last code update: 20.07.2022
last overall update: 19.03.2024
SanKEY.js is a small JavaScript data visualization "package" that provides a simple object-oriented API to create interactive column-ordered Sankey diagrams. It's free of dependencies and written in vanilla JavaScript. SanKEY.js was designed with the goal of maximum performance, so it works faster than any other library or package used for the purpose of creating Sankey diagrams. It allows a high level of customization with features like special link shapes, adjustable nodes width and integrated callback functions. The possibility of easily creating plots with thousands of columns and tens of thousands of nodes and links redefines the capabilities of Sankey diagrams and opens new horizons for their usage.
SanKEY.js is provided as an ES6 module. To use it, download "SanKEY_script.js" and include the line below in your main script file.
import {PlotCreator} from './SanKEY_script.js'Remember to add type=module declaration while including your script in an HTML document.
SanKEY.js also requires "SanKEY_styles.css" to work properly. This CSS file can be modified to change some basic styles. Read more about it in the documentation section. The full HTML declaration should look more or less like this:
<script type="module" src="./your_main_script.js"></script>
<link rel="stylesheet" href='./SanKEY_styles.css'>SanKEY.js offers possibilities like no other library or package.
Code of this example:
import {PlotCreator} from './SanKEY_script.js'
let nodes = [
[{label:'A'}, {label:'C'}],
[{label:'A'}],
[{label:'Any label', width: 0.8, color:'#ff00ff'},{label:'G'}],
[{label:'T'}],
[{label:'A', color: '#6dd9fa'}, {label:'G'}],
[],[{label: 'A', width: 2 }],[],[],
]
let links =[
{from:{column: 0, node: 1}, to:{column: 1, node: 0}, value:1, link_type: 'LL'},
{from:{column: 0, node: 0}, to:{column: 1, node: 0}, value:1, link_type: 'LR'},
{from:{column: 0, node: 0}, to:{column: 1, node: 0}, value:1, link_type: 'LR'},
{from:{column: 0, node: 1}, to:{column: 1, node: 0}, value:2, link_type: 'LR'},
{from:{column: 1, node: 0}, to:{column: 2, node: 0}, value:2, link_type: 'RL'},
{from:{column: 1, node: 0}, to:{column: 2, node: 1}, value:1, link_type: 'RR'},
{from:{column: 2, node: 0}, to:{column: 3, node: 0}, value:1},
{from:{column: 2, node: 0}, to:{column: 4, node: 1}, value:1, link_type: "LR"},
{from:{column: 3, node: 0}, to:{column: 4, node: 0}, value:1.8},
{from:{column: 3, node: 0}, to:{column: 4, node: 1}, value:1, link_type: 'LR', color: 'black'},
{from:{column: 3, node: 0}, to:{column: 4, node: 1}, value:0.5},
{from:{column: 4, node: 1}, to:{column: 6, node: 0}, value:0.5},
{from:{column: 4, node: 0}, to:{column: 4, node: 0}, value:1},
]
new PlotCreator(
document.body, nodes, links, 1000, 400, 0, 5,
{
plot_background_color: '#f4edf7',
default_links_opacity: 0.8,
default_gradient_links_opacity: 0.8,
label_colors_object: {A: "#3680f7", C: "#39db64", T: '#eb4034', G: '#ebc950', 'Any label': 'magenta'},
lines_style_object: {stroke:'black','stroke-opacity':0.2},
column_names_style_object: { color:'black', opacity:0.6,},
}
)To generate a plot, simply create an instance of the PlotCreator class.
Interactions with the plot, such as changing the plot range, can be achieved by using methods of the class. Each time a method is called, the plot is refreshed, that is, the old plot element on the page is removed and a new one is generated. If the nodes were moved, it is remembered, and refreshing the plot won't change their positions in relation to the columns.
Most of the PlotCreator properties are not private (yet) and theoretically could be modified. However, in most cases it's not recommended. It is advised to use proper methods in order to change them.
Nodes and links data are kept as references to reduce memory usage. Thus, modifying them will impact the PlotCreator object. Adding or removing columns/nodes/links is strongly discouraged, however, changing individual properties such as colors or labels can be beneficial. In such a case, to observe a change, the plot needs to be refreshed.
It's important to mention that in some cases links_data will be slightly modified by PlotCreator. More about it below.
The nodes must have a fixed order in the columns. They will be created according to this order, from top to bottom.
Main class for generating and managing the Sankey diagrams. It requires 7 arguments:
- dom_container - HTMLelement - DOM element inside which the plot will be created.
