[traits.build workflow] Time data types get reformatted with read_csv

[yangsophieee]

Time data types get reformatted with read_csv, such that e.g. "9:00:00" becomes "09:00:00". This is not an issue if it's consistent with the output of metadata_add_contexts and process.R. However, right now metadata_add_contexts recognises "09:00:00" but then it gets written to the metadata as "32400.0".

[yangsophieee]

Note: You can't read in all data as character columns during read_csv because then process_custom_code won't work.

I decided to convert context values to character type before they are converted to yaml, which seems to have fixed the problem!

[yangsophieee]

There is a little bit of an issue where someone might manually add a time context value to the metadata in its original format, i.e. "9:00:00". Then they would find that the context value isn't being detected in the data (because the data has been read in as "09:00:00". There is a dataset_test that says when context values are not in the metadata so this would clear things up, and I added a little note here to notify when a time data type has been detected and reformatted.

[ehwenk]

Nice solution. Does the dataset_test now specify which values data are missing from metadata vs which values metadata values are extra? If so, I think that is good enough, because it will immediately indicate what to change.

[yangsophieee]

@ehwenk Not yet, that is still an issue to be fixed (#63) but I will get to it soon!

[yangsophieee]

It seems that David has been converting time variables to character type within custom_R_code, otherwise the dataset_test test about time data types gets flagged. Needs further investigation, but it seems "11:30" in the data file may not be properly written by metadata_add_context to the metadata file.