Optimally, developers can add a new plot and table to the API by simply adding a plotting function. Based on the plotting function, the API framework can automatically generate required code and data. As each plot is generated from a table, the table generation function is included in the plotting function.
This ticket intends to discuss options to automate API endpoint development. This overall direction is conceptualized based on the discussions with @jharenza and @chinwallaa.
Following is a brief description on the current development procedure:
However, integrating the code for generating example plots and tables is time consuming, because the developers generally need to work on the following steps:
Convert data release files into database tables.
Clean up data.
Handle independent samples, all cohorts, n sample filtering. If a plot needs more than one data files, align them so there is no duplicated or missing links.
Load processed data into database.
Dump database for deployment on remote server.
Write R code to generate plots via API:
Query database. Performance may need to be evaluated by adding additional columns to the data development step.
Process query results. There may be missing or duplicated records.
Plot. The plot code for generating example plots needs to be greatly refactored to take care of all possible cases, e.g., plot width/height, title/legend location, text length. The refactoring step usually takes a lot longer than developing example plots.
Currently, API endpoint development for new plots and tables is a complex procedure. As an example, the development procedure for new DNA methylation plots and tables is briefly described in https://github.com/PediatricOpenTargets/OpenPedCan-api/issues/69 and https://github.com/PediatricOpenTargets/OpenPedCan-api/issues/70.
Optimally, developers can add a new plot and table to the API by simply adding a plotting function. Based on the plotting function, the API framework can automatically generate required code and data. As each plot is generated from a table, the table generation function is included in the plotting function.
This ticket intends to discuss options to automate API endpoint development. This overall direction is conceptualized based on the discussions with @jharenza and @chinwallaa.
Following is a brief description on the current development procedure:
For any new table and plot to be added to API, example plots and tables need to be generated using https://github.com/PediatricOpenTargets/OpenPedCan-analysis data. The code for generating example plots and tables are further integrated into the API. For example, CNV example plot creation is tracked by https://github.com/PediatricOpenTargets/ticket-tracker/issues/349.
However, integrating the code for generating example plots and tables is time consuming, because the developers generally need to work on the following steps:
cc @taylordm @kelseykeith @afarrel @ewafula