- nodes_data - (see description) - Nodes data.
- links_data - (see description) - Links data.
- plot_width - number - Width of the plot in pixels.
- plot_height - number - Height of the plot in pixels.
- first_column - number - Index of the first column to be shown.
- last_column - number - Index of the first column that won't be shown.
- settings (Optional) - Object - Object with additional plot settings.
DOM element in which the plot will be created. Creating a plot inside it may override some of its default styles. In particular, it will set the overflow to 'auto'.
A structure that describes the nodes to be created. It should have a form of an array with nested arrays, where each array denotes a new column. The order of the nodes inside the columns is important, because to define a link, the user must specify it for both nodes that the link connects (see links_data).
Arrays describing the columns should be empty or contain objects representing the nodes.
The user can define any properties in node objects. PlotCreator will recognize: label, color and width.
Properties defined here will have a higher priority than those defined as default for the entire plot.
- width (Optional) - a positive floating-point number that indicates what fraction of the column width will the node occupy. Values >1 mean that the node will span horizontally beyond the column.
- color (Optional) - any CSS color. Color of the node.
- label (Optional) - any string. It will be displayed when the user hovers a node.
Example:
[
[{label:'A'}, {}],
[{label:'A'}],
[{label:'Any label name', width:0.8, color:'#ff00ff'},{label:'B'}],
[{label:'T'}],
[{label:'A', color:'#6dd9fa'}, {label:'B'}],
[],
[{label:'A', width:2}],
[],
[]
]It's a data structure that describes the links. It should be an array (with no fixed order needed), in which each object represents one link. Note: the array will be sorted by the PlotCreator. Each link should contain information about the nodes it connects, by specifying the column indexes and positions of the nodes. In addition, it's necessary to specify the value of the link, which will later be converted to its height accordingly. The user can define any properties in link objects. PlotCreator will recognize: link_type, color and sublinks.
Properties defined here will have a higher priority than those defined as defaults for the entire plot.
- from - object with column and position properties defined. It describes the coordinates of the node from which the link goes. Position is the node's index in the column array.
- to - object with column and position properties defined. It Describes the coordinates of the node to which the link goes. Position is the node's index in the column array.
- value - number, value of the link.
- link_type (Optional) - string, one of ['RL','RR','LR','LL']. Describes the shape of the link. If not specified, 'RL' is assumed. The first letter describes on which side of the first node the link will be created, and the second - on which side of the second one. Connections between nodes can be specified from further columns to the previous ones. In such cases, due to the internal architecture, link_type will be overwritten and the reversed=true property will be added.
- color (Optional) - any CSS color. Color of the link.
- sublinks (Optional) - an array of sublink objects. Sublinks are separate links that are included in the area of the link. The shape of the sublink is inherited from its parent link. Objects describing sublinks should have defined value and shift properties. Value is interpreted in the same way as in normal links. Shift is a positive number and defines by what fraction of the parent's link height is the sublink shifted downwards, in relation to the parent's link upper edge. To simplify, shift=0.5 means that the top edge of the sublink will be exactly in the middle of the parent link's height. In addition, each sublink can have its own unique color defined. Unlike normal links, sublinks are parsed every time the plot is refreshed, so they can be used, for example, to dynamically highlight paths in the diagram. Note: nested sublinks won't be parsed.
Example:
[
{from:{column: 0, node: 1}, to:{column: 1, node: 0}, value:1, link_type: 'LL'},
{from:{column: 0, node: 0}, to:{column: 1, node: 0}, value:1, link_type: 'LR'},
{from:{column: 0, node: 0}, to:{column: 1, node: 0}, value:1, link_type: 'LR'},
{from:{column: 0, node: 1}, to:{column: 1, node: 0}, value:2, link_type: 'LR'},
{from:{column: 1, node: 0}, to:{column: 2, node: 0}, value:2, link_type: 'RL'},
{from:{column: 1, node: 0}, to:{column: 2, node: 1}, value:1, link_type: 'RR'},
{from:{column: 2, node: 0}, to:{column: 3, node: 0}, value:1},
{from:{column: 2, node: 0}, to:{column: 4, node: 1}, value:1, link_type: 'LR'},
{from:{column: 3, node: 0}, to:{column: 4, node: 0}, value:1.8, sublinks: [{shift:0.1, value: 1.2, color:'magenta'}]},
{from:{column: 3, node: 0}, to:{column: 4, node: 1}, value:1, link_type: 'LR', color: 'black'},
{from:{column: 3, node: 0}, to:{column: 4, node: 1}, value:0.5},
{from:{column: 4, node: 1}, to:{column: 6, node: 0}, value:0.5},
]Number, width of the plot in pixels.
