Dev

[TimTaylor]

This PR overhauls the implementation and interface of add_parent_interval() and merge_episodes():

• The merge_episodes() generic has had it's signature tweaked to widen how it can be used. This has been done with the addition of a default method that consequently means the following are equivalent:

with(dat, merge_episodes(id, start, end))
reframe(dat, merge_episodes(id, start, end))
merge_episodes(dat, "id", "start", "end")

• add_parent_interval() has been updated in a similar fashion.

• For both merge_episodes() and add_parent_interval(), the id argument can now be of length greater than one. In essence this is akin to constructing unique identifiers across multiple variables (I'm unsure if this is useful or not at the moment). EDIT: This wasn't fully implemented and think it is best left to the user to ensure in hindsight. Have removed partial implementation.

• BREAKING CHANGE: merge_episodes and add_parent_interval() now require column names to be explicitly stated for the data frame methods. I.e. Where, in the previous release, merge_episodes(dat) would have defaults "id", "start" and "end" for the id, start and end arguments, this would now be written as merge_episodes(dat, "id", start", "end").

• Input checking and subsequent error signalling has been improved.

• In merge_episodes() we now catch errors from calling add_parent_interval() and chain together for a more useful output.

• Return to using the cli package for error signalling/messaging.

• C code for .calculate_parent() has been made more robust.

• BREAKING CHANGE: For both merge_episodes() and add_parent_interval(), users can now control the output column names. These arguments have been given slightly different defaults to what was set automatically in the previous version (hence the breaking change warning).

Tests have been updated accordingly but are still rather lacking (hoping to address this in future commits in this PR).
EDIT: Tests now updated.

[TimTaylor]

@Lextuga007 - this is the sort of thing I've been mulling over for the refactor. Marking as draft for the time being to get your thoughts.

[TimTaylor]

@Lextuga007 - I'm going to close this for the moment but keep working on the same dev branch with a slightly revised implementation that utilises tidyselect/non-standard evaluation (per our slack musings). Will do another PR (probably next week now though).