Migrate to Swift 4's Decodable APIs #84

Palleas · 2017-06-15T01:55:41Z

Add CI target for Swift 4 in travis (cf Add Xcode 9 jobs to CI matrix #88)
Use Swift 4 Codable protocol
Update tests
Remove duplicate Result submodule

Let me know if I forgot something in the list of stuff to do!

Palleas · 2017-06-17T21:32:12Z

Getting there!

Palleas · 2017-06-18T16:52:59Z

I'm running the unit tests now and I have an issue. Basically since we treat id in all the resources as String while the JSON returns Int, the decoding fails. The only solution I've found so far is doing is for Release:

    public init(from decoder: Decoder) throws {
        let container = try decoder.container(keyedBy: CodingKeys.self)
        let id = try container.decode(Int.self, forKey: .id)
        self.id = "\(id)"
        self.tag = try container.decode(String.self, forKey: .tag)
        self.url = try container.decode(URL.self, forKey: .url)
        self.name = try container.decode(String.self, forKey: .tag)
        self.isDraft = try container.decode(Bool.self, forKey: .isDraft)
        self.isPrerelease = try container.decode(Bool.self, forKey: .isPrerelease)
        self.assets = try container.decode([Asset].self, forKey: .assets)
    }

Do you see a different approach we could take with this? Right now the only solution I see is going back to treating id as Int. ¯\(ツ)/¯

mdiep · 2017-06-18T17:35:19Z

I think that's likely the only solution. I don't think the built-in JSON encoding is as flexible.

mdiep · 2017-06-18T18:22:39Z

On second thought: I think we should add a property type.

extension Release {
    public struct ID {
        private let id: Int
        public var string: String {
            return "\(id)"
        }
    }
}

That might make decoding easier?

Palleas · 2017-06-18T18:49:53Z

@mdiep Ha! That's exactly what we were talking about in Slack, Cf: https://gist.github.com/Palleas/645b6c30306603feaab148c9e167a307

Although since it will happen for all the Resources, maybe we need to move it to a higher level?

mdiep · 2017-06-18T18:58:17Z

I think it makes sense to have a per-type ID because it adds additional type safety. You can't accidentally pass an ID for the wrong type of resource to an API.

But if it made sense, we could add a top-level protocol that they all conform to.

mdiep

Really neat to see this with the built-in Swift Decodable stuff! 👏

mdiep · 2017-06-28T15:57:03Z

Sources/Tentacle/Client.swift

-            .filter { _, value in value != nil }
-            .map { name, value in URLQueryItem(name: name, value: "\(value!)") }
+            .filter { arg -> Bool in let (_, value) = arg; return value != nil }
+            .map { arg -> URLQueryItem in let (name, value) = arg; return URLQueryItem(name: name, value: "\(value!)") }


If we move to a newer Swift build, we should be able to revert this change since SE-0110 was reverted.

mdiep · 2017-06-28T15:57:52Z

Sources/Tentacle/Client.swift

-    private func execute<Value>(_ request: Request<Value>, page: UInt?, perPage: UInt?) -> SignalProducer<(Response, Any), Error> {
-        return urlSession
+    private func execute<Value: Decodable>(_ request: Request<Value>, page: UInt?, perPage: UInt?) -> SignalProducer<(Response, Data), Error> {
+        let s = urlSession


Should be able to remove this local variable and go back to returning directly?

mdiep · 2017-06-28T15:59:08Z

Sources/Tentacle/Client.swift

+
+            if httpResponse.statusCode >= 400 && httpResponse.statusCode < 600 {
+                let error: Result<GitHubError, Error> = decode(data).mapError { Error.jsonDecodingError($0) }
+                return .failure(.apiError(httpResponse.statusCode, githubResponse, error.value!))


