This patch was authored and released by @0xTim.

This ensures that the Application's configuration is updated correctly when the server starts. This allows you to retrieve correct hostname and ports from the Application.

Resolves #2755

This patch was authored and released by @0xTim.

Adds initial support for async/await

This patch was authored and released by @gwynne.

It is now possible to specify a custom Encoder and Decoder to use in place of JSONEncoder and JSONDecoder for storing and retrieving values cached in Redis via the Cache interface.

This patch was authored and released by @gwynne.

DatabaseError's isSyntaxError and isConnectionClosed properties now correctly respect MySQLError.invalidSyntax and MySQLError.closed errors. A unit tests for these behaviors is included.

Also avoids double-running the FluentBenchmark tests (same change as fluent-sqlite-driver#75).

This patch was authored and released by @gwynne.

This solves users of Fluent's MySQL driver getting silently stuck on version 1.1.6 of swift-crypto (the current, source-compatible version is 2.0.3).

It would be far preferable to replace the usage of Insecure.SHA1 in MySQLDialect with something better suited to the purpose and thus drop the Crypto dependency in MySQLKit altogether, but we can't do that without breaking migrations in existing databases.

Due to the change in dependency version requirement, this is semver-minor.

This patch was authored by @Guang1234567 and released by @gwynne.

Hello dev:

Just support android, thanks.

Usage:

#!/usr/bin/env bash

export ANDROID_HOME=$HOME/dev_kit/sdk/android_sdk
export ANDROID_SDK=$ANDROID_HOME
export ANDROID_SDK_ROOT=$ANDROID_HOME
export ANDROID_NDK_HOME=$ANDROID_HOME/ndk/21.4.7075529
export ANDROID_NDK=$ANDROID_NDK_HOME
export NDK_ROOT=$ANDROID_NDK_HOME
export ANDROID_NDK_ROOT=$ANDROID_NDK_HOME
export ANDROID_NDK_PATH=$ANDROID_NDK_HOME
export NDK_TOOLCHAINS=$HOME/dev_kit/sdk/toolchain-wrapper

ALL_IN_ONE_FOLDER=$(dirname $(pwd))

export SWIFT_ANDROID_HOME=$ALL_IN_ONE_FOLDER/swift-android-5.4.2-release-ndk21
export SWIFT_ANDROID_ARCH=aarch64
#export SWIFT_ANDROID_ARCH=armv7
#export SWIFT_ANDROID_ARCH=x86_64
export SWIFT_ANDROID_API=23

cd console-kit

#echo -e "\n\n\nRunning on macOS:\n=======================================\n"
#swift build
#.build/x86_64-apple-macosx/debug/ConsoleKitExample demo --color --frames "⠋,⠙,⠹,⠸,⠼,⠴,⠦,⠧,⠇,⠏"

echo -e "\n\n\nRunning on androidOS:\n=======================================\n"

${SWIFT_ANDROID_HOME}/build-tools/1.9.7-swift5.4/swift-build --configuration debug -Xswiftc -DDEBUG -Xswiftc -g

echo -e "\n\n\nCopy ELF to real android device :\n_______________________________________\n"

adb push .build/aarch64-unknown-linux-android/debug/ConsoleKitExample /data/local/tmp

echo -e "\n\n\nCopy swift runtime SO to real android device :\n_______________________________________\n"

adb push ${SWIFT_ANDROID_HOME}/toolchain/usr/lib/swift/android/aarch64/*.so /data/local/tmp

echo -e "\n\n\nRunning on real android device :\n_______________________________________\n"

adb shell LD_LIBRARY_PATH=/data/local/tmp /data/local/tmp/ConsoleKitExample demo --color --frames "⠋,⠙,⠹,⠸,⠼,⠴,⠦,⠧,⠇,⠏"

cd ..

Adds the none attribute to SameSite cookies (#376). This resolves vapor/vapor#375.

⚠️ This contains an additional case in a public enum which may cause a breaking change if you're switching on it

This patch was authored and released by @gwynne.

Leverages the new dialect flags from vapor/sql-kit#144 to enable full UNION queries according to SQLite's support.

The increased SQLKit version requirement makes this a semver-minor change.

Fix a compilation error when building in Xcode 13

This patch was authored by @BeauNouvelle and released by @0xTim.

This change allows Leaf to be used within iOS environments. Without it we get the following error:

The package product 'Vapor' requires minimum platform version 13.0 for the iOS platform, but this target supports 9.0

This patch was authored and released by @0xTim.

Adds async APIs for JWT calls when interacting with JWKS servers

This patch was authored and released by @gwynne.

An attempt to improve the error message for this unusual case.

Also ensures the error appears in release builds.

This patch was authored and released by @0xTim.

Remove the platform requirements as they're not needed and it allows MultipartKit to be used on more platforms.

Resolves #77

Fixed:

• Session authentication middleware now correctly short-circuits if authentication already succeeded (#68, #69)

Fixed:

• KeyedCache.get(_:) is no longer internal. (#50, #57)

This patch was authored and released by @tanner0101.

Adds support for serializing URLs in templates (#38, fixes #64).

This patch was authored and released by @gwynne.

• Leverage new APIs available in PostgresNIO to improve row decoding performance, making up for some speed regressions and then some, especially for queries with lots of result fields.
• Minor improvements to connection handling, also partially courtesy of improved APIs in PostgresNIO.

Note: No new APIs, but marked semver-minor due to requiring an updated minor release of postgres-nio.

This patch was authored and released by @gwynne.

Instead of duplicating logic implemented by postgres-kit for row decoding - and incorrectly, at that - we now always call through to that logic. Despite the additional indirection through the SQLRow existential, this should yield a minor improvement in performance, as we perform fewer unnecessary checks and retain fewer copies of data.

