The Split-Apply-Combine Strategy
Design of the split-apply-combine support
Many data analysis tasks involve three steps:
- splitting a data set into groups,
- applying some functions to each of the groups,
- combining the results.
Note that any of the steps 1 and 3 of this general procedure can be dropped, in which case we just transform a data frame without grouping it and later combining the result.
A standardized framework for handling this sort of computation is described in the paper "The Split-Apply-Combine Strategy for Data Analysis", written by Hadley Wickham.
The DataFrames package supports the split-apply-combine strategy through the groupby
function that creates a GroupedDataFrame
, followed by combine
, select
/select!
or transform
/transform!
.
All operations described in this section of the manual are supported both for AbstractDataFrame
(when split and combine steps are skipped) and GroupedDataFrame
. Technically, AbstractDataFrame
is just considered as being grouped on no columns (meaning it has a single group, or zero groups if it is empty). The only difference is that in this case the keepkeys
and ungroup
keyword arguments (described below) are not supported and a data frame is always returned, as there are no split and combine steps in this case.
In order to perform operations by groups you first need to create a GroupedDataFrame
object from your data frame using the groupby
function that takes two arguments: (1) a data frame to be grouped, and (2) a set of columns to group by.
Operations can then be applied on each group using one of the following functions:
combine
: does not put restrictions on number of rows returned per group; the returned values are vertically concatenaded following order of groups inGroupedDataFrame
; it is typically used to compute summary statistics by group; forGroupedDataFrame
if grouping columns are kept they are put as first columns in the result;select
: return a data frame with the number and order of rows exactly the same as the source data frame, including only new calculated columns;select!
is an in-place version ofselect
;transform
: return a data frame with the number and order of rows exactly the same as the source data frame, including all columns from the source and new calculated columns;transform!
is an in-place version oftransform
; existing columns in the source data frame are put as first columns in the result;
As a special case, if a GroupedDataFrame
that has zero groups is passed then the result of the operation is determined by performing a single call to the transformation function with a 0-row argument passed to it. The output of this operation is only used to identify the number and type of produced columns, but the result has zero rows.
All these functions take a specification of one or more functions to apply to each subset of the DataFrame
. This specification can be of the following forms:
- standard column selectors (integers,
Symbol
s, strings, vectors of integers, vectors ofSymbol
s, vectors of strings,All
,Cols
,:
,Between
,Not
and regular expressions) - a
cols => function
pair indicating thatfunction
should be called with positional arguments holding columnscols
, which can be any valid column selector; in this case target column name is automatically generated and it is assumed thatfunction
returns a single value or a vector; the generated name is created by concatenating source column name andfunction
name by default (see examples below). - a
cols => function => target_cols
form additionally explicitly specifying the target column or columns, which must be a single name (as aSymbol
or a string), a vector of names orAsTable
. Additionally it can be aFunction
which takes a string or a vector of strings as an argument containing names of columns selected bycols
, and returns the target columns names (all accepted types exceptAsTable
are allowed). - a
col => target_cols
pair, which renames the columncol
totarget_cols
, which must be single name (as aSymbol
or a string), a vector of names orAsTable
. - column-independent operations
function => target_cols
or justfunction
for specificfunction
s where the input columns are omitted; withouttarget_cols
the new column has the same name asfunction
, otherwise it must be single name (as aSymbol
or a string). Supportedfunction
s are:nrow
to efficiently compute the number of rows in each group.proprow
to efficiently compute the proportion of rows in each group.eachindex
to return a vector holding the number of each row within each group.groupindices
to return the group number.
- vectors or matrices containing transformations specified by the
Pair
syntax described in points 2 to 5 - a function which will be called with a
SubDataFrame
corresponding to each group if aGroupedDataFrame
is processed, or with the data frame itself if anAbstractDataFrame
is processed; this form should be avoided due to its poor performance unless the number of groups is small or a very large number of columns are processed (in which caseSubDataFrame
avoids excessive compilation)
Note! If the expression of the form x => y
is passed then except for the special convenience form nrow => target_cols
it is always interpreted as cols => function
. In particular the following expression function => target_cols
is not a valid transformation specification.
Note! If cols
or target_cols
are one of All
, Cols
, Between
, or Not
, broadcasting using .=>
is supported and is equivalent to broadcasting the result of names(df, cols)
or names(df, target_cols)
. This behaves as if broadcasting happened after replacing the selector with selected column names within the data frame scope.
All functions have two types of signatures. One of them takes a GroupedDataFrame
as the first argument and an arbitrary number of transformations described above as following arguments. The second type of signature is when a Function
or a Type
is passed as the first argument and a GroupedDataFrame
as the second argument (similar to map
).
As a special rule, with the cols => function
and cols => function => target_cols
syntaxes, if cols
is wrapped in an AsTable
object then a NamedTuple
containing columns selected by cols
is passed to function
.
What is allowed for function
to return is determined by the target_cols
value:
- If both
cols
andtarget_cols
are omitted (so only afunction
is passed), then returning a data frame, a matrix, aNamedTuple
, or aDataFrameRow
will produce multiple columns in the result. Returning any other value produces a single column. - If
target_cols
is aSymbol
or a string then the function is assumed to return a single column. In this case returning a data frame, a matrix, aNamedTuple
, or aDataFrameRow
raises an error. - If
target_cols
is a vector ofSymbol
s or strings orAsTable
it is assumed thatfunction
returns multiple columns. Iffunction
returns one ofAbstractDataFrame
,NamedTuple
,DataFrameRow
,AbstractMatrix
then rules described in point 1 above apply. Iffunction
returns anAbstractVector
then each element of this vector must support thekeys
function, which must return a collection ofSymbol
s, strings or integers; the return value ofkeys
must be identical for all elements. Then as many columns are created as there are elements in the return value of thekeys
function. Iftarget_cols
isAsTable
then their names are set to be equal to the key names except ifkeys
returns integers, in which case they are prefixed byx
(so the column names are e.g.x1
,x2
, ...). Iftarget_cols
is a vector ofSymbol
s or strings then column names produced using the rules above are ignored and replaced bytarget_cols
(the number of columns must be the same as the length oftarget_cols
in this case). Iffun
returns a value of any other type then it is assumed that it is a table conforming to the Tables.jl API and theTables.columntable
function is called on it to get the resulting columns and their names. The names are retained whentarget_cols
isAsTable
and are replaced iftarget_cols
is a vector ofSymbol
s or strings.