This ! isn't really safe. We should go back to the previous structure using flatMap.

mdiep · 2017-06-28T15:59:21Z

Sources/Tentacle/Client.swift

+        let s: SignalProducer<(Response, Data), Error> = execute(request, page: nil, perPage: nil)
+
+        return s.attemptMap({ (arg) -> Result<(Response, Resource), Client.Error> in
+            let (response, data) = arg


Same here WRT SE-0110.

mdiep · 2017-06-28T15:59:37Z

Sources/Tentacle/Client.swift

-                    .mapError(Error.jsonDecodingError)
-            }
+    ) -> SignalProducer<(Response, Resource), Error> {
+        let s: SignalProducer<(Response, Data), Error> = execute(request, page: nil, perPage: nil)


Should be able to remove s and just return.

mdiep · 2017-06-28T16:04:46Z

Sources/Tentacle/Homepage.swift

+
+import Foundation
+
+public struct Homepage: Decodable {


Maybe make this an enum instead?

enum Homepage { case url(URL) case string(String) }

mdiep · 2017-06-28T16:05:58Z

Sources/Tentacle/Label.swift

+    private enum CodingKeys: String, CodingKey {
+        case name
+        case color
+    }


Do we need these since they match what would be generated?

mdiep · 2017-06-28T16:07:12Z

Tests/TentacleTests/FileResponseTests.swift

+//            return
+//        }
+//
+//        XCTAssertEqual(expected, decoded)


mdiep · 2017-06-28T16:07:20Z

Tests/TentacleTests/FileTests.swift

+//        ])
+//
+//        let encoded = file.encode()
+//        XCTAssertEqual(expected, encoded)


mdiep · 2017-06-28T16:07:23Z

Tests/TentacleTests/FileTests.swift

+//        ])
+//
+//        let encoded = file.encode()
+//        XCTAssertEqual(expected, encoded)


dcaunt · 2017-06-30T09:12:56Z

.gitmodules

-[submodule "Carthage/Checkouts/Runes"]
-	path = Carthage/Checkouts/Runes
-	url = https://github.com/thoughtbot/Runes.git
+[submodule "Carthage/Checkouts/result"]


Result is now duplicated in this file, and (nit) it should be capitalized Result

Palleas · 2017-07-04T13:07:29Z

Thank you all for the feedback, I'll take care of most of them this week. The rollback of SE-0110 should be in the next Xcode seed right?

mdiep · 2017-07-04T16:40:06Z

The rollback of SE-0110 should be in the next Xcode seed right?

I think so, yes. Otherwise you can probably download a snapshot from swift.org.

Palleas · 2017-07-10T19:46:27Z

Xcode 9b3 reverts SE-110 so I'll tackle all the feedback this week!

mdiep · 2017-08-25T12:21:06Z

Sources/Tentacle/Client.swift

 import Foundation
 import ReactiveSwift
 import Result

+//extension Array where Element == ResourceType : ResourceType {}


mdiep · 2017-08-25T12:21:22Z

Sources/Tentacle/Client.swift

-            .filter { _, value in value != nil }
-            .map { name, value in URLQueryItem(name: name, value: "\(value!)") }
+            .filter { (_, value) -> Bool in value != nil }
+            .map { (name, value) -> URLQueryItem in URLQueryItem(name: name, value: "\(value!)") }


You should be able to remove these return types?

mdiep · 2017-08-25T12:22:30Z

Sources/Tentacle/Client.swift

+                }
+
+                return current.concat(self.execute(request, page: nextPage, perPage: perPage))
+            })


Trailing closure syntax here, please. 🙏

Not sure how I missed it, fixed!

mdiep · 2017-08-25T12:23:03Z

Sources/Tentacle/Client.swift

@@ -272,7 +266,7 @@ extension Client.Error: Hashable {
            return (error1 as NSError) == (error2 as NSError)

        case let (.jsonDecodingError(error1), .jsonDecodingError(error2)):
-            return error1 == error2
+            return error1 == error2 // FIXME


Why is this labeled FIXME?

mdiep · 2017-08-25T12:26:53Z

Sources/Tentacle/Tree.swift

+            case mode
+            case url
+            case size
+        }


This shouldn't be needed since all the cases have the default value?