Note: Depends on vapor/postgres-kit#221 to receive the full benefit of these changes.

This patch was authored by @madsodgaard and released by @0xTim.

Adds option to configure the connection pool timeout (#74)

databases.use(.sqlite(.memory, connectionPoolTimeout: .minutes(1)), as: .sqlite)

This patch was authored by @GNMoseke and released by @0xTim.

This removes the content-length: 0 header being automatically applied to outgoing websocket handshakes.

As per RFC 7230 Section 3.3.2:

A user agent SHOULD NOT send a Content-Length header field when the request message does not contain a payload body and the method semantics do not anticipate such a body.

This was causing issues with certain cloud providers, most notably Google Cloud Run, that would force close the websocket as soon as data was received from a client.

This patch was authored and released by @gwynne.

• Fixes SQLUnionBuilder lacking SQLQueryFetcher conformance.
• Fixes support for SQLite.
• Adds support for INTERSECT [DISTINCT|ALL] and EXCEPT [DISTINCT|ALL] unions.
• Adds documentation for SQLDialect and its subtypes.
• Improves test infrastructure.

SemVer Minor

• Allow unescaped SQL in PostgresQuery in https://github.com/vapor/postgres-nio/pull/258
• Remove connection state machine log in https://github.com/vapor/postgres-nio/pull/266
• Add EventLoop API that uses PostgresQuery in https://github.com/vapor/postgres-nio/pull/265
• Add contains(_:) to PostgresRandomAccessRow in https://github.com/vapor/postgres-nio/pull/270

SemVer Patch

• Use Int(exactly:) instead of restricting to 64 bit platforms in https://github.com/vapor/postgres-nio/pull/267
• Continue to make forward progress when cancelled in https://github.com/vapor/postgres-nio/pull/261

Other Changes

• Fix CI test link in README in https://github.com/vapor/postgres-nio/pull/260
• Update CI to 5.6 release and checkout@v3 in https://github.com/vapor/postgres-nio/pull/269

Full Changelog: https://github.com/vapor/postgres-nio/compare/1.8.0...1.9.0

This patch was authored by @rjhancock and released by @gwynne.

removed due to depreciation warnings issued with latest Vapor.

Removes depreciated parameters file:line: from method signature and calling method.

This patch was authored by @iKenndac and released by @0xTim.

This PR allows client apps to conform to the previously-added AsyncScheduledJob protocol by making it, and the associated adapter function from ScheduledJob, public.

This patch was authored by @Andrewangeta and released by @jdmcd.

Adds full conformance to the RedisClient protocol to use the context's eventLoop for commands vs the default RedisConnectionPools RedisConnection eventLoop for commands.

This patch was authored and released by @0xTim.

Resolves a compiler warning introduced in Swift 5.6

Resolves #99

This patch was authored and released by @gwynne.

This loosens the requirement on the value of a KeyPath passed to Fields.path(for:) to AnyQueryableProperty, which has no associated types or Self requirements, and can thus be treated as an existential (e.g. any AnyQueryableProperty). While existentials are best avoided when possible, in this case, the relaxed definition allows usefully operating on the resolved database field names of an arbitrary set of key paths without the use of variadic generics.

As this is a very confusing and opaque description of the behavior, here is an example of a helper method which was impossible to write before:

extension SchemaBuilder {
    func fields<M: FluentKit.Model>(from _: M.Type = M.self, _ keypaths: KeyPath<M, AnyQueryableProperty>...) -> Self {
        return keypaths.reduce(self) { builder, keypath in
            // N.B.: Despite the name and array type, `Fields.path(for:)` never returns more than one `FieldKey` in practice.
            // Implementing `figureOutDataType(for:)` is left as an exercise for the reader.
            return builder.field(M.path(for: keypath).first!, figureOutDataType(for: keypath))
        }
    }
}

// Usage would be something like:
database.schema(M.schema)
    .fields(\.$id, \.$foo, \.$bar, \$.baz)
    .create()

Previously, such a helper could only have taken a single KeyPath parameter. This is admittedly a contrived example (and by design very far from a complete "automatic migration" implementation), and it's a fairly small convenience to call .fields(a,b,c,d) versus .field(a).field(b).field(c).field(d), but it's nonetheless useful. The generic constraint should never have been so strict in any event. (And field keys should not be wrapped in arrays all the time, but that's a PR for another day.)

Note: While technically the relaxed constraint is a "new" public API, the change is small enough and the difference invisible enough that I've marked this semver-patch; even for me, there's such a thing as being too pedantic about the rules 🙂.

This patch was authored and released by @gwynne.

• "Connection closed" errors are now correctly reported when trying to issue queries on closed connections, instead of throwing various NIO errors.
• Query syntax errors are now explicitly reported.
• Unique constraint violation errors are now more consistently reported.

semver-minor due to adding a new case on MySQLError.

This patch was authored by @wibed and released by @tanner0101.

Allows for a boolean field to be stored as string.

.field("is_true", .string, .required)

Additionally it is now its separate type, together with the other types, instead of referring to an anonymous int for a return, which makes It cleaner.

Before: If the field type of a presumed boolean is set to string. it results in below error msg

typeMismatch(Swift.Bool, Swift.DecodingError.Context(codingPath: [], debugDescription: "Could not initialize Bool from \"0\".", underlyingError: nil))

This patch was authored by @danramteke and released by @0xTim.

Exposes the Curve on ECDSAKeys

This patch was authored by @MFranceschi6 and released by @0xTim.

Make FluentMongoError public so it can be examined by end-users.

Fixes #45