[swift-build-dev] [swift-evolution] [SwiftPM] Proposal: Add support for test-only dependencies

[Ankit Aggarwal]

- swift-evolution, + swift-build-dev

It might be worth discussing this feature separately from product proposal because IMO the test-only (or local) dependencies should not be considered in dependency resolution. With product proposal we will *not* compile the targets which are not vended by a package but we will still clone and resolve all the dependencies. This means two packages can not depend on conflicting versions of Quick (which might not make sense for local dependencies).


> On 28-Dec-2016, at 5:31 AM, Xiaodi Wu via swift-evolution <swift-evolution at swift.org> wrote:
> 
> The need for a feature such as this was pointed out some time ago. A proposal to allow dependencies to be declared with increased granularity has already been written, reviewed, and accepted! Here it is:
> https://github.com/apple/swift-evolution/blob/master/proposals/0146-package-manager-product-definitions.md <https://github.com/apple/swift-evolution/blob/master/proposals/0146-package-manager-product-definitions.md>
> 
> According to the authors, the proposal was meant to be a more general solution that enables specifying test dependencies as well as other use cases:
> https://lists.swift.org/pipermail/swift-build-dev/Week-of-Mon-20161114/000735.html <https://lists.swift.org/pipermail/swift-build-dev/Week-of-Mon-20161114/000735.html>
> 
> The proposal came up for review, received no feedback, and was accepted as-is.
> 
> 
> On Tue, Dec 27, 2016 at 5:41 PM, thislooksfun via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> wrote:
> Does anyone else have any thoughts on this, or should I go ahead an assemble a PR?
> 
> -thislooksfun (tlf)
> 
>> On Dec 26, 2016, at 1:56 PM, thislooksfun <thislooksfun at repbot.org <mailto:thislooksfun at repbot.org>> wrote:
>> 
>> (I think this is the right place for this suggestion, but please let me know if I'm mistaken)
>> 
>> There is currently no supported way to have some dependencies only used for testing (`swift test`), and the workarounds for it, while workable, are unnecessary crude.
>> 
>> Currently, what a lot of projects are doing is defining a `.Package.test.swift` file that is then copied over `Package.swift` when tests are to be run (on external CI, for example). While this works, it adds a extra step and another point of failure (if you forget to add a new dependency to both `Package` files, for example.
>> 
>> What I propose is a new section of `Package.swift`, labelled either `testDependencies` or `devDependencies`, and the modules referenced within are only loaded/compiled when running `swift test`
>> 
>> // Package.swift
>> import PackageDescription
>> let package = Package(
>>     name: "Project",
>>     targets: [
>>       Target(name: "BoxioDebug", dependencies: ["Core"]),
>>     ],
>>     dependencies: [
>> 	//Same as before
>>     ],
>>     testDependencies: [
>>         .Package(url: "https://github.com/FooBar/PackageOnlyNeededForTesting.git <https://github.com/FooBar/PackageOnlyNeededForTesting.git>", majorVersion: 1),
>>     ]
>> )
>> 
>> This solves having to manually switch out Package files or `.Package` statements inside one Package file when trying to run tests, and keeps all dependencies neatly organized in one file.
>> 
>> This change is purely additive and optional, no existing code needs to change for this to be added.
>> 
>> A real world example is a dependency on Quick (or another such testing framework), that itself uses XCTest, and thus crashes when being run outside of `swift test` (add Quick to Package.swift, run `swift build` and run the compiled binary). With the new `test[dev]Dependencies`, running `swift build` wouldn't even see Quick, and thus would compile and run correctly, but `swift test` would both see and use Quick for running the test suite.
>> 
>> 
>> Other solutions considered:
>> Another potential solution is to only compile the dependencies that were actually used when running `swift build`, since that seems to be the cause of the above problem.
>> 
>> 
>> -thislooksfun (tlf)
>> 
> 
> 
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org <mailto:swift-evolution at swift.org>
> https://lists.swift.org/mailman/listinfo/swift-evolution <https://lists.swift.org/mailman/listinfo/swift-evolution>
> 
> 
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-build-dev/attachments/20161229/0be9de08/attachment.html>