If think they are because url and size are not exposed as properties and we are using them for encoding, but maybe I'm missing something..?

mdiep · 2017-08-25T12:28:31Z

I'm liking these changes!

If it'd be helpful, we could add ID/Identifiable in a separate PR before this one.

ikesyo · 2017-09-28T14:01:59Z

Sources/Tentacle/Homepage.swift

+
+import Foundation
+
+public enum Homepage: Decodable {


What is the purpose of this type? Keep using URL seems better to me.

Unless the API is expected to respond with an arbitrary String and not a valid URL, I agree.

I understand now that the problem here is empty string handling (""). Empty string is not valid input even if a property is URL?.

In that case it seems like we should revert to URL? and change the decoding?

GitHub will let you set this field to any string.

BUT your github.com profile will only show the string if it's actually a URL. So I think we could revert to a URL and just set it to nil if it's not a valid URL.

Pushed aa43141 to address this.

Oops, missed this, removed my comment 🙊
Are we sure it’s the same with a repository? I remember having this conversation a while ago 🤔

Palleas · 2017-10-07T17:27:12Z

Sources/Tentacle/Tree.swift

-
-        if let base = base {
-            try container.encode(base, forKey: .base)
-        }


mdiep

Looks good overall! Thanks for tackling this! It's no small effort. 🎉

mdiep · 2017-10-10T01:20:46Z

Sources/Tentacle/Client.swift

+            .flatMap(.concat) { response, data -> SignalProducer<(Response, [Resource]), Error> in
+                let current = SignalProducer<(Response, [Resource]), Error>(value: (response, data))
+                guard let _ = response.links["next"] else {
+                    return current.concat(.empty)


.concat(.empty) is effectively a noop. This can just be return current.

mdiep · 2017-10-10T01:21:31Z

Sources/Tentacle/Commit.swift

    /// SHA of the commit
-    public let sha: SHA
+    public let sha: Hash


Why did this change from a SHA?

Because it’s not exactly the same type IIRC

Can you elaborate on this? I think it should be the same thing.

mdiep · 2017-10-10T01:21:46Z

Sources/Tentacle/Commit.swift

        /// URL to see the parent commit in a browser
        public let url: URL

        /// SHA of the parent commit
-        public let sha: SHA
+        public let sha: Hash


Why did this change from a SHA?

mdiep · 2017-10-10T01:23:16Z

Sources/Tentacle/Content.swift

+                        debugDescription: "Invalid content-type \(type)"
+                    )
+                }
+            }


How else would you handle this? 🤔

Made a proposal in 261c768, let me know what you think. At least we're no longer switching over strings.

mdiep · 2017-10-10T01:25:29Z

Sources/Tentacle/JSONExtensions.swift

+extension DecodingError: Equatable {
+    static public func ==(lhs: DecodingError, rhs: DecodingError) -> Bool {
+        switch (lhs, rhs) {
+        default: return false // FIXME


Any idea how to do proper Equatable implementation for DecodingError? 🤔

I have a proposition suggested to me in Slack in f9d214e, let me know what you think!

mdiep · 2017-10-10T01:27:08Z

Sources/Tentacle/Label.swift

+    }
+
+    // TODO: remove this, or replace `Color` with something (De)codable
+    private enum CodingKeys: String, CodingKey {


I'm fine with being explicit. 👍

mdiep · 2017-10-10T01:27:52Z

Sources/Tentacle/Label.swift

+        self.color = color
+    }
+
+    // TODO: remove this, or replace `Color` with something (De)codable


I think this comment can go away?

mdiep · 2017-10-10T01:30:02Z

Sources/Tentacle/RepositoryInfo.swift

+        self.owner = try container.decode(UserInfo.self, forKey: .owner)
+        self.name = try container.decode(String.self, forKey: .name)
+        self.nameWithOwner = try container.decode(String.self, forKey: .nameWithOwner)
+        self.body = try? container.decode(String.self, forKey: .body)


I think this should be try container.decodeIfPresent?

