Graphexp: graph explorer with D3.js

Graphexp is a lightweight web interface to explore and display a graph stored in a Gremlin graph database, via the Gremlin server (version 3.2.x, 3.3.x or 3.4.x). Update for version 3.5 is required, see issue #102.

Graphexp is under the Apache 2.0 license.

Versions of Graphexp with the same backend but a nicer UI (using bootstrap) are available here github.com/erandal/graphexp and here github.com/ddmx/graphexp.

Configuration

To use Graph Explorer, you need a Gremlin server running with REST or websocket protocol and a recent web browser to display the visualization. On your web browser, just access the file graphexp.html.

Next step, configure the server settings on the bottom of the page. The default Gremlin server address is localhost:8182. You will have to specify the communication protocol websocket or REST and the gremlin server version. Graphexp is not able to handle secure connections yet and a contribution on this topic would be welcome.

Graphexp works with Amazon Neptune. With this database, set SINGLE_COMMANDS_AND_NO_VARS = true in the file graphConf.js. if you use REST over https you may need to set REST_USE_HTTPS = true as well.

Make sure you choose the correct version of Gremlin on the bottom right corner. Setting a wrong version may lead to unexpected problems such as not displaying the edges.

Additional parameters can be configured inside the file graphConf.js.

Getting started

Installing a Gremlin server

If you have not yet installed a gremlin server, download the last release of the Gremlin server and follow the documentation. In the server folder just run

bin/gremlin-server.sh conf/gremlin-server-rest-modern.yaml

or on windows

bin/gremlin-server.bat conf/gremlin-server-rest-modern.yaml

This default server comes with a small graph database of 6 nodes. The server should start on port 8182. Replace gremlin-server-rest-modern.yaml by gremlin-server-modern.yaml if you want to use websocket.

Alternatively, if you have Docker installed on your machine, you may run a Docker container with an already configured Gremlin server. You can find one on this page. This server has a graph database containing a demo graph: the tree of life, with 35960 nodes and 35959 edges. You can download it and run it using

docker pull bricaud/gremlin-server-with-demo-graph
docker run -p 8182:8182 -it --name gremlin-server-rest bricaud/gremlin-server-with-demo-graph

or for the websocket version:

docker pull bricaud/gremlin-server-with-demo-graph:websocket
docker run -p 8182:8182 -it --name gremlin-server-websocket bricaud/gremlin-server-with-demo-graph:websocket

Running a graphexp Demo with Docker

You may also try out a Graphexp demo on joov's Github repository. It uses Docker compose and can work on Windows.

Graphexp guidelines

To get some first visualization of your graph, you may click on the Search button, without filling any box. Graphexp will then send a query to the graph DB, asking for the first 50 nodes and their edges.

The node and edge properties can be automatically retrieved using the get graph info button. Pushing this button will also display some graph properties on the left side of the page. If it is not the case, check your configuration, it means Graphexp can not query the graphDB. To get the properties, Graphexp should consider all the nodes and edges. This may be overwhelming for the server if the graph is very large. A limit to the 10000 first nodes and edges is set to avoid that. You may change it in graphConf.js with the parameter limit_graphinfo_request.

When a node of the visualization is clicked, it will become 'active' with a circle surrounding it and its information will be displayed on the right side of the page. Moreover, this action will trigger the display of its neighbors. Clicking on an edge will show its properties (without highlighting the edge).

When appearing for the first time the nodes will be positioned following a force layout. Drag and drop can be used to pin them in a particular position. Once dragged the nodes will stay at their position. Drag and drop is allowed only for the nodes on the active layer (most recent layer) with no connection with nodes in other layers. See "Visualization concepts" section for more information on the layers.

Querying the graphDB

In the top bar, you can search the graphDB to display a particular node or group of nodes.

• The box Node label allows to filter nodes with a particular label during the search.
• The box Node property, in combination with the Property value box, allows to find nodes with a particular keyword or value in their properties. The Type of search allows for a perfect (equals) or partial match (Contains). Note that the 'contains' option will only work with Janusgraph.
• The box Traverse by edge acts directly in the interactive visualization. If an edge label is entered in the box, clicking on a node will only display its neighbors connected with that type of edge label.
• The Results limit is here to avoid overwhelming the visualization. It fixes the maximal le number of nodes to display per query.
• If Freeze exploration is ticked, the graph displayed will stay the same even if nodes are clicked on. It is useful when you just need to display the node properties.
• Number of layers is explained below in the "Visualization concept" section.

Note that the input is case-sensitive.

URL query string parameters

• ts specifies TraversalSource in case of multiple different graphs stored in the same database. If unspecified, the default is just g. For Example http://localhost:8183/graphexp.html?ts=gTreeOfLife replaces g by gTreeOfLife in all the gremlin queries (for example g.V() becomes gTreeOfLife.V()).

