Investigating the advantages and disadvantages of available open-source frameworks for prototyping data-based applications
The Data Science team at ESR focus on exploring and developing innovative solutions that can then be passed on to clients or other areas of the organisation to use and maintain. Data Engineers employed within this team create platforms which enable users to easily interact with and interpret data and results. Traditionally these have been developed in Tableau or R Shiny, but more recent projects have been built in Streamlit or React.
The shift from R Shiny to Streamlit was initially made because Streamlit was supported by Snowflake, which the team were moving towards as a data management platform, and applications could be created and deployed rapidly. ESR are now using a High-Performance Computing cluster for data management and model development, so other options beyond Streamlit could be considered. While Streamlit is easy to use, it does have several disadvantages that may mean that another framework could be a better choice for certain projects.
The goal of this research project was to compare Streamlit to potential alternatives. Options such as Plotly Dash and Python Shiny were compared on factors such as the overall look and feel of the application, complexity of the visualisations, optimisation of performance, the use and persistence of input widgets, and the ability to incorporate data modelling. A decision-support tool was developed to aid the Data Engineering team in determining which package to use for their prototype applications. As part of the exploration process, a template application was built in several frameworks for comparison purposes. Future research could use this template to explore further options.
ABOUT THE AUTHOR
Rebekah Au is a Data Science intern at Environmental Science and Research who is currently working on an industry-based student project with the Data Science team. This project is a contribution towards a Master of Applied Data Science from the University of Canterbury. During her internship, Rebekah has developed dashboards and formed strategies to visualise and model time series and geospatial data.
-----
For more information about the eResearch NZ / eRangahau Aotearoa conference, visit:
https://eresearchnz.co.nz/