In all of these cases, function
can return either a single row or multiple rows. As a particular rule, values wrapped in a Ref
or a 0
-dimensional AbstractArray
are unwrapped and then treated as a single row.
select
/select!
and transform
/transform!
always return a data frame with the same number and order of rows as the source (even if GroupedDataFrame
had its groups reordered), except when selection results in zero columns in the resulting data frame (in which case the result has zero rows).
For combine
, rows in the returned object appear in the order of groups in the GroupedDataFrame
. The functions can return an arbitrary number of rows for each group, but the kind of returned object and the number and names of columns must be the same for all groups, except when a DataFrame()
or NamedTuple()
is returned, in which case a given group is skipped.
It is allowed to mix single values and vectors if multiple transformations are requested. In this case single value will be repeated to match the length of columns specified by returned vectors.
By default (threads=true
) a separate task is spawned for each specified transformation; each transformation then spawns as many tasks as Julia threads, and splits processing of groups across them (however, currently transformations with optimized implementations like sum
and transformations that return multiple rows use a single task for all groups). This allows for parallel operation when Julia was started with more than one thread. Passed transformation functions must therefore not modify global variables (i.e. they must be pure), use locks to control parallel accesses, or threads=false
must be passed to disable multithreading.
To apply function
to each row instead of whole columns, it can be wrapped in a ByRow
struct. cols
can be any column indexing syntax, in which case function
will be passed one argument for each of the columns specified by cols
or a NamedTuple
of them if specified columns are wrapped in AsTable
. If ByRow
is used it is allowed for cols
to select an empty set of columns, in which case function
is called for each row without any arguments and an empty NamedTuple
is passed if empty set of columns is wrapped in AsTable
.
The following keyword arguments are supported by the transformation functions (not all keyword arguments are supported in all cases; in general they are allowed in situations when they are meaningful, see the documentation of the specific functions for details):
keepkeys
: whether grouping columns should be kept in the returned data frame.ungroup
: whether the return value of the operation should be a data frame or aGroupedDataFrame
.copycols
: whether columns of the source data frame should be copied if no transformation is applied to them.renamecols
: whether in thecols => function
form automatically generated column names should include the name of transformation functions or not.threads
: whether transformations may be run in separate tasks which can execute in parallel
Examples of the split-apply-combine operations
We show several examples of these functions applied to the iris
dataset below:
julia> using DataFrames, CSV, Statistics
julia> iris = CSV.read((joinpath(dirname(pathof(DataFrames)),
"..", "docs", "src", "assets", "iris.csv")),
DataFrame)
150×5 DataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼──────────────────────────────────────────────────────────────────
1 │ 5.1 3.5 1.4 0.2 Iris-setosa
2 │ 4.9 3.0 1.4 0.2 Iris-setosa
3 │ 4.7 3.2 1.3 0.2 Iris-setosa
4 │ 4.6 3.1 1.5 0.2 Iris-setosa
5 │ 5.0 3.6 1.4 0.2 Iris-setosa
6 │ 5.4 3.9 1.7 0.4 Iris-setosa
7 │ 4.6 3.4 1.4 0.3 Iris-setosa
8 │ 5.0 3.4 1.5 0.2 Iris-setosa
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
144 │ 6.8 3.2 5.9 2.3 Iris-virginica
145 │ 6.7 3.3 5.7 2.5 Iris-virginica
146 │ 6.7 3.0 5.2 2.3 Iris-virginica
147 │ 6.3 2.5 5.0 1.9 Iris-virginica
148 │ 6.5 3.0 5.2 2.0 Iris-virginica
149 │ 6.2 3.4 5.4 2.3 Iris-virginica
150 │ 5.9 3.0 5.1 1.8 Iris-virginica
135 rows omitted
julia> iris_gdf = groupby(iris, :Species)
GroupedDataFrame with 3 groups based on key: Species
First Group (50 rows): Species = "Iris-setosa"
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼───────────────────────────────────────────────────────────────
1 │ 5.1 3.5 1.4 0.2 Iris-setosa
2 │ 4.9 3.0 1.4 0.2 Iris-setosa
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
49 │ 5.3 3.7 1.5 0.2 Iris-setosa
50 │ 5.0 3.3 1.4 0.2 Iris-setosa
46 rows omitted
⋮
Last Group (50 rows): Species = "Iris-virginica"
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼──────────────────────────────────────────────────────────────────
1 │ 6.3 3.3 6.0 2.5 Iris-virginica
2 │ 5.8 2.7 5.1 1.9 Iris-virginica
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
50 │ 5.9 3.0 5.1 1.8 Iris-virginica
47 rows omitted
julia> combine(iris_gdf, :PetalLength => mean)
3×2 DataFrame
Row │ Species PetalLength_mean
│ String15 Float64
─────┼───────────────────────────────────
1 │ Iris-setosa 1.464
2 │ Iris-versicolor 4.26
3 │ Iris-virginica 5.552
julia> combine(iris_gdf, nrow, proprow, groupindices)
3×4 DataFrame
Row │ Species nrow proprow groupindices
│ String15 Int64 Float64 Int64
─────┼────────────────────────────────────────────────
1 │ Iris-setosa 50 0.333333 1
2 │ Iris-versicolor 50 0.333333 2
3 │ Iris-virginica 50 0.333333 3