Editing the graph

There is now the possibility to add/edit the vertices and edges of the graph. A small button was added by sandecho at the bottom Edit Graph. You can modify your graph using Graphexp but you have to update the view to see the result. You can check if the modification has been taken into account by the server in the message window on the top right of the interface.

Visualization concept

The visualization is based on a concept of layers of visualization. The idea is to progress in the graph as in a jungle. The clicked node immediately shows its neighbors, opening new paths for the exploration. If not clicked, a node vanishes little by little as we progress in the exploration. Coming back during the exploration is allowed. Before it completely disappears, a node can be clicked and will become active again. This visualization concept is aimed at providing a precise, local view rather than a global one.

During your exploration, you can set up milestones by clicking on the small circle on the upper right side of a node. This will pin the node in time, preventing it from disappearing.

You may also freeze the exploration, by ticking the appropriate checkbox. The evolution of the exploration will stop, allowing to gather information on the nodes displayed, without displaying their neighbors.

Node and edge information

The Id and label of each node can be displayed by hovering the cursor over the node. The full information on the properties is displayed on the right of the page when clicking on the node or edges. Once the get graph info button has been clicked, a choice of properties to display appears on the left side.

Node color

If a node property called 'color' exists in the node properties with a hexadecimal color code (string), it will be displayed automatically on the graph. Otherwise, the default node color can be set in the graphConf.js file. The node color can be set interactively after the get graph info button has been pressed. A select tab appears on the left sidebar allowing to set the color according to one of the property values present in the graph.

Predefined node positions

Graphexp can display nodes at specific positions if they are stored in the DB. For that, modify node_position_x and node_position_y in graphConf.js (by default graphexpx and graphexpy), to whatever keys are refering to the node positions in the graphDB. Values must be numbers. According to Sim Bamford, reasonable values should be below 500 to stay within the page limits. Node with predefined positions are not subject to the force layout and will stay at the same position, while the others may move. It may be useful for plotting a hierarchical graph for example.

Curved edges

GraphExp has now curved links and can display multiple edges between 2 nodes, thanks to a contribution from agussman. This is the default, you can still come back to straight edges by setting use_curved_edges = false in graphConf.js.

Program description

The program uses:

• the D3.js library to visualize a graph in an interactive manner, API Reference,
• an ajax request (with Jquery) that query the graph database (Gremlin Tinkerpop via REST).

Contributing

Contribution as pull requests are very welcome. If you want to contribute, you may have a look at the issues. You can also submit a pull request with a new feature. When contributing, keep in mind that graphexp must stays simple. The idea is to have a simple tool for a quick (and efficient) graph exploration.

Tutorial with the tree of life

Once your gremlin server is up and running (from the Docker repository), click on the get graph info button. Information should appear on the left side of the page, like on the following image.

This graph has a single type of nodes (label 'vertex') and a single type of edges (label 'edge'). Each node is a species (taxon) living on earth or extinct, and directed edges represent the link ancestor-descendant. The different node properties are displayed on the left.

• CHILDCOUNT the number of descendent nodes
• name the name of the species
• HASPAGE whether there is a page of information on the Tree Of Life Project website
• ID Tree of Life Project unique id
• CONFIDENCE confidence in classification, from confident (0) to less confident (1) and (2)
• EXTINCT whether the node is extinct (2) or not (0)
• LEAF the node is a leaf of the tree (1) or the node does not represent a leaf (it has or will have descendent nodes on the Tree of Life) (0)
• PHYLESIS (0) monophyletic, (1) uncertain, (2) not monophyletic

On the top navigation bar, choose the field name, enter 'Dinosauria' as value in the input and click on the Search button. Do not forget the capital letter, as the search is case-sensitive. A single node, corresponding to the Dinosaurs clade, should appear in the middle of the page. Click on the node to display node details on the right as well as its ancestors and descendants on the graph. Check the box name on the left bar to display the node names. You should see appearing the two subgroups of dinosaurs Saurischia and Ornithischia, as in the Wikipedia dinosaur page and an additional none node which is the ancestor. This latter node is a taxon that has ancestors and descendants but does not have a name. Note that there are different versions of the tree of life and it is always evolving as researchers find new species. You may now enjoy the exploration of the dinosaur order by clicking on nodes and following ascendant and descendant lines. The oldest nodes will vanish as you explore the data and if you want more nodes to be displayed, just increase the number of layers on the top navigation bar.

You may also color the nodes according to the values of some of their properties by clicking on the color tab on the left side. The color scale is computed using the range of values of the nodes already displayed and a palette of 20 colors. You should refresh the color after a few steps of exploration.

During the exploration of the Dinosauria clade you may find the bird class Aves. They are the only survivors of the Dinosaur group and descendant of dinosaurs with feathers. To see it, enter Aves in the value field, press search and climb up the tree.

If you want to explore the world of insects, you may start with the taxon Insecta and follow the links. Did you know that spiders are not insects but have they own group Arachnida? Can you tell what is the common ancestor between spiders and insects?

You may also be interested in the Homo group.

Have a try on the live demo of Graphexp on the project Github page.