Exclude unneeded files from built package

[darnuria]

By default cargo publish all the content versionned by git.
following insight given by cargo diet I used an explicit approach
to exclude test dataset etc. from the final package.

May clause #202

[darnuria]

Seems like the files are included, look like a bug/edge case in the reporting of cargo-diet I fill a bug in the repository of the project because, the line suggested by cargo dietdo not have the same semantics in cargo itself than in the size analisys of cargo diet.

I will push soon something to avoid test folder in each sub-folder or at least in src it seems to work. (Need some test to be sure)

[darnuria]

I think it might be easier to just use exclude instead of include: https://doc.rust-lang.org/cargo/reference/manifest.html#the-exclude-and-include-fields

Just made a commit with explicit exclusion if it's okay, I can rebase it, if it not I can change the Pull-request.

Edit: Random-idea a possibility with exclusion is to just exclude big samples test file, it may need to change some test to need an option if test are desired in the final package.

Quick analysis of the size of the package

I checked the package size by doing:

cargo clean
cargo publish --dry-run
ls -al ./target/package

Results:

# Before sha c240586a31f19d06b4a82a2280e7735832aa0c80
78972 juil. 24 23:46 bson-1.0.0.crate
# After 
43993 juil. 24 23:33 bson-1.0.0.crate

Something like 34979 bytes saved!

Sorry for my English I am a not an English native writer and also thanks for your patience! :)

[darnuria]

There was a effectively a bug in crago-dietabout the interpretation of globing in cargo manifest format: the-lean-crate/cargo-diet#6 seems solved.

[saghm]

Any chance you'd be able to rebase this branch with master? I think that should fix the lint failures, and then this should be review for the rest of the team to review!

[darnuria]

Hi, sure thanks for the follow-up!

I rebased it onto master, but I didn't squashed the two commit to keep information while reviewing of the first approach but if the exclusion one is selected, I will rebase my commits to keep just one.

[saghm]

Sounds good! I agree there's no reason to modify the commits you made until we're ready to merge.

I just authorized the CI run; once it passes, I'll mark this as ready for review and add the others!

[darnuria]

Thanks I took the liberty to make it available for review since the result of the CI test were OK.

[patrickfreed]

LGTM, and thanks for your contribution!

I think something like this would definitely be useful in the driver too, so I filed RUST-526 to track the work for that.

[saghm]

Just merged; thanks for the PR!

[darnuria]

Thanks for time and review! :)