Allow passing vectors of dates

[rofinn]

This should be non-breaking as we're just loosening some of the existing restrictions. Most of the code already worked for abstract vectors, but now we hard code step(dates) as Day(1) and throw an ArgumentError if we're passed a different step size. Needed to update a single pre-existing test which didn't allow vectors and switched the step size tests to check the ArgumentError vs MethodError.

Closes #18

Also may satisfy the requirements for #14 and #17. Although, maybe having a separate PR for an ExcludeSelector would be useful?

[codecov]

Codecov Report

Merging #25 (0dc394a) into main (845b79f) will increase coverage by 0.95%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #25      +/-   ##
==========================================
+ Coverage   89.04%   90.00%   +0.95%     
==========================================
  Files           5        5              
  Lines          73       80       +7     
==========================================
+ Hits           65       72       +7     
  Misses          8        8              

Impacted Files	Coverage Δ
src/common.jl	80.64% <ø> (ø)
src/NoneSelector.jl	100.00% <100.00%> (ø)
src/PeriodicSelector.jl	100.00% <100.00%> (ø)
src/RandomSelector.jl	91.30% <100.00%> (+0.82%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[oxinabox]

1. Can you tell me more about why we want this
2. I worry that if not careful we will lose the key invariant feature that wether or not a date is selected should not be a property of the dates offerred but rather of only of the parameters of the selector
3. isn't excluding dates, and selecting a subset of dates easily enough done in post processing via setdiff and intersect?

[rofinn]

Can you tell me more about why we want this?

Mostly just that it seems overly restrictive?

I worry that if not careful we will lose the key invariant feature that whether or not a date is selected should not be a property of the dates offered but rather of only of the parameters of the selector

That feels like an assertion that should be tested for each selector regardless of whether we allow vectors.

Isn't excluding dates, and selecting a subset of dates easily enough done in post processing via setdiff and intersect?

True, but we already pass around a selector as a kind of declarative representation of what dates we want. I think eventually allowing nesting or composition of selectors with an explicit inclusion/exclusion could be a useful way of declaratively specifying the dates considered for a project. I suppose we could choose not to loosen the types here and simply support wrapping directly and have a ExcludeSelector which does that post-processing on the selector outputs?