngroup) are also included.
This function can also be followed by another `ColumnSet` filter-function like `colsOf<>()`, `single()`,
This function can also be followed by another [`ColumnSet`](#column-resolvers) filter-function like `colsOf<>()`, `single()`,
or `valueCols()`.

**For example:**
Expand Down Expand Up @@ -165,8 +166,8 @@ All value columns at any depth nested under a column group named "myColGroup":
##### Cols in Groups {collapsible="true"}
`colsInGroups {}`, `colsInGroups()`

Creates a `ColumnSet` containing all columns that are nested in the [column groups](DataColumn.md#columngroup) at
the top-level, specified [column group](DataColumn.md#columngroup), or `ColumnSet` adhering to an optional predicate.
Creates a [`ColumnSet`](#column-resolvers) containing all columns that are nested in the [column groups](DataColumn.md#columngroup) at
the top-level, specified [column group](DataColumn.md#columngroup), or [`ColumnSet`](#column-resolvers) adhering to an optional predicate.
This is useful if you want to select all columns that are "one level down".

This function used to be called `children()` in the past.
Expand All @@ -186,28 +187,28 @@ or with filter:

`df.select { colsInGroups { "user" in it.name } }`

Similarly, you can take the columns inside all [column groups](DataColumn.md#columngroup) in a `ColumnSet`:
Similarly, you can take the columns inside all [column groups](DataColumn.md#columngroup) in a [`ColumnSet`](#column-resolvers):

`df.select { colGroups { "my" in it.name }.colsInGroups() }`

##### Take (Last) (Cols) (While) {collapsible="true"}
`take(5)`, `takeLastCols(2)`, `takeLastWhile {}`, `takeColsWhile {}`,

Creates a `ColumnSet` containing the first / last `n` columns from the top-level,
specified [column group](DataColumn.md#columngroup), or `ColumnSet` or those that adhere to the given condition.
Creates a [`ColumnSet`](#column-resolvers) containing the first / last `n` columns from the top-level,
specified [column group](DataColumn.md#columngroup), or [`ColumnSet`](#column-resolvers) or those that adhere to the given condition.
Note, to avoid ambiguity, `take` is called `takeCols` when called on a [column group](DataColumn.md#columngroup).

##### Drop (Last) (Cols) (While) {collapsible="true"}
`drop(5)`, `dropLastCols(2)`, `dropLastWhile {}`, `dropColsWhile {}`

Creates a `ColumnSet` without the first / last `n` columns from the top-level,
specified [column group](DataColumn.md#columngroup), or `ColumnSet` or those that adhere to the given condition.
Creates a [`ColumnSet`](#column-resolvers) without the first / last `n` columns from the top-level,
specified [column group](DataColumn.md#columngroup), or [`ColumnSet`](#column-resolvers) or those that adhere to the given condition.
Note, to avoid ambiguity, `drop` is called `dropCols` when called on a [column group](DataColumn.md#columngroup).

##### Select from [Column Group](DataColumn.md#columngroup) {collapsible="true"}
`colGroupA.select {}`, `"colGroupA" {}`

Creates a `ColumnSet` containing the columns selected by a `ColumnsSelector` relative to the specified
Creates a [`ColumnSet`](#column-resolvers) containing the columns selected by a `ColumnsSelector` relative to the specified
[column group](DataColumn.md#columngroup). In practice, this means you're opening a new selection DSL scope inside a
[column group](DataColumn.md#columngroup) and selecting columns from there.
The selected columns are referenced individually and "unpacked" from their parent
Expand Down Expand Up @@ -242,14 +243,14 @@ This function is best explained in parts:

**On Column Sets:** `except {}`

This function can be explained the easiest with a `ColumnSet`.
This function can be explained the easiest with a [`ColumnSet`](#column-resolvers).
Let's say we want all `Int` columns apart from `age` and `height`.

We can do:

`df.select { colsOf<Int>() except (age and height) }`

which will 'subtract' the `ColumnSet` created by `age and height` from the `ColumnSet` created by
which will 'subtract' the [`ColumnSet`](#column-resolvers) created by `age and height` from the [`ColumnSet`](#column-resolvers) created by
[`colsOf<Int>()`](ColumnSelectors.md#cols-of).

This operation can also be used to exclude columns that are originally in [column groups](DataColumn.md#columngroup).
Expand All @@ -261,7 +262,7 @@ For instance, excluding `userData.age`:
Note that the selection of columns to exclude from column sets is always done relative to the outer scope.
Use the [Extension Properties API](extensionPropertiesApi.md) to prevent scoping issues if possible.

> Special case: If a column that needs to be removed appears multiple times in the `ColumnSet`,
> Special case: If a column that needs to be removed appears multiple times in the [`ColumnSet`](#column-resolvers),
> it is excepted each time it is encountered (including inside [Column Groups](DataColumn.md#columngroup)).
> You could say the receiver `ColumnSet` is [simplified](ColumnSelectors.md#simplify) before the operation is performed:
>
Expand Down Expand Up @@ -319,24 +320,24 @@ or:
##### Column Name Filters {collapsible="true"}
`nameContains()`, `colsNameContains()`, `nameStartsWith()`, `colsNameEndsWith()`

Creates a `ColumnSet` containing columns from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet` that have names that satisfy the given function. These functions accept a `String` as argument, as
Creates a [`ColumnSet`](#column-resolvers) containing columns from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers) that have names that satisfy the given function. These functions accept a `String` as argument, as
well as an optional `ignoreCase` parameter. For the `nameContains` variant, you can also pass a `Regex` as an argument.
Note, on [column groups](DataColumn.md#columngroup), the functions have names starting with `cols` to avoid
ambiguity.

