Future

• Refactor to reduce file IO: Functions should take and return dataframes. CLI commands should still operate via CSVs/files. This has benefits such as easier to test, chain multiple commands w/o intermediate IO (possible speed improvement), etc.
• Logging for visibility: Want to know where cytotools is slow and can be improved. Logging here will promote visibility into slow operations. Can "drill down" with logging (e.g., into cytominer) to get finer grained view.
• Refactor to improve usability: Consider whether or not the fixed file system is working for all use cases. Consider CLI flags or a configuration file to set values.
• Integration tests: Create a larger end-to-end test to run nightly/on changes on, e.g., external platform like Jenkins. Talk to CMAP for inspiration/advice.
• Provenance for reproducibility of analysis: Consider adding tags or logging to create a record of the analysis. CMAP team has some experience and thoughts on this that would be useful.