Code Generation for Use-case specific code

[krbundy]

Create code generation for the use-case specific code. This could be a generator for the functions that converts class to integers and back. This could also generate other use-case specific files, provide format validation for inputs, generate templates, etc.

There is already some asset generation (e.g. taxonomic conversion JSON file is automatically generated) but additional code generation and documentation would significantly reduce the time to start a new use case.

[nelsopet]

To do this, the data standard for labels and spectra so that it was clean. Rules in the schema for the different data types, including categorical, continuous and ordinal factors. Couple this approach with validation in single set of functionality to save time on computing (eg. not having to convert validation images into dataframes mulltiple times).

[krbundy]

Based on our meeting regarding this, there are some takeaways:

• Generate a file like Functions/pfts.R that is used for converting between plant functional types.
• Can be generated from a CSV/spreadsheet of plant functional types, or lists of factors.
• Should also add support for continuous response. This includes updating datatypes for saving raster data (remove hard-coded datatypes by adding a categorical/continuous switch)
• Development can happen in either location

[nelsopet]

This set of additional functionality may need it's own config.json to control settings. The steps that would be included in all these additional functionality would be crop and extract pixels from images, parsing the labeled pixels and then writing them to disk as .csv and/or spectral libraries. Then the use-case specific conversion functions do/would use that speclib as input.