##### (Cols) Without Nulls {collapsible="true"}
`withoutNulls()`, `colsWithoutNulls()`

Creates a `ColumnSet` containing columns from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet` that have no `null` values. This is a shorthand for `cols { !it.hasNulls() }`.
Creates a [`ColumnSet`](#column-resolvers) containing columns from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers) that have no `null` values. This is a shorthand for `cols { !it.hasNulls() }`.
Note, to avoid ambiguity, `withoutNulls` is called `colsWithoutNulls` when called on a
[column group](DataColumn.md#columngroup).

##### Distinct {collapsible="true"}
`colSet.distinct()`

Returns a new `ColumnSet` from the specified `ColumnSet` containing only distinct columns (by path).
Returns a new [`ColumnSet`](#column-resolvers) from the specified [`ColumnSet`](#column-resolvers) containing only distinct columns (by path).
This is useful when you've selected the same column multiple times but only want it once.

This does not cover the case where a column is selected individually and through its enclosing
Expand All @@ -348,30 +349,30 @@ For this, you'll need to [rename](ColumnSelectors.md#rename) one of the columns.
##### None {collapsible="true"}
`none()`

Creates an empty `ColumnSet`, essentially selecting no columns at all.
Creates an empty [`ColumnSet`](#column-resolvers), essentially selecting no columns at all.
This is the opposite of [`all()`](ColumnSelectors.md#all-cols).

This function mostly exists for completeness, but can be useful in some very specific cases.

##### Cols Of {collapsible="true"}
`colsOf<T>()`, `colsOf<T> {}`

Creates a `ColumnSet` containing columns from the top-level, specified [column group](DataColumn.md#columngroup),
or `ColumnSet` that are a subtype of the specified type `T` and adhere to the optional condition.
Creates a [`ColumnSet`](#column-resolvers) containing columns from the top-level, specified [column group](DataColumn.md#columngroup),
or [`ColumnSet`](#column-resolvers) that are a subtype of the specified type `T` and adhere to the optional condition.

##### Simplify {collapsible="true"}
`colSet.simplify()`

Returns a new `ColumnSet` from the specified `ColumnSet` in 'simplified' form.
This function simplifies the structure of the `ColumnSet` by removing columns that are already present in
Returns a new [`ColumnSet`](#column-resolvers) from the specified [`ColumnSet`](#column-resolvers) in 'simplified' form.
This function simplifies the structure of the [`ColumnSet`](#column-resolvers) by removing columns that are already present in
[column groups](DataColumn.md#columngroup), returning only these groups,
plus columns not belonging in any of the groups.

In other words, this means that if a column in the `ColumnSet` is inside a [column group](DataColumn.md#columngroup)
in the `ColumnSet`, it will not be included in the result.
In other words, this means that if a column in the [`ColumnSet`](#column-resolvers) is inside a [column group](DataColumn.md#columngroup)
in the [`ColumnSet`](#column-resolvers), it will not be included in the result.

It's useful in combination with [`colsAtAnyDepth {}`](ColumnSelectors.md#cols-at-any-depth), as that function can
create a `ColumnSet` containing both a column and the [column group](DataColumn.md#columngroup) it's in.
create a [`ColumnSet`](#column-resolvers) containing both a column and the [column group](DataColumn.md#columngroup) it's in.

In the past, was named `top()` and `roots()`, but these names have been deprecated.

Expand All @@ -382,13 +383,13 @@ In the past, was named `top()` and `roots()`, but these names have been deprecat
##### Filter {collapsible="true"}
`colSet.filter {}`

Returns a new `ColumnSet` from the specified `ColumnSet` containing only columns that satisfy the given condition.
Returns a new [`ColumnSet`](#column-resolvers) from the specified [`ColumnSet`](#column-resolvers) containing only columns that satisfy the given condition.
This function behaves the same as [`cols {}` and `[{}]`](ColumnSelectors.md#cols), but only exists on column sets.

##### And {collapsible="true"}
`colSet and colB`

Creates a `ColumnSet` containing the columns from both the left and right side of the function. This allows
Creates a [`ColumnSet`](#column-resolvers) containing the columns from both the left and right side of the function. This allows
you to combine selections or simply select multiple columns at once.

Any combination of [AccessApi](apiLevels.md) can be used on either side of the `and` operator.
Expand Down Expand Up @@ -595,3 +596,27 @@ df.select { (colsOf<Int>() and age).distinct() }

<inline-frame src="resources/org.jetbrains.kotlinx.dataframe.samples.api.Access.columnSelectorsModifySet.html" width="100%"/>
<!---END-->

### Column Resolvers

`ColumnsResolver` is the base type used to resolve columns within the **Columns Selection DSL**,
as well as the return type of columns selection expressions.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe write it like "column(s)", because it's the return type of both the singular and multiple columns dsl

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I use it as generalized name of Dsl


All functions described above for selecting columns in various ways return a `ColumnResolver` of a specific kind:

- **`SingleColumn`** — resolves to a single [`DataColumn`](DataColumn.md).
- **`ColumnAccessor`** — a specialized `SingleColumn` with a defined path and type argument.
It can also be renamed during selection.
- **`ColumnPath`** — a wrapper for a [`DataColumn`](DataColumn.md) path
in a [`DataFrame`](DataFrame.md) also can serve as a `ColumnAccessor`.
```kotlin
// Select all columns from the group by path "group2"/"info":
df.select { pathOf("group2", "info").allCols() }
// For each selected column, place it under its ancestor group
// from two levels up in the column path hierarchy:
df.group { colsAtAnyDepth().colsOf<String>() }
.into { it.path.dropLast(2) }
```
- **`ColumnSet`** — resolves to an ordered list of [`DataColumn`s](DataColumn.md).