Yeah, it should 😞

mdiep · 2017-10-10T17:12:20Z

Sources/Tentacle/Tree.swift

+        case isTruncated = "truncated"
+    }
+
+    public struct SHA: Codable {


We shouldn't need this one either. A single SHA type should be enough.

Palleas · 2017-10-10T22:45:06Z

I remember why there are 2 SHAs:

On Branch, we have:

Branch
- Commit
  - Sha (String)

But on Tree, we have:

Tree
- Sha (String)

I'll change the way it works for Branch, it feels like a hack.

ikesyo · 2017-10-10T23:45:45Z

Sources/Tentacle/RepositoryInfo.swift

+        self.nameWithOwner = try container.decode(String.self, forKey: .nameWithOwner)
+        self.body = try container.decodeIfPresent(String.self, forKey: .body)
+        self.url = try container.decode(URL.self, forKey: .url)
+        self.homepage = try container.decodeIfPresent(URL.self, forKey: .homepage)


This doesn't work for empty string. That should be why try? was used here.

That doesn't totally match the intent, since a string that's not a URL should be nil, but a number should throw. But it's probably close enough. 🤷‍♂️

Soooo, back to try? is what you meant?

Yeah, just for homepage.

Palleas · 2017-10-11T13:13:36Z

Also fixed the remaining failing test (in Tree encoding)

mdiep

The question of the SHA types is really the only remaining thing.

mdiep · 2017-10-15T01:01:24Z

Sources/Tentacle/Branch.swift

-    /// Sha of the commit the branch points to
-    public let sha: SHA
+    /// The commit the branch points to
+    public let commit: Commit


I'm still confused about why this isn't just a SHA. 🙈

This is what the payload returns:

[ { "name": "master", "commit": { "sha": "6dcb09b5b57875f334f61aebed695e2e4193db5e", "url": "https://api.github.com/repos/octocat/Hello-World/commits/c5b97d5ae6c19d5c5df71a34c7fbeeda2479ccbc" }, "protected": true, "protection_url": "https://api.github.com/repos/octocat/Hello-World/branches/master/protection" } ]

I can retrieve the sha directly with a custom init, is that what you have in mind? Maybe I'm just missing something 🤷‍♂️

Ah, okay! I didn't realize there was another property returned on commit! 😅

mdiep · 2017-10-15T01:01:47Z

Sources/Tentacle/Client.swift

+                            .mapError(Error.jsonDecodingError)
+                            .flatMap { error in
+                                .failure(Error.apiError(response.statusCode, Response(headerFields: headers), error))
+                        }


This } is under indented.

mdiep · 2017-10-15T01:02:42Z

Sources/Tentacle/Commit.swift

    /// SHA of the commit
-    public let sha: SHA
+    public let sha: Hash


Can you elaborate on this? I think it should be the same thing.

mdiep · 2017-10-17T00:54:50Z

Sources/Tentacle/Branch.swift

-    /// Sha of the commit the branch points to
-    public let sha: SHA
+    /// The commit the branch points to
+    public let commit: Commit


Ah, okay! I didn't realize there was another property returned on commit! 😅

mdiep · 2017-10-17T01:02:19Z

This is great! Thank you SO SO SO much for doing this. ✨ It's great to see what Swift's built-in decoding looks like, warts and all.