julia> combine(iris_gdf, nrow, :PetalLength => mean => :mean)
3×3 DataFrame
Row │ Species nrow mean
│ String15 Int64 Float64
─────┼─────────────────────────────────
1 │ Iris-setosa 50 1.464
2 │ Iris-versicolor 50 4.26
3 │ Iris-virginica 50 5.552
julia> combine(iris_gdf,
[:PetalLength, :SepalLength] =>
((p, s) -> (a=mean(p)/mean(s), b=sum(p))) =>
AsTable) # multiple columns are passed as arguments
3×3 DataFrame
Row │ Species a b
│ String15 Float64 Float64
─────┼────────────────────────────────────
1 │ Iris-setosa 0.292449 73.2
2 │ Iris-versicolor 0.717655 213.0
3 │ Iris-virginica 0.842744 277.6
julia> combine(iris_gdf,
AsTable([:PetalLength, :SepalLength]) =>
x -> std(x.PetalLength) / std(x.SepalLength)) # passing a NamedTuple
3×2 DataFrame
Row │ Species PetalLength_SepalLength_function
│ String15 Float64
─────┼───────────────────────────────────────────────────
1 │ Iris-setosa 0.492245
2 │ Iris-versicolor 0.910378
3 │ Iris-virginica 0.867923
julia> combine(x -> std(x.PetalLength) / std(x.SepalLength), iris_gdf) # passing a SubDataFrame
3×2 DataFrame
Row │ Species x1
│ String15 Float64
─────┼───────────────────────────
1 │ Iris-setosa 0.492245
2 │ Iris-versicolor 0.910378
3 │ Iris-virginica 0.867923
julia> combine(iris_gdf, 1:2 => cor, nrow)
3×3 DataFrame
Row │ Species SepalLength_SepalWidth_cor nrow
│ String15 Float64 Int64
─────┼────────────────────────────────────────────────────
1 │ Iris-setosa 0.74678 50
2 │ Iris-versicolor 0.525911 50
3 │ Iris-virginica 0.457228 50
julia> combine(iris_gdf, :PetalLength => (x -> [extrema(x)]) => [:min, :max])
3×3 DataFrame
Row │ Species min max
│ String15 Float64 Float64
─────┼───────────────────────────────────
1 │ Iris-setosa 1.0 1.9
2 │ Iris-versicolor 3.0 5.1
3 │ Iris-virginica 4.5 6.9
To get row number for each observation within each group use the eachindex
function:
julia> combine(iris_gdf, eachindex)
150×2 DataFrame
Row │ Species eachindex
│ String15 Int64
─────┼───────────────────────────
1 │ Iris-setosa 1
2 │ Iris-setosa 2
3 │ Iris-setosa 3
⋮ │ ⋮ ⋮
148 │ Iris-virginica 48
149 │ Iris-virginica 49
150 │ Iris-virginica 50
144 rows omitted
Contrary to combine
, the select
and transform
functions always return a data frame with the same number and order of rows as the source. In the example below the return values in columns :SepalLength_SepalWidth_cor
and :nrow
are broadcasted to match the number of elements in each group:
julia> select(iris_gdf, 1:2 => cor)
150×2 DataFrame
Row │ Species SepalLength_SepalWidth_cor
│ String Float64
─────┼────────────────────────────────────────────
1 │ Iris-setosa 0.74678
2 │ Iris-setosa 0.74678
3 │ Iris-setosa 0.74678
4 │ Iris-setosa 0.74678
⋮ │ ⋮ ⋮
148 │ Iris-virginica 0.457228
149 │ Iris-virginica 0.457228
150 │ Iris-virginica 0.457228
143 rows omitted
julia> transform(iris_gdf, :Species => x -> chop.(x, head=5, tail=0))
150×6 DataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species Species_function
│ Float64 Float64 Float64 Float64 String SubString…
─────┼────────────────────────────────────────────────────────────────────────────────────
1 │ 5.1 3.5 1.4 0.2 Iris-setosa setosa
2 │ 4.9 3.0 1.4 0.2 Iris-setosa setosa
3 │ 4.7 3.2 1.3 0.2 Iris-setosa setosa
4 │ 4.6 3.1 1.5 0.2 Iris-setosa setosa
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
148 │ 6.5 3.0 5.2 2.0 Iris-virginica virginica
149 │ 6.2 3.4 5.4 2.3 Iris-virginica virginica
150 │ 5.9 3.0 5.1 1.8 Iris-virginica virginica
143 rows omitted
All functions also support the do
block form. However, as noted above, this form is slow and should therefore be avoided when performance matters.
julia> combine(iris_gdf) do df
(m = mean(df.PetalLength), s² = var(df.PetalLength))
end
3×3 DataFrame
Row │ Species m s²
│ String15 Float64 Float64
─────┼─────────────────────────────────────
1 │ Iris-setosa 1.464 0.0301061
2 │ Iris-versicolor 4.26 0.220816
3 │ Iris-virginica 5.552 0.304588
To apply a function to each non-grouping column of a GroupedDataFrame
you can write:
julia> combine(iris_gdf, valuecols(iris_gdf) .=> mean)
3×5 DataFrame
Row │ Species SepalLength_mean SepalWidth_mean PetalLength_mean P ⋯
│ String15 Float64 Float64 Float64 F ⋯
─────┼──────────────────────────────────────────────────────────────────────────
1 │ Iris-setosa 5.006 3.418 1.464 ⋯
2 │ Iris-versicolor 5.936 2.77 4.26
3 │ Iris-virginica 6.588 2.974 5.552
1 column omitted
Note that GroupedDataFrame
is a view: therefore grouping columns of its parent data frame must not be mutated, and rows must not be added nor removed from it. If the number or rows of the parent changes then an error is thrown when a child GroupedDataFrame
is used:
julia> df = DataFrame(id=1:2)
2×1 DataFrame
Row │ id
│ Int64
─────┼───────
1 │ 1
2 │ 2
julia> gd = groupby(df, :id)
GroupedDataFrame with 2 groups based on key: id
First Group (1 row): id = 1
Row │ id
│ Int64
─────┼───────
1 │ 1
⋮
Last Group (1 row): id = 2
Row │ id
│ Int64
─────┼───────
1 │ 2
julia> push!(df, [3])
3×1 DataFrame
Row │ id
│ Int64
─────┼───────
1 │ 1
2 │ 2
3 │ 3
julia> gd[1]
ERROR: AssertionError: The current number of rows in the parent data frame is 3 and it does not match the number of rows it contained when GroupedDataFrame was created which was 2. The number of rows in the parent data frame has likely been changed unintentionally (e.g. using subset!, filter!, deleteat!, push!, or append! functions).
