Branch: master
Find file Copy path
5e3f016 May 27, 2018
1 contributor

Users who have contributed to this file

132 lines (77 sloc) 6.4 KB

Key Concepts

This document explains the key concepts of Data-Forge.


A series is an indexed sequence of values and is implemented by the Series class. By default a series has an integer index starting at 0 and counting up (just like a JavaScript array). Series will often be used with a date-time index, something usually known as a time series.

All values in a series are generally expected to have the same type, although this is not specifically a requirement of Data-Forge.

A series can easily be constructed from a JavaScript array of data. It is also easily exported back to a JavaScript array. We can also extract a series from a column in a dataframe. Or plug a series into an existing dataframe to create a new column.


A dataframe is really the main concept. It is implemented by the DataFrame class.

A dataframe contains a sequence of rows. Each row is just a JavaScript object where each field contains the value for a column of data at that row. You might also think of it as a matrix (rows and columns) of structured data. You can also think of it as a spreadsheet in memory.

Although a dataframe ostensibly represents tabular data, it's actually pretty flexible and different rows can contain different data types and field values can contain deeply nested data.

A dataframe is composed of multiple series, where each series has a name and represents a column of tabular data.

A dataframe can be easily constructed from a JavaScript array containing data or parsed from CSV and JSON data formats. A dataframe can be exported back to a JavaScript array or serialized/stringified to CSV or JSON.

DataFrame and Series are very similar although not exactly the same. For example, because a dataframe has columns it therefore has functions such as getSeries for working with its columns. Series on the hand is specialised for working with a sequence of values and thus has functions that dataframe does not have, such as average which computes the average value of the series.


A column is a single named series of data contained within a dataframe. Each column is simply a series with a name, the values of the series are the values of the column.

A column can be thought of a as a slice of data that cuts through all rows of the dataframe.


An index is a sequence of values that is used to index a dataframe or series. When the data is a time-series the index is expected to contain Date values.

An index is used for operations that search and merge data-farmes and series.

If not specified an integer index (starting at 0) is generated based on row position. An index can be explicitly by specifying a column by name, from a JavaScript array of data or generated on the fly from another column or series.


Through the documentation and the code you will occasionally see a reference to a pair or pairs. Series and dataframes are actually sequences of pairs, where each pair contains a index and a value or row.

Lazy evaluation

Dataframe, series and index are only fully evaluated when necessary. Operations are queued up (like a data-pipeline) and only fully evaluated as needed and when required, for example when serializing to csv or json (eg toCSV or toJSON) or when baking to values (eg toArray or toRows).

A dataframe, series or index can be forcibly evaluated and baked into memory by calling the bake function.

Iterable / Iterator

These are JavaScript concepts defined in ES6 and they are at the core of Data-Forge. To use Data-Forge you don't really need to know about these - but it's useful to understand them to understand the internals of Data-Forge and how lazy evaluation is implemented.

An iterator allows the rows of a dataframe, series or index to be iterated. Iterators allow lazy evaluation (row by row evaluation) of data frames, series and index. This is the same concept as an enumerator in C#, the concept that powers lazy evaluation in LINQ.


A selector is a user-defined function (usually anonymous) that is passed to various Data-Forge functions to process or transform each value in the sequence. Selectors are also used to instruct Data-Forge on which part of the data to work with.

For example say you have a row that looks as follows:

	Column1: "some data",
	Column2: 42,

Here is an example a selector that identifies Column2:

var mySelector = row => {
    return row.Column2;

Selectors are usually applied to each row in the series or dataframe.

Selectors are often also applied to the index (although we ignored this in the previous snippet).

An example of a selector that works with index rather than row:

var mySelector = (row, index) => {
	return index;


A predicate function is similar to a selector, but returns a boolean value (technically you can return any value that can be considered truthy or falsey).

An example predicate function:

var myPredicate = row => {
	return row.Column2 >= 42;	


A comparer method is used to compare to values for equality. It returns true (or truthy) to indicate equality or false (or falsey) to indicate inequality.

An example:

var myComparer = (row1, row2) => {
	return row1.ClientName === row.ClientName; // Row comparison based on client name.


A generator is a function that produces zero or more values to be inserted into a Series or DataFrame.

A generator may take arguments and it can return an array of values or rows:

var myGenerator = function (... appropriate arguments ...) {
	return [
		[ .. generated row 1 .. ],
		[ .. row 2 .. ],
		[ .. row 3 .. ],
		[ .. etc .. ]

Alternatively (to support lazy evaluation) a generator may return a lazily evaluated iterable, that is a function that returns an iterator for a sequence of values or rows:

var myGenerator = function (... appropriate arguments ...) {
	return function () {
		var myIterator = ... some iterator for a sequence of values or rows ...
		return myIterator;

Group / Window

A series where each value in the series is itself another series or dataframe. Think of it as a sequence of groups or batches of data. This concept is used by the multiple Data-Forge functions that create groups and windows, for example groupBy, window and rollingWindow.