Sorry about the long review times and my general cluelessness in this PR. Things have been a little crazy here. 😅

Palleas force-pushed the swift-4 branch from a6be31b to c0c74e4 Compare June 16, 2017 17:44

Palleas force-pushed the swift-4 branch from e6ac4b7 to b431236 Compare June 18, 2017 22:10

Palleas requested a review from mdiep June 22, 2017 17:05

mdiep requested changes Jun 28, 2017

View reviewed changes

dcaunt reviewed Jun 30, 2017

View reviewed changes

Palleas force-pushed the swift-4 branch from a8be0cd to a35d0b7 Compare August 18, 2017 15:30

mdiep requested changes Aug 25, 2017

View reviewed changes

Palleas mentioned this pull request Aug 27, 2017

Implement Identifiable type in model with ids #87

Merged

Palleas added 12 commits August 29, 2017 15:53

Migrate to Swift 4 using Xcode

0c32ba7

Remove dependencies to Argo / Curry and Runes

d7148c9

Remove Argo dependencies

f822788

First attempt at refactoring the execute() method

dbfc430

First attempt at execute() with [T]

de06905

Fix execute method

7996c78

Remove Argo from tests

4c3d301

Use Identifiable protocol

ab781aa

Replace all the string with ID<T>

4f0fa14

Fix method called

956281e

Update decoding code

d56cc01

Fix renamed range API

8a3b327

ikesyo added 5 commits September 28, 2017 22:04

Use DecodingError.dataCorruptedError method

58389ad

Use encodeIfPresent

e90b8cb

Remove unnecessary Encodable implementations

f71bf9d

Minor refactoring

1cae102

Update to recommended settings

ce36bf6

ikesyo mentioned this pull request Sep 28, 2017

[RepositoryInfo] Conform to Identifiable #89

Merged

ikesyo reviewed Sep 28, 2017

View reviewed changes

ikesyo and others added 2 commits September 29, 2017 00:24

Merge branch 'master' into swift-4

17fac50

Use URL instead of Homepage type

aa43141

Palleas commented Oct 7, 2017

View reviewed changes

Sources/Tentacle/Tree.swift Outdated

if let base = base {

try container.encode(base, forKey: .base)

}

Copy link

Collaborator Author

Palleas Oct 7, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TIL!

mdiep requested changes Oct 10, 2017

View reviewed changes

mdiep reviewed Oct 10, 2017

View reviewed changes

Palleas added 2 commits October 10, 2017 13:37

Use decodeIfPresent

74933bf

Remove comment

284c6db

Palleas added 4 commits October 10, 2017 18:55

Use only one SHA

3d0997a

Implement Equatable for DecodingError

f9d214e

Add ContentTypeName to switch on enums

261c768

Remove noop concat

c43c3c1

ikesyo reviewed Oct 10, 2017

View reviewed changes

Palleas added 2 commits October 11, 2017 09:08

Use try? for homepage

c1bcdf5

Fix Tree encoding

fe72556

mdiep reviewed Oct 15, 2017

View reviewed changes

Palleas added 2 commits October 15, 2017 11:18

Remove Hash subtype

2967a0d

Fix indent

11b759b

mdiep approved these changes Oct 17, 2017

View reviewed changes

mdiep merged commit 90c5757 into master Oct 17, 2017

mdiep deleted the swift-4 branch October 17, 2017 01:02

Migrate to Swift 4's Decodable APIs #84

Migrate to Swift 4's Decodable APIs #84

Conversation

Palleas commented Jun 15, 2017 • edited Loading

Palleas commented Jun 17, 2017

Palleas commented Jun 18, 2017

mdiep commented Jun 18, 2017

mdiep commented Jun 18, 2017

Palleas commented Jun 18, 2017

mdiep commented Jun 18, 2017

mdiep left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Palleas commented Jul 4, 2017

mdiep commented Jul 4, 2017

Palleas commented Jul 10, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mdiep commented Aug 25, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mdiep left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Palleas commented Oct 10, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Palleas commented Oct 11, 2017

mdiep left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mdiep commented Oct 17, 2017

Palleas commented Jun 15, 2017 •

edited

Loading