TODO

- Look into nonexistent time error: 1997-03-30 02:00:00
- Document that nonexistent times will return NaT...

- Add tests for forecasts code

- Enable limiting forecast data to specific parameters!
- Add tests to forecast import
- Unique station ids prefixed with domain, e.g. CDC:00005
- Correspondence table between observation and forecast measurement parameters; convert to consistent units
- Setup ongoing data collection via Azure pipelines (hourly?)

- Start on blogpost about library
- Work on couple of viz (with forecasts & how change over time...)

- Add badges to readme
- Move to cloud (Azure?)
- https://pypi.org/project/azure-storage-blob/
- Process data files -> parquet with partitioning
- Put together exploratory visualizations
    - Blog post about sleeping weather
    - Blog post about heat & duration
    - Blog post about cloudiness (overall vs daytime)

- Migrate resource url search to use link lists, not link crawling...

- Add Metadata import..
- Change "height" field to "elevation"
- Identify and correct timezones! (Starting in 2000 UTC, up til 1999 in CET);
    - For correct parsing, metadata Needs to be extracted

- Add file check step for parquet files; delete corrupted files...

- Log processed files (to avoid reprocessing!) (check update timestamps to avoid special case for 'latest' files)
- Cleanup directory after downloading files...
- Adapt parser for zipped files in hourly+ measurements