Implemented groupBy for DataFrame

Author: olekscode

DataFrame-Core.package/DataFrame.class/instance/group.by..st

@@ -0,0 +1,12 @@
+as yet unclassified
+group: colNameOrArray by: colName
+
+	| left right |
+	
+	left := colNameOrArray isArray
+		ifTrue: [ self columns: colNameOrArray ]	"a DataFrame"
+		ifFalse: [ self column: colNameOrArray ].	"a DataSeries"
+		
+	right := self column: colName.
+	
+	^ left groupBy: right.

DataFrame-Core.package/DataFrame.class/instance/groupBy..st

@@ -0,0 +1,11 @@
+as yet unclassified
+groupBy: colName
+
+	| groupedColNames |
+	
+	"We exclude the column by which we are grouping"
+	groupedColNames := self columnNames copyWithout: colName.
+
+	^ DataFrameGrouped
+		group: (self columns: groupedColNames)
+		by: (self column: colName)

DataFrame-Core.package/DataFrameGrouped.class/class/group.by..st

@@ -0,0 +1,4 @@
+as yet unclassified
+group: aDataFrame by: aSeries
+
+	^ self new split: aDataFrame by: aSeries

DataFrame-Core.package/DataFrameGrouped.class/instance/apply..st

@@ -1,4 +1,20 @@
 private
 apply: aBlock
 
-	"TODO"
+	| colNames numberOfRows numberOfColumns result |
+	
+	colNames := (groups at: 1) columnNames.
+	
+	numberOfRows := groups size.
+	numberOfColumns := colNames size.
+	
+	result := DataFrame new: (numberOfRows @ numberOfColumns).
+	result rowNames: groups keys.
+	result columnNames: colNames.
+	
+	groups doWithIndex: [ :df :i |
+		1 to: colNames size do: [ :j |
+			result at: i at: j put:
+				(aBlock value: (df columnAt: j)) ] ].
+		
+	^ result

DataFrame-Core.package/DataFrameGrouped.class/instance/printOn..st

@@ -0,0 +1,13 @@
+printing
+printOn: aStream
+
+	super printOn: aStream.
+	aStream cr.
+	
+	groups doWithIndex: [ :eachDataFrame :i |
+		(groups keys at: i) printOn: aStream.
+		aStream cr.
+		eachDataFrame printOn: aStream.
+		
+		i = groups size
+			ifFalse: [ aStream cr; cr ] ] 

DataFrame-Core.package/DataFrameGrouped.class/instance/split.by..st

@@ -0,0 +1,26 @@
+private
+split: aDataFrame by: aSeries
+
+	| seriesUnique |
+	
+	aDataFrame numberOfRows = aSeries size
+		ifFalse: [ SizeMismatch signal ].
+	
+	seriesUnique := aSeries unique asArray.
+	
+	groups := seriesUnique collect: [ :eachUnique |
+		| aList df |
+		aList := LinkedList new.
+		
+		aSeries doWithIndex: [ :each :i |
+			each = eachUnique 
+				ifTrue: [ aList add: (aDataFrame rowAt: i) ] ].
+		
+		df := DataFrame fromRows: aList.
+		df columnNames: aDataFrame columnNames.
+		df ].
+	
+	groups := groups asDataSeries.
+	groups keys: seriesUnique.
+	
+	^ self

DataFrame-Core.package/DataFrameGrouped.class/properties.json

@@ -5,7 +5,9 @@
 	"classinstvars" : [ ],
 	"pools" : [ ],
 	"classvars" : [ ],
-	"instvars" : [ ],
+	"instvars" : [
+		"groups"
+	],
 	"name" : "DataFrameGrouped",
 	"type" : "normal"
 }

README.md

@@ -31,19 +31,38 @@ series keys: #(k1 k2 k3).
 ```
 
 ### Creating a DataFrame
-The easiest way to create an object of `DataFrame` class is by passing it an array of rows or columns (an array of arrays). Creating a DataFrame from rows allows us to see how the data will be arranged in a table. It's very readable and can be handy if we need to write some simple examples for testing
+There are four ways of creating a data frame:
+1. Creating an empty data frame, then filling it with data
+2. Creating a data frame from an array of rows
+3. Creating a data frame from an array of columns
+4. Reading data from a file
+
+#### Creating an empty DataFrame
+You can create an empty instance of `DataFrame` using the `new` message
+
+```smalltalk
+df := DataFrame new.
+```
+The data can be added later using the `add:` message.
+```smalltalk
+df add: #('Barcelona' 1.609 true).
+```
+
+#### Creating a DataFrame from an array of rows
+This way is the best for creating simple examples for testing since you can see how the data will be arranged in your data frame.
 
 ```smalltalk
-df := DataFrame fromRows: #(
+df := DataFrame rows: #(
    ('Barcelona' 1.609 true)
    ('Dubai' 2.789 true)
    ('London' 8.788 false)).
 ```
 
+#### Creating a DataFrame from an array of columns
 We can do the same by passing an array of columns
 
 ```smalltalk
-df := DataFrame fromColumns: #(
+df := DataFrame rows: #(
    ('Barcelona' 'Dubai' 'London')
    (1.609 2.789 8.788)
    (true true false)).