Add include_lazy_artifacts flag to create_app.

[Sacha0]

This pull request reconstitutes @rbvermaa's stab at #464 in https://github.com/RelationalAI-oss/PackageCompiler.jl/tree/rv-include-lazy. Rob notes

I tried to add an include_lazy_artifacts option [1], however, it is probably not a correct solution, as it ignores the situation where there is a Dict in the Artifact section. See first part of the if at https://github.com/RelationalAI-oss/PackageCompiler.jl/blob/rv-include-lazy/src/PackageCompiler.jl#L722 .

which I haven't looked into yet. But putting this pull request up to get the ball rolling :).

Co-Authored-By: Rob Vermaas [email protected]

[codecov-io]

Codecov Report

Merging #489 (d5229b1) into master (e408e64) will increase coverage by 0.01%.
The diff coverage is 60.00%.

@@            Coverage Diff             @@
##           master     #489      +/-   ##
==========================================
+ Coverage   82.65%   82.66%   +0.01%     
==========================================
  Files           2        2              
  Lines         392      398       +6     
==========================================
+ Hits          324      329       +5     
- Misses         68       69       +1     

Impacted Files	Coverage Δ
src/PackageCompiler.jl	81.74% <60.00%> (-0.31%)	⬇️
src/juliaconfig.jl	90.47% <0.00%> (+2.67%)	⬆️

---

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 e408e64...d5229b1. Read the comment docs.

[Sacha0]

Complete solution or no, CI seems happy! 😄

[Sacha0]

Friendly bump. Thoughts? :)

[fguevaravas]

I think this is a great idea and will test this PR in my usage case and report back in a few days.

I have issues with apps generated with PackageCompiler.jl getting huge because every single artifact that might be remotely needed is downloaded. To give an example: compiling an app that uses FFTW.jl, downloads the whole Intel MKL and OpenMP even if the app never ends up using these (we're talking about 1.5GB of things that don't need to be packaged with app). Ditto for all the timezone data in TimeZone.jl

I am concerned about one thing though, what is the behavior if there is an artifact that is really needed by the app and it wasn't downloaded at compile time? Will the app download it and crash if it can't? If that is the case, it may be better to have a more granular control? Perhaps providing a list of artifacts to include/exclude?

[jlperla]

Sorry to bump this, but is this PR safe to use (or is there a good reason it hasn't been merged)? We have been bitten by the JuliaTime/TimeZones.jl#325 issue.

[Sacha0]

We have been using this patch continuously for several months now, for what that's worth :). IIRC this pull request is waiting on formation of consensus on how best to move forward (with this or another approach to addressing the underlying issue).

[KristofferC]

Sorry, yes, we should clearly get this in.

[KristofferC]

#529 should hopefully do it.