Make registering same DataDep twice a no-op

[briochemc]

Trying to fix issue #86 (not sure what I am doing at this point)

[codecov-io]

Codecov Report

Merging #88 into master will increase coverage by 0.4%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master      #88     +/-   ##
=========================================
+ Coverage    53.5%   53.91%   +0.4%     
=========================================
  Files          12       12             
  Lines         228      230      +2     
=========================================
+ Hits          122      124      +2     
  Misses        106      106

Impacted Files	Coverage Δ
src/registration.jl	60% <100%> (+10%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c94424...f43cb0f. Read the comment docs.

[oxinabox]

Suggested change

	return registry
	return registry[name]

Might be a cleaner way to do this, using say:

existing = get(regisitry, name, nothing)
if existing != nothing
     if isequal(datadep, existing)
          return existing
    else
          @warning...
     end
end

or something like that?

[oxinabox]

Cool! thanks. I know digging into a new code-base can be tough and I appreciate it.

Needs tests.
Take a look at the Test.logger stuff for checking what warning messages are, or are not printed.

I think isequal (and thus hash) will need to be defined for a DataDep to make this work, but I am not sure.
Tests will prove it

[zerefwayne]

@oxinabox Is any work remaining on this PR? I need this functionality for my new package.

[oxinabox]

Something is wrong if you need this for your package.
Can you give more info?

[zerefwayne]

@oxinabox

// Returns the URL of dataset
function download(country) {
 register(Datadep(country, ...))
...
}

function get(country) {
resource_path = download(country)

...further processing
}

There's a single get() function in my package which may be called multiple times by a user during runtime. Which leads to register being called multiple times. There's no way to get the path of the dataset or checking the existence if I don't register it once.

If register is not a very heavy process, I don't think doing it multiple times during runtime should be a problem.

UPDATE: I've solved the problem by catching the KeyError thrown by @datadep_str.

[oxinabox]

recommendation is that register is called from the __init__ of your module.
It registers it into a global dictionary so that datadep"abc" works.
Basically so one can pretend it is a filepath string, which is how it acts.

If you have a function you call that return the DataDep object, you actually don't need to register it.
You can resolve it directly, to get a file path (which may mean it will also do a download if not already done)

DataDeps.jl/src/resolution.jl

Line 27 in c3a522d

function resolve(datadep::AbstractDataDep, inner_filepath, calling_filepath)::String

Though in many packages i have that could do that, I register in the __init__ anyway.
Its just the way normally it is worked with.
Something to rethink maybe if we ever did a big overhall of the API.