Sometimes it is useful to append rows to the source data frame of a GroupedDataFrame
, without affecting the rows used for grouping. In such a scenario you can create the grouped data frame using a view
of the parent data frame to avoid the error:
julia> df = DataFrame(id=1:2)
2×1 DataFrame
Row │ id
│ Int64
─────┼───────
1 │ 1
2 │ 2
julia> gd = groupby(view(df, :, :), :id)
GroupedDataFrame with 2 groups based on key: id
First Group (1 row): id = 1
Row │ id
│ Int64
─────┼───────
1 │ 1
⋮
Last Group (1 row): id = 2
Row │ id
│ Int64
─────┼───────
1 │ 2
julia> push!(df, [3])
3×1 DataFrame
Row │ id
│ Int64
─────┼───────
1 │ 1
2 │ 2
3 │ 3
julia> gd[1]
1×1 SubDataFrame
Row │ id
│ Int64
─────┼───────
1 │ 1
Using GroupedDataFrame
as an iterable and indexable object
If you only want to split the data set into subsets, use the groupby
function. You can then iterate SubDataFrame
s that constitute the identified groups:
julia> for subdf in iris_gdf
println(size(subdf, 1))
end
50
50
50
To also get the values of the grouping columns along with each group, use the pairs
function:
julia> for (key, subdf) in pairs(iris_gdf)
println("Number of data points for $(key.Species): $(nrow(subdf))")
end
Number of data points for Iris-setosa: 50
Number of data points for Iris-versicolor: 50
Number of data points for Iris-virginica: 50
The value of key
in the example above where we iterated pairs(iris_gdf)
is a DataFrames.GroupKey
object, which can be used in a similar fashion to a NamedTuple
.
Grouping a data frame using the groupby
function can be seen as adding a lookup key to it. Such lookups can be performed efficiently by indexing the resulting GroupedDataFrame
with DataFrames.GroupKey
(as it was presented above) a Tuple
, a NamedTuple
, or a dictionary. Here are some more examples of such indexing.
julia> iris_gdf[(Species="Iris-virginica",)] # a NamedTuple
50×5 SubDataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼──────────────────────────────────────────────────────────────────
1 │ 6.3 3.3 6.0 2.5 Iris-virginica
2 │ 5.8 2.7 5.1 1.9 Iris-virginica
3 │ 7.1 3.0 5.9 2.1 Iris-virginica
4 │ 6.3 2.9 5.6 1.8 Iris-virginica
5 │ 6.5 3.0 5.8 2.2 Iris-virginica
6 │ 7.6 3.0 6.6 2.1 Iris-virginica
7 │ 4.9 2.5 4.5 1.7 Iris-virginica
8 │ 7.3 2.9 6.3 1.8 Iris-virginica
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
44 │ 6.8 3.2 5.9 2.3 Iris-virginica
45 │ 6.7 3.3 5.7 2.5 Iris-virginica
46 │ 6.7 3.0 5.2 2.3 Iris-virginica
47 │ 6.3 2.5 5.0 1.9 Iris-virginica
48 │ 6.5 3.0 5.2 2.0 Iris-virginica
49 │ 6.2 3.4 5.4 2.3 Iris-virginica
50 │ 5.9 3.0 5.1 1.8 Iris-virginica
35 rows omitted
julia> iris_gdf[[("Iris-virginica",), ("Iris-setosa",)]] # a vector of Tuples
GroupedDataFrame with 2 groups based on key: Species
First Group (50 rows): Species = "Iris-virginica"
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼──────────────────────────────────────────────────────────────────
1 │ 6.3 3.3 6.0 2.5 Iris-virginica
2 │ 5.8 2.7 5.1 1.9 Iris-virginica
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
49 │ 6.2 3.4 5.4 2.3 Iris-virginica
50 │ 5.9 3.0 5.1 1.8 Iris-virginica
46 rows omitted
⋮
Last Group (50 rows): Species = "Iris-setosa"
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼───────────────────────────────────────────────────────────────
1 │ 5.1 3.5 1.4 0.2 Iris-setosa
2 │ 4.9 3.0 1.4 0.2 Iris-setosa
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
50 │ 5.0 3.3 1.4 0.2 Iris-setosa
47 rows omitted
julia> key = keys(iris_gdf) |> last # last key in iris_gdf
GroupKey: (Species = String15("Iris-virginica"),)
julia> iris_gdf[key]
50×5 SubDataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼──────────────────────────────────────────────────────────────────
1 │ 6.3 3.3 6.0 2.5 Iris-virginica
2 │ 5.8 2.7 5.1 1.9 Iris-virginica
3 │ 7.1 3.0 5.9 2.1 Iris-virginica
4 │ 6.3 2.9 5.6 1.8 Iris-virginica
5 │ 6.5 3.0 5.8 2.2 Iris-virginica
6 │ 7.6 3.0 6.6 2.1 Iris-virginica
7 │ 4.9 2.5 4.5 1.7 Iris-virginica
8 │ 7.3 2.9 6.3 1.8 Iris-virginica
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