Number, height of the plot in pixels.
Number, index of the first column to be shown when the plot is initially generated.
Number, index of the first column that won't be shown when the plot is initially generated.
As an eighth argument, the user can additionally pass an object with settings, that will impact plot's styling and usability. All settings are optional. PlotCreator will recognize the following properties:
Number between 0 and 1. Minimal size of the total vertical gap between nodes in a column (as a height fraction). E.g. 0.6 means that in each column at least 60% of its height will be the spacing between nodes. The default value is 0.7.
Number between 0 and 1. Standard node width as a fraction of the column width. E.g. 0.6 means that each node will occupy 60% of the column width. The default value is 0.3.
Boolean, determines whether to show the lines between columns. The default value is true.
Boolean, determines whether to show the column names. The default value is true.
Boolean, determines whether to show links in a situation where both nodes are not in the range of the plot, but the link between them is. If set to true, each time the plot is being refreshed, PlotCreator will look for such links, by scanning k nearest columns on one side of the column range, where k is the length of the longest link in the graph and length being defined as |column_of_the_first_node - column_of_the_second_node|. This may negatively impact the performance, however, unless there are links with the length around several thousand columns, the slowdown should not be noticeable. The default value is true.
Boolean, determines whether the nodes can be moved vertically. The default value is true.
Boolean, determines whether the links should have a gradient color. The default value is true.
Any CSS color. Plot background color. The default value is '#f5ecec'.
Any CSS color. Default nodes color. The default value is 'grey'.
Any CSS color. Default links color. The default value is 'blue'.
Number between 0 and 1. Default links opacity. The default value is 0.25.
Number between 0 and 1. Default opacity of links if those have a gradient color. The default value is 0.43.
Any CSS color. Default sublinks color. The default value is 'red'.
Number between 0 and 1. Default sublinks opacity. The default value is 0.8.
An object in which node labels are properties and node colors are the values. By defining such an object, the user can easily link node labels to their color, thus avoiding code redundancy. The default value is {}.
An array of column names. If passed, then instead of column indexes, names from this array will be displayed. The default value is undefined.
Number, determines from which number columns are being indexed. Important: this will only impact the display and PlotCreator’s interface. While defining nodes and links data, the user still needs to use the standard array indexes. The default value is 0.
Number, determines from which number nodes in the columns are being indexed. Important: this will only impact the display and PlotCreator’s interface. While defining nodes and links data, the user still needs to use the standard array indexes. The default value is 0.
Number, minimal curvature of the arc when the links have a curved shape, for example when a link connects the left sides of both nodes. The default value is 5.
Number, determines by how many pixels should increase the arc of the curved links when these are stacked. The default value is 5.
An object that defines styles of the lines between the columns. PlotCreator will recognize all properties that are also defined in the default object and only overwrite those defined by the user. The default values are: {stroke: 'red', 'stroke-width': 2, 'stroke-opacity': 0.35, 'stroke-dasharray': '5,5'}.
An object that defines styles of the column names. PlotCreator will recognize all properties that are also defined in the default object and only overwrite those defined by the user. The default values are: {'font-size':'15px', color:'red', opacity:0.50,'font-weight':'bold'}.
String, appearance of the cursor when a node is being hovered. The default value is 'pointer'.
String, appearance of the cursor when a link is being hovered. The default value is 'help'.
String, appearance of the cursor when a node is being grabbed. The default value is 'grabbing'.
Function that will be executed after a node is clicked. PlotCreator will try to pass four arguments to it: node_info, node_data_reference, node_element and event.
- node_info - an object with the information about a clicked node.
- label - node's label
- height - node's height
- number_of_links - total number of links in a node
- number_of_links_left - number of links on the left side of the node
- number_of_links_right - number of links on the right side of the node
- left_side_sum - sum of the values of the links on the left side of the node
- right_side_sum - sum of the values of the links on the right side of the node
- column - index of the node's column
- position - index of the node in a column
- node_data_reference - a reference to a defined data object about the chosen node
- node_element - node's DOM SVG rectangle element
- event - event object
The default value is undefined.
Function that will be executed after a link is clicked. PlotCreator will try to pass four arguments to it: link_info, link_data_reference, link_element and event.
- link_info - an object with information about the clicked link.
- from_column - name/index of the column, from which the link goes
- to_column - name/index of the column, to which the link goes
- from_node - index of the node from which the link goes
- to_node - index of the node to which the link goes
- from_label - label of the node from which the link goes
- to_label - label of the node to which the link goes
- value - value of the link
- link_data_reference - a reference to a defined data object about the chosen llink
- link_element - link's DOM SVG path element
- event - event object
The default value is undefined.
