The idea of implementing sankey diagram using visualization using scripting library like D3 is not very old. The number of expert people in this field is also limited
comparing other fields of visualization. Therefore, while working on this project we found the shortage of information as well as the unavailability of proper and detailed
documentation (still so many unanswered questions while dealing with some exclusive features). So, it could be a topic to work on further more and try to develop its
capacity of accepting exceptional case.
Introducing circular behavior (allowing rotation of nodes) of normal sankey is unique. So far our research says nobody implemented it before (or at least we don't know of). It was not very easy to handle because the defined functions of original sankey library works very closely and they are strongly inter-related. Changing one particular function or one certain part of library brought unbalance or disorientation to other remaining parts. We were able to manage this problem but we still believe that further more study needed to improve the performance.
We did not focus much on data parsing. We assume that user will give data input in our specified format. Exception of this format will bring error in the end. While modern technologies offer so much user friendly behavior, we feel that we somehow restricted users and asking them to use only one target data format. So, efficient data engineering could be one area we may focus.
Another important issue is to introduce several type of curves and nodes in sankey diagram. So far we are using rectangles and Cubic Bézier Curves. It would be interesting to add other type of geometrical shape like polygon or circle etc.. In terms of curves instead of using only Cubic Bézier Curves we can introduce some other curve style which will bring lot of varieties to our circular sankey. The more varieties we can add here the more user satisfaction can be ensured.
Introducing circular behavior (allowing rotation of nodes) of normal sankey is unique. So far our research says nobody implemented it before (or at least we don't know of). It was not very easy to handle because the defined functions of original sankey library works very closely and they are strongly inter-related. Changing one particular function or one certain part of library brought unbalance or disorientation to other remaining parts. We were able to manage this problem but we still believe that further more study needed to improve the performance.
We did not focus much on data parsing. We assume that user will give data input in our specified format. Exception of this format will bring error in the end. While modern technologies offer so much user friendly behavior, we feel that we somehow restricted users and asking them to use only one target data format. So, efficient data engineering could be one area we may focus.
Another important issue is to introduce several type of curves and nodes in sankey diagram. So far we are using rectangles and Cubic Bézier Curves. It would be interesting to add other type of geometrical shape like polygon or circle etc.. In terms of curves instead of using only Cubic Bézier Curves we can introduce some other curve style which will bring lot of varieties to our circular sankey. The more varieties we can add here the more user satisfaction can be ensured.