boilerdata

Data processing pipeline for a nucleate pool boiling apparatus.

Overview

The data processing approach taken in this repository started over at pdpipewrench. It was initially conceptualized as a way to outfit pdpipe pipelines from configuration files, allowing for Pandas pipeline orchestration with minimal code. I have since adopted a less aggressive tact, where I still separate configuration out into YAML files (constants, file paths, pipeline function arguments, etc.), but pipeline logic is handled in pipeline.py. I have also done away with using pdpipe in this approach, as it doesn't lend itself particularly well to ETL. Besides, my data processing need is not quite the "flavor" of statistical data science type approaches supported by pdpipe.

This new approach maintains the benefits of writing logic in Python, while allowing configuration in files. I am using Pydantic as the interface between my configs and my logic, which allows me to specify allowable values with Enums and other typing constructs. Expressing allowable configurations with Pydantic allows for generation of schema for your config files, raising errors on typos or missing keys, for example. I also specify the "shape" of my input and output data in configs, and validate my dataframes with pandera. Once these components are in place, it is easy to implement new functionality in the pipeline.

Usage

If you would like to adopt this approach to processing your own data, you may clone this repository and begin swapping configs and logic for your own, or use a similar architecture for your data processing. To run a working example with some actual data from this study, perform the following steps:

1. Clone this repository and open it in your terminal or IDE (e.g. git clone https://github.com/blakeNaccarato/boilerdata.git boilerdata).
2. Navigate to the clone directory in a terminal window (e.g. cd boilerdata).
3. Create a Python 3.10 virtual environment (e.g. py -3.10 -m venv .venv on Windows w/ Python 3.10 installed from python.org).
4. Activate the virtual environment (e.g. .venv/scripts/activate on Windows).
5. Run pip install --editable . to install boilerdata package in an editable fashion. This step may take awhile.
6. Delete the top-level data and config directories, then copy the config and data folders inside of tests/data to the root directory.
7. Copy the .propshop folder in tests/data/.propshop to your user-folder (e.g. C:/Users/<you>/.propshop on Windows).
8. Run dvc repro metrics to execute the data process up to that stage.

The data process should run the following stages: axes, modelfun, runs, parse_benchmarsk, pipeline, and metrics. Some stages are skipped because we specified to run just the necessary stages up to metrics (the example data doesn't currently include the literature data). You may inspect the pipeline stages of the same name in src/boilerdata/stages, such as pipeline.py to see the logic that runs during that stage. This example happens to use Python scripts, but you could define a stage in dvc.yaml that instead runs Matlab scripts, or any arbitrary action. This approach allows for the data process to be reliably reproduced over time, and for the process to be easily modified and extended in a collabroative effort.

There are other details of this process, such as the hosting of data in the data folder in a Google Cloud Bucket (alternatively it can be hosted on Google Drive), and more. This has to do with the need to store data (especially large datasets) outside of the repository, and access it in an authenticated fashion.

Project information

• Changes
• Docs
• Contributing

Contributors

Blake Naccarato 💻