Proposal: Migrate from CocoaPods to Swift Package Manager

[hi2gage]

Summary

Proposal to migrate the Home Assistant iOS project from CocoaPods to Swift Package Manager (SPM) for dependency management.

Reason

With CocoaPods moving to maintenance mode, I believe it’s a good time to consider migrating this project to SPM.

Benefits of Migrating to SPM

• Native Integration: SPM is integrated directly into Xcode, reducing the need for third-party tools and simplifying workflows for new developers.
• Simplified Dependency Management: SPM's package resolution and versioning are straightforward and align with Swift's ecosystem.
• Reduced Maintenance: By using SPM, the project can avoid dependency-related issues tied to CocoaPods updates or compatibility.
• Encourages Modularization: SPM makes it easier to move code into LocalPackage that are tracked with the project, incentivizing modular architecture.

Proposed Plan

• Evaluate the current dependencies in the Podfile to ensure compatibility with SPM.
• Replace CocoaPods with SPM in the project setup.
• Update the README and any developer documentation to reflect the changes.
• Test the migration thoroughly to ensure there are no breaking changes for contributors or CI workflows.

Implementation

I am interested in leading this migration and would be happy to volunteer for the implementation. I have experience with Swift Package Manager and have already managed to get the project building with SPM. However, completing the pre-migration tasks is necessary before proceeding with the full transition.

If there is interest in proceeding with this migration, we can begin tackling the smaller pre-migration steps outlined below. These incremental changes will help pave the way for a smooth transition to SPM.

That said, I don’t want to begin this project if there is no interest in migrating away from CocoaPods. I’m looking forward to hearing feedback from the community and maintainers to determine if this is the right direction for the project.

Pre-Migration Plan:

• Fix crash due to race condition in Communicator #1505 needs to be reverted and use non-forked Communicator dependency which includes Package.swift
  • The queue can be passed into the observe function without exposing the store

static func observe(
    queue: DispatchQueue = .communicator, 
    handler: @escaping (Self) -> Void
) -> Observation

• Improv Wi-Fi iOS needs a Package.Swift file
• EMTLoadingIndicator needs a Package.Swift but we can just use the fork from this PR
• CPDAcknowledgements needs to replaced (This could be done in the same PR but probably better to do prior.

Questions for Discussion

• Are there any specific concerns about migrating from CocoaPods to SPM?
• Are all current dependencies compatible with SPM? If not, are there alternatives we could consider?
  • Semi answered in the Pre-Migration Plan
• What is the best way to test and ensure a smooth transition for contributors?
• Any other questions / concerns?

Looking forward to hearing your thoughts and feedback on this proposal!

[bgoncal]

Past context:
#1242

I am curious why do you feel the need for moving to SPM right now, I see the benefits you listed but I also have some mixed experiences with SPM in other projects that are actually way simpler in terms of structure compared to Home Assistant.

[hi2gage]

Thank you for the context. Not use how I missed that searching through the repo.

I am curious why do you feel the need for moving to SPM right now

To be fully transparent, I’ve been looking for a new open-source project to contribute to, and as a Home Assistant user, I thought this project would be a great fit. I’ve generally avoided projects using CocoaPods because of the context switching when using different tools. From my perspective, SPM seems to be the future of dependency management for the Swift ecosystem and it would be a rewarding experience to help move a project that direction (if there is a desire of course).

I see the benefits you listed but I also have some mixed experiences with SPM in other projects that are actually way simpler in terms of structure compared to Home Assistant.

That’s a fair point. My perspective comes from working on projects with different complexity structures that use SPM successfully, though I wasn’t directly involved in setting up those systems. As a result, I didn’t feel some of the pain points firsthand, which may explain why my experiences with SPM have been generally positive. 😬 But I do agree with you it would be a huge change structurally which would affect all aspects of the project including CI which could possibly suffer from a package caching standpoint.

I’m curious whether the issues described in #1242 have been resolved with newer SPM updates and/or if they were compounded by using both CocoaPods and SPM together? As I mentioned I was able to get a rough build that imported Realm, Sodium, etc.

Next Steps:

Is there any interest in moving forward and/or investigating this migration? Should we dive deeper into the issues outlined in #1242, or explore if this transition to SPM is even necessary or wanted at this time? Another thing that will need to investigated is the CI impacts, which could be a deal breaker I suppose. If the migration is something the team is interested in, I'd be happy to collaborate on identifying the smaller tasks involved.

If not, no worries. I bet I could find some less invasive tasks.

[bgoncal]

I can say this isn't a priority right now, currently I am solo iOS dev in Nabu Casa so as you can imagine the priority is to solve major issues and implement core funcionality, so if you want o make POC, I can checkout your branch and evaluate the result, but I won't be able to actively help you with that unfortunately, but thanks a lot for proposing this.