Fix to the CI tests I submitted

[SynapticSage]

"random" but no "ground truth" tests failed on edge cases.

When the sought element is smaller than all search library numbers, method 2 yields a 0, whereas searchsortedprevious yields a 1 (first element in the library). Likewise, for the next method, if it's largest than all elements, searchsortedfirst will give a number that is one larger than the total set length(a)+1.

Forgot to place bounds on the second method when the test element is either larger than all elements in the search list a or smaller than all elements a.

[codecov]

Codecov Report

Merging #5 (d3dcbc7) into main (baa4310) will increase coverage by 9.09%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##             main        #5      +/-   ##
===========================================
+ Coverage   90.90%   100.00%   +9.09%     
===========================================
  Files           1         1              
  Lines          22        24       +2     
===========================================
+ Hits           20        24       +4     
+ Misses          2         0       -2     

Impacted Files	Coverage Δ
src/SearchSortedNearest.jl	100.00% <100.00%> (+9.09%)	⬆️

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

[SynapticSage]

Also on a related note, there is a possible choice here on edge cases that I hadn't considered.

When the search number is smaller than everything, Matlab's interp1(...,'prev') shoots out a NaN. Same when the search is larger than everything, Matlab's interp1(..., 'next') shoots out a NaN if you're number is bigger than the search set. It's sort of telling the user that "no element satisfies the requested nearest but higher or nearest but lower situation".

I wonder if it would make more sense to output an empty [] in these scenarios. If the user asks for a nearest but higher number, and there is none, or the user asks for a nearest but lower number, and there is none, [] would index their array a[ [] ] and show them an empty set. In a sense saying "there is no element satisfying this request".

[joshday]

I wonder if it would make more sense to output an empty [] in these scenarios.

I think the searchsorted functions in base return nothing for cases like these and we should follow suit here.

[SynapticSage]

That's reasonable. I updated the outputs to nothing for edge cases.

[joshday]

Cool, thanks!