@superset-ui/react-pivottable

react-pivottable

react-pivottable is a React-based pivot table library with drag'n'drop functionality. It is a React port of the jQuery-based PivotTable.js by the same author.

react-pivottable is part of Plotly's React Component Suite for building data visualization Web apps and products.

What does it do & where is the demo?

react-pivottable's function is to enable data exploration and analysis by summarizing a data set into table or Plotly.js chart with a true 2-d drag'n'drop UI, very similar to the one found in older versions of Microsoft Excel.

A live demo can be found here.

How can I use it in my project?

Drag'n'drop UI with Table output only

Installation is via NPM and has a peer dependency on React:

npm install --save react-pivottable react react-dom

Basic usage is as follows. Note that PivotTableUI is a "dumb component" that maintains essentially no state of its own.

import React from 'react';
import ReactDOM from 'react-dom';
import PivotTableUI from 'react-pivottable/PivotTableUI';
import 'react-pivottable/pivottable.css';

// see documentation for supported input formats
const data = [
  ['attribute', 'attribute2'],
  ['value1', 'value2'],
];

class App extends React.Component {
  constructor(props) {
    super(props);
    this.state = props;
  }

  render() {
    return (
      <PivotTableUI
        data={data}
        onChange={(s) => this.setState(s)}
        {...this.state}
      />
    );
  }
}

ReactDOM.render(<App />, document.body);

Drag'n'drop UI with Plotly charts as well as Table output

The Plotly react-plotly.js component can be passed in via dependency injection. It has a peer dependency on plotly.js.

Important: If you build your project using webpack, you'll have to follow these instructions in order to successfully bundle plotly.js. See below for how to avoid having to bundle plotly.js.

npm install --save react-pivottable react-plotly.js plotly.js react react-dom

To add the Plotly renderers to your app, you can use the following pattern:

import React from 'react';
import PivotTableUI from 'react-pivottable/PivotTableUI';
import 'react-pivottable/pivottable.css';
import TableRenderers from 'react-pivottable/TableRenderers';
import Plot from 'react-plotly.js';
import createPlotlyRenderers from 'react-pivottable/PlotlyRenderers';

// create Plotly renderers via dependency injection
const PlotlyRenderers = createPlotlyRenderers(Plot);

// see documentation for supported input formats
const data = [
  ['attribute', 'attribute2'],
  ['value1', 'value2'],
];

class App extends React.Component {
  constructor(props) {
    super(props);
    this.state = props;
  }

  render() {
    return (
      <PivotTableUI
        data={data}
        onChange={(s) => this.setState(s)}
        renderers={Object.assign({}, TableRenderers, PlotlyRenderers)}
        {...this.state}
      />
    );
  }
}

ReactDOM.render(<App />, document.body);

With external plotly.js

If you would rather not install and bundle plotly.js but rather get it into your app via something like <script> tag, you can ignore react-plotly.js' peer-dependcy warning and handle the dependency injection like this:

import React from 'react';
import PivotTableUI from 'react-pivottable/PivotTableUI';
import 'react-pivottable/pivottable.css';
import TableRenderers from 'react-pivottable/TableRenderers';
import createPlotlyComponent from 'react-plotly.js/factory';
import createPlotlyRenderers from 'react-pivottable/PlotlyRenderers';

// create Plotly React component via dependency injection
const Plot = createPlotlyComponent(window.Plotly);

// create Plotly renderers via dependency injection
const PlotlyRenderers = createPlotlyRenderers(Plot);

// see documentation for supported input formats
const data = [
  ['attribute', 'attribute2'],
  ['value1', 'value2'],
];

class App extends React.Component {
  constructor(props) {
    super(props);
    this.state = props;
  }

  render() {
    return (
      <PivotTableUI
        data={data}
        onChange={(s) => this.setState(s)}
        renderers={Object.assign({}, TableRenderers, PlotlyRenderers)}
        {...this.state}
      />
    );
  }
}

ReactDOM.render(<App />, document.body);

Properties and layered architecture

• <PivotTableUI {...props} />
  • <PivotTable {...props} />
    • <Renderer {...props} />
      • PivotData(props)

The interactive component provided by react-pivottable is PivotTableUI, but output rendering is delegated to the non-interactive PivotTable component, which accepts a subset of its properties. PivotTable can be invoked directly and is useful for outputting non-interactive saved snapshots of PivotTableUI configurations. PivotTable in turn delegates to a specific renderer component, such as the default TableRenderer, which accepts a subset of the same properties. Finally, most renderers will create non-React PivotData object to handle the actual computations, which also accepts a subset of the same props as the rest of the stack.