44 │ 6.8 3.2 5.9 2.3 Iris-virginica
45 │ 6.7 3.3 5.7 2.5 Iris-virginica
46 │ 6.7 3.0 5.2 2.3 Iris-virginica
47 │ 6.3 2.5 5.0 1.9 Iris-virginica
48 │ 6.5 3.0 5.2 2.0 Iris-virginica
49 │ 6.2 3.4 5.4 2.3 Iris-virginica
50 │ 5.9 3.0 5.1 1.8 Iris-virginica
35 rows omitted
julia> iris_gdf[Dict("Species" => "Iris-setosa")] # a dictionary
50×5 SubDataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼───────────────────────────────────────────────────────────────
1 │ 5.1 3.5 1.4 0.2 Iris-setosa
2 │ 4.9 3.0 1.4 0.2 Iris-setosa
3 │ 4.7 3.2 1.3 0.2 Iris-setosa
4 │ 4.6 3.1 1.5 0.2 Iris-setosa
5 │ 5.0 3.6 1.4 0.2 Iris-setosa
6 │ 5.4 3.9 1.7 0.4 Iris-setosa
7 │ 4.6 3.4 1.4 0.3 Iris-setosa
8 │ 5.0 3.4 1.5 0.2 Iris-setosa
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
44 │ 5.0 3.5 1.6 0.6 Iris-setosa
45 │ 5.1 3.8 1.9 0.4 Iris-setosa
46 │ 4.8 3.0 1.4 0.3 Iris-setosa
47 │ 5.1 3.8 1.6 0.2 Iris-setosa
48 │ 4.6 3.2 1.4 0.2 Iris-setosa
49 │ 5.3 3.7 1.5 0.2 Iris-setosa
50 │ 5.0 3.3 1.4 0.2 Iris-setosa
35 rows omitted
Note that although GroupedDataFrame
is iterable and indexable it is not an AbstractVector
. For this reason currently it was decided that it does not support map
nor broadcasting (to allow for making a decision in the future what result type they should produce). To apply a function to all groups of a data frame and get a vector of results either use a comprehension or collect
GroupedDataFrame
into a vector first. Here are examples of both approaches:
julia> sdf_vec = collect(iris_gdf)
3-element Vector{Any}:
50×5 SubDataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼───────────────────────────────────────────────────────────────
1 │ 5.1 3.5 1.4 0.2 Iris-setosa
2 │ 4.9 3.0 1.4 0.2 Iris-setosa
3 │ 4.7 3.2 1.3 0.2 Iris-setosa
4 │ 4.6 3.1 1.5 0.2 Iris-setosa
5 │ 5.0 3.6 1.4 0.2 Iris-setosa
6 │ 5.4 3.9 1.7 0.4 Iris-setosa
7 │ 4.6 3.4 1.4 0.3 Iris-setosa
8 │ 5.0 3.4 1.5 0.2 Iris-setosa
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
44 │ 5.0 3.5 1.6 0.6 Iris-setosa
45 │ 5.1 3.8 1.9 0.4 Iris-setosa
46 │ 4.8 3.0 1.4 0.3 Iris-setosa
47 │ 5.1 3.8 1.6 0.2 Iris-setosa
48 │ 4.6 3.2 1.4 0.2 Iris-setosa
49 │ 5.3 3.7 1.5 0.2 Iris-setosa
50 │ 5.0 3.3 1.4 0.2 Iris-setosa
35 rows omitted
50×5 SubDataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼───────────────────────────────────────────────────────────────────
1 │ 7.0 3.2 4.7 1.4 Iris-versicolor
2 │ 6.4 3.2 4.5 1.5 Iris-versicolor
3 │ 6.9 3.1 4.9 1.5 Iris-versicolor
4 │ 5.5 2.3 4.0 1.3 Iris-versicolor
5 │ 6.5 2.8 4.6 1.5 Iris-versicolor
6 │ 5.7 2.8 4.5 1.3 Iris-versicolor
7 │ 6.3 3.3 4.7 1.6 Iris-versicolor
8 │ 4.9 2.4 3.3 1.0 Iris-versicolor
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
44 │ 5.0 2.3 3.3 1.0 Iris-versicolor
45 │ 5.6 2.7 4.2 1.3 Iris-versicolor
46 │ 5.7 3.0 4.2 1.2 Iris-versicolor
47 │ 5.7 2.9 4.2 1.3 Iris-versicolor
48 │ 6.2 2.9 4.3 1.3 Iris-versicolor
49 │ 5.1 2.5 3.0 1.1 Iris-versicolor
50 │ 5.7 2.8 4.1 1.3 Iris-versicolor
35 rows omitted
50×5 SubDataFrame
Row │ SepalLength SepalWidth PetalLength PetalWidth Species
│ Float64 Float64 Float64 Float64 String15
─────┼──────────────────────────────────────────────────────────────────
1 │ 6.3 3.3 6.0 2.5 Iris-virginica
2 │ 5.8 2.7 5.1 1.9 Iris-virginica
3 │ 7.1 3.0 5.9 2.1 Iris-virginica
4 │ 6.3 2.9 5.6 1.8 Iris-virginica
5 │ 6.5 3.0 5.8 2.2 Iris-virginica
6 │ 7.6 3.0 6.6 2.1 Iris-virginica
7 │ 4.9 2.5 4.5 1.7 Iris-virginica
8 │ 7.3 2.9 6.3 1.8 Iris-virginica
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮
44 │ 6.8 3.2 5.9 2.3 Iris-virginica
45 │ 6.7 3.3 5.7 2.5 Iris-virginica
46 │ 6.7 3.0 5.2 2.3 Iris-virginica
47 │ 6.3 2.5 5.0 1.9 Iris-virginica
48 │ 6.5 3.0 5.2 2.0 Iris-virginica
49 │ 6.2 3.4 5.4 2.3 Iris-virginica
50 │ 5.9 3.0 5.1 1.8 Iris-virginica
35 rows omitted
julia> map(nrow, sdf_vec)
3-element Vector{Int64}:
50
50
50
julia> nrow.(sdf_vec)
3-element Vector{Int64}:
50
50
50
Since GroupedDataFrame
is iterable, you can achieve the same result with a comprehension:
julia> [nrow(sdf) for sdf in iris_gdf]
