Change scan so that compiler specialisation will be active

[SamuelBrand1]

Minor fix of scan to address #82 . Also removed docstring for scan because being actively developed in response to #35 .

Closes #82

[seabbs]

Could you give a bit more on why this is a fix and what the impact on the fix is (the latter only if its not that much trouble). It looks a bit subjective from what I have seen so far.

[seabbs]

Also I assume you checked that it wasn't already specialising?

[seabbs]

This looks good to me and reasoning makes sense as to the change. A little more justification would be perfect in this PR but as is is enough for me to sign off.

[SamuelBrand1]

In light of discussion here and on #82 I changed scan implementation to take a where F <: AbstractEpiModel argument.

The reason for this change is to

1. avoid potential problems with specialisation on generic functions (flagged by @seabbs in Activate specialisation for functions with function input #82).
2. Create safety for the code; the only context we want EpiAware.scan used is as part of the latent infection dynamics.
   • The behaviour of latent infs (e.g. model structure) is now determined by the type of EpiModel.
   • The fixed parameters (e.g. GI) are fields of an immutable EpiModel object.
   • Optionally, a callable/functor for the EpiModel can be defined which defines the dynamics and is used with EpiAware.scan.

Having said the above simple performance testing shows no change between doing this and scan(f::F,...) where {F} or even scan(f,...).

Therefore, point 1. should be considered fairly subjective/based on general principles at this point and point 2. is more persuasive.

[seabbs]

Create safety for the code; the only context we want

This is really nice by the way and I agree its persuasive

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.74%. Comparing base (8b9edf9) to head (d676612).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #83   +/-   ##
=======================================
  Coverage   97.74%   97.74%           
=======================================
  Files           6        6           
  Lines         133      133           
=======================================
  Hits          130      130           
  Misses          3        3           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[seabbs]

Having said the above simple performance testing

Its interesting it should make a difference but in practice doesn't really. I wonder if its due to julia docs being a little out of step with improvements that have been made or just this case. Shows I guess that its hard to reason on.

[SamuelBrand1]

Having said the above simple performance testing

Its interesting it should make a difference but in practice doesn't really. I wonder if its due to julia docs being a little out of step with improvements that have been made or just this case. Shows I guess that its hard to reason on.

Yeah, at some point we should do a bit of deeper dive.