Function that will be executed after a node is hovered. PlotCreator will try to pass four arguments to it: node_info, node_data_reference, node_element and event, same as in on_node_click_function. Important: the return value of this function will be set as info bubble's inner HTML. The default value is:
(node_info,node_data_reference,node_element,event)=>{return `${node_info['label']}`}Function that will be executed after a link is hovered. PlotCreator will try to pass four arguments to it: link_info, link_data_reference, link_element and event, same as in on_link_click_function. Important: the return value of this function will be set as info bubble's inner HTML. The default value is:
(link_info,link_data_reference,link_element,event)=>{
return `From: (column - ${link_info['from_column']}, node - ${link_info['from_node']}, label - "${link_info['from_label']}") To: (column - ${link_info['to_column']}, node - ${link_info['to_node']}, label - "${link_info['to_label']}") Value: ${link_info['value']}`
}After the plot is initially generated, the methods allow interacting with it to, for example, change some of the default properties.
Changes current column range. Refreshes the plot. First_column and last_column are column indexes. *Note, that if the default start_column_count_from was changed, then the indexes here must be shifted accordingly.
Allows changing whether column names are being displayed. Refreshes the plot.
Allows changing whether lines between columns are being displayed. Refreshes the plot.
Allows changing whether nodes can be moved vertically. Refreshes the plot.
Allows changing whether links have a gradient color. Refreshes the plot.
Method that should be used when one wants to remove the plot from DOM and from the memory. Note, that all references pointing to that object should also be removed.
Plot refresh. Useful when the plot data is manually modified.
Additional function. Sets multiple attributes to a DOM element. Can be imported.
- dom_element - HTMLelement - Target DOM Element.
- arg_val_object - Object - Object with the attributes to be set.
As aforementioned, SanKEY.js requires SanKEY_styles.css to work properly. The file is short and contains classes that allow modifying the following things:
- on hover appearance and animations of nodes and links
- appearance and animations of info bubbles
- mobile functionality
We recommend customizing it to one's needs.
A simple benchmark scripts can be found in the "benchmarks" directory.
Results can be compared with the Plotly library. To run the tests, set up a simple local server, for example, with Node.js live-server pacakge, and open the HTML with the tests that you want to run.
Be aware, that the tests can be very time-consuming. It's strongly suggested that you adjust the number of test runs by modifying testData variables in JS files, especially for the configurations with a large number of nodes and links.
Since the tests are run in the browser, many things can influence the final result: hardware, resources assigned to your browser, processes running in the background, garbage collection cycles, and so on. There's no guarantee that the tests will yield the same results for different machines, or even the same machine if run more than once.
The tests may put a big pressure on the browser, due to intense resource exploitation. It's recommended that you run the tests for different test data in isolation. It's highly likely, that the browser will crash during the tests.
The results presented below were run on: Chrome 118, Ubuntu 22.04, 12th Gen Intel® Core™ i7-1265U × 12, Mesa Intel® Graphics (ADL GT2).
| Number of nodes | Number of links | Columns (for SanKEY.js) | Mean generation time, Plotly (ms) | Mean generation time, SanKEY.js (Initial, reload) (ms) |
|---|---|---|---|---|
| 10 | 40 | 5 | 6.33 | 0.48, 0.35 |
| 100 | 400 | 20 | 56.78 | 3.38, 2.81 |
| 500 | 2000 | 100 | 1129.32 | 17.82, 14.89 |
| 1000 | 4000 | 200 | 7012.73 | 38.04, 33.00 |
| 5000 | 10000 | 1000 | 59 483.17 | 102.66, 93.49 |
| 10000 | 20000 | 2000 | 222 179.5 | 233.26, 187.63 |
The results show up to 1000 times better performance in comparison to Plotly. The difference grows with the size of the dataset.
- Work on this project is (technically) still in progress. Please contact me if you find any bugs or mistakes.
- It is forbidden to distribute the internal code as your own in any external repositories. Appropriate mention or citation is required.
- Ideally I'd like to release a NPM package and add TypeScript type definitions, but it's not of any priority for me at the moment. The code is quite old, so it would require a thorough refactoring. Maybe if the project sparks some interest...
The project was created and is maintained by Krzysztof Zdąbłasz.
I would like to thank Zuzanna Czerska for her help with not coding-related works.
e-mail: krzysiek.zdablasz@gmail.com