3-element Vector{Int64}:
50
50
50
Note that using the split-apply-combine strategy with the operation specification syntax in combine
, select
or transform
will usually be faster for large GroupedDataFrame
objects than iterating them, with the difference that they produce a data frame. An operation corresponding to the example above is:
julia> combine(iris_gdf, nrow)
3×2 DataFrame
Row │ Species nrow
│ String15 Int64
─────┼────────────────────────
1 │ Iris-setosa 50
2 │ Iris-versicolor 50
3 │ Iris-virginica 50
Simulating the SQL where
clause
You can conveniently work on subsets of a data frame by using SubDataFrame
s. Operations performed on such objects can either create a new data frame or be performed in-place. Here are some examples:
julia> df = DataFrame(a=1:5)
5×1 DataFrame
Row │ a
│ Int64
─────┼───────
1 │ 1
2 │ 2
3 │ 3
4 │ 4
5 │ 5
julia> sdf = @view df[2:3, :]
2×1 SubDataFrame
Row │ a
│ Int64
─────┼───────
1 │ 2
2 │ 3
julia> transform(sdf, :a => ByRow(string)) # create a new data frame
2×2 DataFrame
Row │ a a_string
│ Int64 String
─────┼─────────────────
1 │ 2 2
2 │ 3 3
julia> transform!(sdf, :a => ByRow(string)) # update the source df in-place
2×2 SubDataFrame
Row │ a a_string
│ Int64 String?
─────┼─────────────────
1 │ 2 2
2 │ 3 3
julia> df # new column was created filled with missing in filtered-out rows
5×2 DataFrame
Row │ a a_string
│ Int64 String?
─────┼─────────────────
1 │ 1 missing
2 │ 2 2
3 │ 3 3
4 │ 4 missing
5 │ 5 missing
julia> select!(sdf, :a => -, renamecols=false) # update the source df in-place
2×1 SubDataFrame
Row │ a
│ Int64
─────┼───────
1 │ -2
2 │ -3
julia> df # the column replaced an existing column; previously stored values are re-used in filtered-out rows
5×1 DataFrame
Row │ a
│ Int64
─────┼───────
1 │ 1
2 │ -2
3 │ -3
4 │ 4
5 │ 5
Similar operations can be performed on GroupedDataFrame
as well:
julia> df = DataFrame(a=[1, 1, 1, 2, 2, 3], b=1:6)
6×2 DataFrame
Row │ a b
│ Int64 Int64
─────┼──────────────
1 │ 1 1
2 │ 1 2
3 │ 1 3
4 │ 2 4
5 │ 2 5
6 │ 3 6
julia> sdf = @view df[2:4, :]
3×2 SubDataFrame
Row │ a b
│ Int64 Int64
─────┼──────────────
1 │ 1 2
2 │ 1 3
3 │ 2 4
julia> gsdf = groupby(sdf, :a)
GroupedDataFrame with 2 groups based on key: a
First Group (2 rows): a = 1
Row │ a b
│ Int64 Int64
─────┼──────────────
1 │ 1 2
2 │ 1 3
⋮
Last Group (1 row): a = 2
Row │ a b
│ Int64 Int64
─────┼──────────────
1 │ 2 4
julia> transform(gsdf, nrow) # create a new data frame
3×3 DataFrame
Row │ a b nrow
│ Int64 Int64 Int64
─────┼─────────────────────
1 │ 1 2 2
2 │ 1 3 2
3 │ 2 4 1
julia> transform!(gsdf, nrow, :b => :b_copy)
3×4 SubDataFrame
Row │ a b nrow b_copy
│ Int64 Int64 Int64? Int64?
─────┼──────────────────────────────
1 │ 1 2 2 2
2 │ 1 3 2 3
3 │ 2 4 1 4
julia> df
6×4 DataFrame
Row │ a b nrow b_copy
│ Int64 Int64 Int64? Int64?
─────┼────────────────────────────────
1 │ 1 1 missing missing
2 │ 1 2 2 2
3 │ 1 3 2 3
4 │ 2 4 1 4
5 │ 2 5 missing missing
6 │ 3 6 missing missing
julia> select!(gsdf, :b_copy, :b => sum, renamecols=false)
3×3 SubDataFrame
Row │ a b_copy b
│ Int64 Int64? Int64
─────┼──────────────────────
1 │ 1 2 5
2 │ 1 3 5
3 │ 2 4 4
julia> df
6×3 DataFrame
Row │ a b_copy b
│ Int64 Int64? Int64
─────┼───────────────────────
1 │ 1 missing 1
2 │ 1 2 5
3 │ 1 3 5
4 │ 2 4 4
5 │ 2 missing 5
6 │ 3 missing 6
Column-independent operations
The operation specification language used with combine
, select
and transform
supports the following column-independent operations:
- getting the number of rows in a group (
nrow
); - getting the proportion of rows in a group (
proprow
); - getting the group number (
groupindices
); - getting a vector of indices within groups (
eachindex
).
These operations are column-independent, because they do not require specifying the input column name in the operation specification syntax.
These four exceptions to the standard operation specification syntax were introduced for user convenience as these operations are often needed in practice.
Below each of them is explained by example.
