redkite

Kaio Motawara - Redkite data engineer assignment

• Instead of doing a code review, I've taken the liberty to refactor and restructure the project to reflect the approach I would have taken. The changes from and bug fixes in the provided script effectively are the review.
• To run the pipeline, please create the environment specified in the environment.yml' file and run main.py, which orchestrates the pipeline.
• There are folders for schemas and processing functions. I've used separate .py files to store the three schemas. I've used a similar structure for processing functions. While it might be overkill for this problem, it's meant to represent a real-world situation where I might have to deal with a large number of complex schemas and processing functions.
• While I have created the prescribed schemas for the parquet files, when I try to load the files with those schemas, I get an error that I haven't been able to resolve. So I've gone ahead and loaded it without enforcing the schema just so that I could continue with the rest of the assignment.
• the final step of the pipeline provided includes .coalesce() and writes to a databricks .csv. Since the number of partitions is already 1 and since I don't have access to databricks, I would have replaced the step with a simple save to .csv. However, when I tried to write the dataframe to csv, Spark threw a Java error that I couldn't resolve. Therefore, in order to complete the assignment, I converted the Spark dataframe to a Pandas dataframe and then saved it as a tab-separated file as required.
• I've created one unittest in the tests folder just to demonstrate how tests might be written. To run the test, please run python -m unittest discover tests from the root directory of the project (ie the same directory that contains main.py).