Specials thanks to Caleb Ruth who encouraged me to hold an un-planned un-session on this subject at dotFMP un-conference in Berlin 2018. This consequently brought some great tips from the community and made this writing happen much faster then it probably would.
The d3-hierarchy is a module that implements several popular techniques for visualizing hierarchical data, such as Node-link diagrams, Adjacency diagrams and Enclosure diagrams. It facilitates rapid multi-scale inference: micro-observations of individual elements and macro-observations of large groups.
The d3.tree layout implements the Reingold-Tilford algorithm for efficient, tidy arrangement of layered nodes. The depth of nodes is computed by distance from the root, leading to a ragged appearance. Radial orientations are also supported. Implementation based on work by Jeff Heer and Jason Davies using Buchheim et al.’s linear-time variant of the Reingold-Tilford algorithm. Data shows the Flare class hierarchy, also courtesy Jeff Heer.
Related to this article:
Github url to the files used as a starting point in this article: https://bl.ocks.org/mbostock/4339083
Download the example FileMaker file here: JSONFlare.fmp12 (please note that the file can only be useful with FileMaker Pro 16 Advanced and higher)
Credentials for this file are: Username: admin; Password: admin
Wikipedia link: https://en.wikipedia.org/wiki/JSON
I have been introduced to the “Flare graph” (the graphical representation of data using flare class hierarchy…) by Davide Puppo and i liked it straight away. It is probably one of the better ways to show the data and it also comes with the flare – it looks great. In dotFMP 2018. un-conference in Berlin i presented the part of what we did to make it possible to implement this type of visualization in our favorite FileMaker Pro platform. As the session was not planned, the notice was short and the previous night was long, my audience might not have got the complete picture of what it is or could be. I hope this article will fix that fairly.
If we study the files from the Github link provided above, we will see that the flare hierarchy graph produced is dependent on the “d3.v3.min.js” library and also on the text file sitting in the same root called “flare.json” which holds all of the data in a JSON formatted chunk of text.
We where able to define two major challenges standing in the way of implementing this graph into FileMakers WebViewer:
1 ) Exporting the data from FileMaker into the “flare.json” file would not be the best option (i could accept a challenge on this conclusion). First of all, we would have to transfer the data to some distant location only to bring it back in, secondly, we need to have the file exported in utf-8 format which would require a plugin or a work around that might be to lengthy and ugly for the purpose, and finally, it does not work well when the file is called from FileMakers WebViewer.
2) Creating a hierarchy JSON code from FileMaker with the embedded child records down the infinite line of possible relationships did not look like an easy doable task, on the contrary.
Once we resolve these 2 issues we thought it would be an easy path forward to put it all together and make it work. And we where right.
In the provided example file, we can see this switch inside the script “Set WebViewer” on line 19 where the global variable $$HTML is built. If FileMaker calculation dialog had line numbers you would see that the global variable $$JSON is called on line 46. The global variable $$JSON is created in the script step on line 15 of the same script. $$JSON is just reading the calculation field “Product::_ini_JSON_UCL” where the actual magic of creating the JSON lies.
Once we got that in place we could call JSON from the text file or a global variable and use it as an JS Object to show the graph. Fine.
How do we now build the JSON document we need from the actual FileMaker record set?
Challenge #2 – building a JSON document to represent the array of child records down the infinite line
In the provided example file, in table “Product” we can find two calculation fields, both resulting in a text output:
NOTE: These calculations are using JSON functions added in FileMaker Pro 16 to build the JSON text we need to represent the data.
$$JSON could probably also be generated with any of the plugins that we use to generate JSON in Filemaker as the speed on higher numbers of records could become an issue, in theory. The plugins might help as we know some of them could have up to 20 times and more faster performance when generating JSON compared to the native functions. I have not tested the speed thoroughly. On the amount of records in the testing file provided (1822) on MacBook Pro 2017 with macOS High Sierra version 10.13.4 it loads the $$JSON almost instantly and further shows the graph instantly.
This was only a relatively simple example of the way to implement this type of visualization into FileMaker. From here it could take all kind of directions. I am looking into the best way to apply the calculation that builds JSON from FileMaker data on the relationship graph consisting of various tables, for example. Other interesting thing to try would be creating dynamic picture nodes instead of just little circles as presented here. Making nodes that link to some data in or out of FileMaker would be a great feature to.
Presenting the graph in the web browser will have to require a completely different approach, of course. In that case we might need to consider exporting the JSON outcome into the actual “flare.json” text file into a place where the library script can fetch it, and we might want to adjust the actual library to get it from the designated location, if we would not want to keep it in the same folder.
In this example the web viewer output is defined by a global variable. This is done by purpose for the sake of easy testing and controlling the output, and to facilitate further research. You might want to investigate into the possibilities of giving the portion of data to each user according to the privileges.
I hope the article can be helpful as a first step into the matter and a clear proof that it is possible to implement this kind of graph in our good old FileMaker. From here you can really take it where ever you want.