First create a data frame we will work with:
julia> df = DataFrame(customer_id=["a", "b", "b", "b", "c", "c"],
transaction_id=[12, 15, 19, 17, 13, 11],
volume=[2, 3, 1, 4, 5, 9])
6×3 DataFrame
Row │ customer_id transaction_id volume
│ String Int64 Int64
─────┼─────────────────────────────────────
1 │ a 12 2
2 │ b 15 3
3 │ b 19 1
4 │ b 17 4
5 │ c 13 5
6 │ c 11 9
julia> gdf = groupby(df, :customer_id, sort=true);
julia> show(gdf, allgroups=true)
GroupedDataFrame with 3 groups based on key: customer_id
Group 1 (1 row): customer_id = "a"
Row │ customer_id transaction_id volume
│ String Int64 Int64
─────┼─────────────────────────────────────
1 │ a 12 2
Group 2 (3 rows): customer_id = "b"
Row │ customer_id transaction_id volume
│ String Int64 Int64
─────┼─────────────────────────────────────
1 │ b 15 3
2 │ b 19 1
3 │ b 17 4
Group 3 (2 rows): customer_id = "c"
Row │ customer_id transaction_id volume
│ String Int64 Int64
─────┼─────────────────────────────────────
1 │ c 13 5
2 │ c 11 9
Getting the number of rows
You can get the number of rows per group in a GroupedDataFrame
by just writing nrow
, in which case the generated column name with the number of rows is :nrow
:
julia> combine(gdf, nrow)
3×2 DataFrame
Row │ customer_id nrow
│ String Int64
─────┼────────────────────
1 │ a 1
2 │ b 3
3 │ c 2
Additionally you are allowed to pass target column name:
julia> combine(gdf, nrow => "transaction_count")
3×2 DataFrame
Row │ customer_id transaction_count
│ String Int64
─────┼────────────────────────────────
1 │ a 1
2 │ b 3
3 │ c 2
Note that in both cases we did not pass source column name as it is not needed to determine the number of rows per group. This is the reason why column-independent operations are exceptions to standard operation specification syntax.
The nrow
expression also works in the operation specification syntax applied to a data frame. Here is an example:
julia> combine(df, nrow => "transaction_count")
1×1 DataFrame
Row │ transaction_count
│ Int64
─────┼───────────────────
1 │ 6
Finally, recall that nrow
is also a regular function that returns a number of rows in a data frame:
julia> nrow(df)
6
This dual use of nrow
does not lead to ambiguities, and is meant to make it easier to remember this exception.
Getting the proportion of rows
If you want to get a proportion of rows per group in a GroupedDataFrame
you can use the proprow
and proprow => [target column name]
column-independent operations. Here are some examples:
julia> combine(gdf, proprow)
3×2 DataFrame
Row │ customer_id proprow
│ String Float64
─────┼───────────────────────
1 │ a 0.166667
2 │ b 0.5
3 │ c 0.333333
julia> combine(gdf, proprow => "transaction_fraction")
3×2 DataFrame
Row │ customer_id transaction_fraction
│ String Float64
─────┼───────────────────────────────────
1 │ a 0.166667
2 │ b 0.5
3 │ c 0.333333
As opposed to nrow
, proprow
cannot be used outside of the operation specification syntax and is only allowed when processing a GroupedDataFrame
.
Getting the group number
Another common operation is getting group number. Use the groupindices
and groupindices => [target column name]
column-independent operations to get it:
julia> combine(gdf, groupindices)
3×2 DataFrame
Row │ customer_id groupindices
│ String Int64
─────┼───────────────────────────
1 │ a 1
2 │ b 2
3 │ c 3
julia> transform(gdf, groupindices)
6×4 DataFrame
Row │ customer_id transaction_id volume groupindices
│ String Int64 Int64 Int64
─────┼───────────────────────────────────────────────────
1 │ a 12 2 1
2 │ b 15 3 2
3 │ b 19 1 2
4 │ b 17 4 2
5 │ c 13 5 3
6 │ c 11 9 3
julia> combine(gdf, groupindices => "group_number")
3×2 DataFrame
Row │ customer_id group_number
│ String Int64
─────┼───────────────────────────
1 │ a 1
2 │ b 2
3 │ c 3
Outside of the operation specification syntax, groupindices
is also a regular function which returns group indices for each row in the parent data frame of the passed GroupedDataFrame
:
julia> groupindices(gdf)
6-element Vector{Union{Missing, Int64}}:
1
2
2
2
3
3
Getting a vector of indices within groups
The last column-independent operation supported by the operation specification syntax is getting the index of each row within each group:
julia> combine(gdf, eachindex)
6×2 DataFrame
Row │ customer_id eachindex
│ String Int64
─────┼────────────────────────
1 │ a 1
2 │ b 1
3 │ b 2
4 │ b 3
5 │ c 1
6 │ c 2
julia> select(gdf, eachindex, groupindices)
6×3 DataFrame
Row │ customer_id eachindex groupindices
│ String Int64 Int64
─────┼──────────────────────────────────────
1 │ a 1 1
2 │ b 1 2
3 │ b 2 2
4 │ b 3 2
5 │ c 1 3
6 │ c 2 3
julia> combine(gdf, eachindex => "transaction_number")
6×2 DataFrame
Row │ customer_id transaction_number
│ String Int64
─────┼─────────────────────────────────
1 │ a 1
2 │ b 1
3 │ b 2
4 │ b 3
5 │ c 1
6 │ c 2
Note that this operation also makes sense in a data frame context, where all rows are considered to be in the same group:
julia> transform(df, eachindex)
6×4 DataFrame
Row │ customer_id transaction_id volume eachindex
│ String Int64 Int64 Int64
─────┼────────────────────────────────────────────────
1 │ a 12 2 1
2 │ b 15 3 2
3 │ b 19 1 3
4 │ b 17 4 4
5 │ c 13 5 5
6 │ c 11 9 6
Finally recall that eachindex
is a standard function for getting all indices in an array. This similarity of functionality was the reason why this name was picked:
julia> collect(eachindex(df.customer_id))
6-element Vector{Int64}:
1
2
3
4
5
6
