easydata
easydata copied to clipboard
hackalog
A flexible template for doing reproducible data science in Python.
We recently pulled #211 which included reproducer code. We need to document **how to write a test for Easydata**. (this is a bit non-obvious, because to run a test, we...
Cannot fetch files from data sources to user-defined download directory. The attribute `download_dir` of `src.datasets.DataSource` class in NOT passed to `fetch` method as a kword argument. Way to reproduce issue:...
Changed ARCH to use python for platform independence, but this can lead to different lockfile names depending on the version of python (e.g. Darwin vs MacOS)
Add unit tests for notebooks that are part of the documentation (via mkdocs).
If I change my data_path in `config.ini` to another location, it doesn't survive restarting my kernel in my notebook. Reloading the kernel overwrites `config.ini`.
Add a warning message (or something like it) if there is no license.
Dataset needs a notion of a hash. This should be the hash of its metadata dict. (the metadata dict contains the hashes of data, target, and eventually, extra). as is,...
We've now overloaded the config.ini to include both paths[] info, and local dataset configuration. This is fragile in two ways: 1. references to local_config (paths._config) rely on the implementation of...
In a datasource that was created via: * add_datasource * add_manual_download * add_metadata all the paths show up in the datasource_catalog as absolute paths.