Here is a table of the properties accepted by this stack, including an indication of which layer consumes each, from the bottom up:

Layer	Key & Type	Default Value	Description
PivotData	data see below for formats	(none, required)	data to be summarized
PivotData	rows array of strings	[]	attribute names to prepopulate in row area
PivotData	cols array of strings	[]	attribute names to prepopulate in cols area
PivotData	vals array of strings	[]	attribute names used as arguments to aggregator (gets passed to aggregator generating function)
PivotData	aggregators object of functions	aggregators from Utilites	dictionary of generators for aggregation functions in dropdown (see original PivotTable.js documentation)
PivotData	aggregatorName string	first key in aggregators	key to aggregators object specifying the aggregator to use for computations
PivotData	valueFilter object of arrays of strings	{}	object whose keys are attribute names and values are objects of attribute value-boolean pairs which denote records to include or exclude from computation and rendering; used to prepopulate the filter menus that appear on double-click
PivotData	sorters object or function	{}	accessed or called with an attribute name and can return a function which can be used as an argument to array.sort for output purposes. If no function is returned, the default sorting mechanism is a built-in "natural sort" implementation. Useful for sorting attributes like month names, see original PivotTable.js example 1 and original PivotTable.js example 2.
PivotData	rowOrder string	"key_a_to_z"	the order in which row data is provided to the renderer, must be one of "key_a_to_z", "key_z_to_a", "value_a_to_z", "value_z_to_a", ordering by value orders by row total
PivotData	colOrder string	"key_a_to_z"	the order in which column data is provided to the renderer, must be one of "key_a_to_z", "key_z_to_a", "value_a_to_z", "value_z_to_a", ordering by value orders by column total
PivotData	derivedAttributes object of functions	{}	defines derived attributes (see original PivotTable.js documentation)
Renderer	<any>	(none, optional)	Renderers may accept any additional properties
PivotTable	renderers object of functions	TableRenderers	dictionary of renderer components
PivotTable	rendererName string	first key in renderers	key to renderers object specifying the renderer to use
PivotTableUI	onChange function	(none, required)	function called every time anything changes in the UI, with the new value of the properties needed to render the new state. This function must be hooked into a state-management system in order for the "dumb" PivotTableUI component to work.
PivotTableUI	hiddenAttributes array of strings	[]	contains attribute names to omit from the UI
PivotTableUI	hiddenFromAggregators array of strings	[]	contains attribute names to omit from the aggregator arguments dropdowns
PivotTableUI	hiddenFromDragDrop array of strings	[]	contains attribute names to omit from the drag'n'drop portion of the UI
PivotTableUI	menuLimit integer	500	maximum number of values to list in the double-click menu
PivotTableUI	unusedOrientationCutoff integer	85	If the attributes' names' combined length in characters exceeds this value then the unused attributes area will be shown vertically to the left of the UI instead of horizontally above it. 0 therefore means 'always vertical', and Infinity means 'always horizontal'.

Accepted formats for data

Arrays of objects

One object per record, the object's keys are the attribute names.

Note: missing attributes or attributes with a value of null are treated as if the value was the string "null".

const data = [
  {
    attr1: 'value1_attr1',
    attr2: 'value1_attr2',
    //...
  },
  {
    attr1: 'value2_attr1',
    attr2: 'value2_attr2',
    //...
  },
  //...
];

Arrays of arrays

One sub-array per record, the first sub-array contains the attribute names. If subsequent sub-arrays are shorter than the first one, the trailing values are treated as if they contained the string value "null". If subsequent sub-arrays are longer than the first one, excess values are ignored. This format is compatible with the output of CSV parsing libraries like PapaParse.

const data = [
  ['attr1', 'attr2'],
  ['value1_attr1', 'value1_attr2'],
  ['value2_attr1', 'value2_attr2'],
  //...
];

Functions that call back

The function will be called with a callback that takes an object as a parameter.

Note: missing attributes or attributes with a value of null are treated as if the value was the string "null".

const data = function(callback) {
    callback({
        "attr1": "value1_attr1",
        "attr2": "value1_attr2",
        //...
    });
    callback({
        "attr1": "value2_attr1",
        "attr2": "value2_attr2",
        //...
    };
    //...
};