This, for example, means that in the following example the two created columns have the same contents:
julia> combine(gdf, eachindex, :customer_id => eachindex)
6×3 DataFrame
Row │ customer_id eachindex customer_id_eachindex
│ String Int64 Int64
─────┼───────────────────────────────────────────────
1 │ a 1 1
2 │ b 1 1
3 │ b 2 2
4 │ b 3 3
5 │ c 1 1
6 │ c 2 2
Column-independent operations versus functions
When discussing column-independent operations it is important to remember that operation specification syntax allows you to pass a function (without source and target column names), in which case such a function gets passed a SubDataFrame
that represents a group in a GroupedDataFrame
. Here is an example comparing a column-independent operation and a function:
julia> combine(gdf, eachindex, sdf -> axes(sdf, 1))
6×3 DataFrame
Row │ customer_id eachindex x1
│ String Int64 Int64
─────┼───────────────────────────────
1 │ a 1 1
2 │ b 1 1
3 │ b 2 2
4 │ b 3 3
5 │ c 1 1
6 │ c 2 2
Notice that the column-independent operation eachindex
produces the same result as using the anonymous function sdf -> axes(sdf, 1)
that takes a SubDataFrame
as its first argument and returns indices along its first axes. Importantly if it wasn't defined as a column-independent operation the eachindex
function would fail when being passed as you can see here:
julia> combine(gdf, sdf -> eachindex(sdf))
ERROR: MethodError: no method matching keys(::SubDataFrame{DataFrame, DataFrames.Index, Vector{Int64}})
The reason for this error is that the eachindex
function does not allow passing a SubDataFrame
as its argument.
The same applies to proprow
and groupindices
: they would not work with a SubDataFrame
as stand-alone functions.
The nrow
column-independent operation is a different case, as the nrow
function accepts SubDataFrame
as an argument:
julia> combine(gdf, nrow, sdf -> nrow(sdf))
3×3 DataFrame
Row │ customer_id nrow x1
│ String Int64 Int64
─────┼───────────────────────────
1 │ a 1 1
2 │ b 3 3
3 │ c 2 2
Notice that columns :nrow
and :x1
have identical contents, but the difference is that they do not have the same names. nrow
is a column-independent operation generating the :nrow
column name by default with number of rows per group. On the other hand, the sdf -> nrow(sdf)
anonymous function does gets a SubDataFrame
as its argument and returns its number of rows. The :x1
column name is the default auto-generated column name when processing anonymous functions.
Passing a function taking a SubDataFrame
is a flexible functionality allowing you to perform complex operations on your data. However, you should bear in mind two aspects:
- Using the full operation specification syntax (where source and target column names are passed) or column-independent operations will lead to faster execution of your code (as the Julia compiler is able to better optimize execution of such operations) in comparison to passing a function taking a
SubDataFrame
. - Although writing
nrow
,proprow
,groupindices
, andeachindex
looks like just passing a function they internally do not take aSubDataFrame
as their argument. As we explained in this section,proprow
,groupindices
, andeachindex
would not work withSubDataFrame
as their argument, andnrow
would work, but would produce a different column name. Instead, these four operations are special column-independent operations that are exceptions to the standard operation specification syntax rules. They were added for user convenience.
Specifying group order in groupby
By default order of groups produced by groupby
is undefined. If you want the order of groups to follow the order of first appereance in the source data frame of a grouping key then pass the sort=false
keyword argument to groupby
:
julia> push!(df, ["a", 100, 100]) # push row with large integer values to disable default sorting
7×3 DataFrame
Row │ customer_id transaction_id volume
│ String Int64 Int64
─────┼─────────────────────────────────────
1 │ a 12 2
2 │ b 15 3
3 │ b 19 1
4 │ b 17 4
5 │ c 13 5
6 │ c 11 9
7 │ a 100 100
julia> keys(groupby(df, :volume))
7-element DataFrames.GroupKeys{GroupedDataFrame{DataFrame}}:
GroupKey: (volume = 2,)
GroupKey: (volume = 3,)
GroupKey: (volume = 1,)
GroupKey: (volume = 4,)
GroupKey: (volume = 5,)
GroupKey: (volume = 9,)
GroupKey: (volume = 100,)
If you want to have them sorted in ascending order pass sort=true
:
julia> keys(groupby(df, :volume, sort=true))
7-element DataFrames.GroupKeys{GroupedDataFrame{DataFrame}}:
GroupKey: (volume = 1,)
GroupKey: (volume = 2,)
GroupKey: (volume = 3,)
GroupKey: (volume = 4,)
GroupKey: (volume = 5,)
GroupKey: (volume = 9,)
GroupKey: (volume = 100,)
You can also use the order
wrapper when passing a column name to group by or pass a named tuple as sort
keyword argument containing one or more of alg
, lt
, by
, rev
, and order
fields that will be treated just like in sortperm
:
julia> keys(groupby(df, [:customer_id, order(:volume, rev=true)]))
6-element DataFrames.GroupKeys{GroupedDataFrame{DataFrame}}:
GroupKey: (customer_id = "a", volume = 2)
GroupKey: (customer_id = "b", volume = 4)
GroupKey: (customer_id = "b", volume = 3)
GroupKey: (customer_id = "b", volume = 1)
GroupKey: (customer_id = "c", volume = 9)
GroupKey: (customer_id = "c", volume = 5)
julia> keys(groupby(df, :customer_id, sort=(rev=true,)))
3-element DataFrames.GroupKeys{GroupedDataFrame{DataFrame}}:
GroupKey: (customer_id = "c",)
GroupKey: (customer_id = "b",)
GroupKey: (customer_id = "a",)