8000 GitHub - quran/mobile-sync: Quran Mobile Sync Engine · GitHub
[go: up one dir, main page]

Skip to content

About

Quran Mobile Sync Engine

Resources

Stars

2 stars

Watchers

4 watching

Forks

Repository files navigation

mobile-sync

Kotlin Multiplatform sync/data stack for Quran mobile apps.

This repository contains shared modules used by Android and iOS apps for:

  • Authentication (OIDC)
  • Local persistence (SQLDelight)
  • Sync engine orchestration
  • A unified managed data API (QuranDataService)

Table of Contents

Architecture

auth + persistence + sync-v2 are composed in sync-pipelines, exposed through a managed DI graph (AppGraph / SharedDependencyGraph), and exported to iOS through umbrella. The repositories write plain domain tables (no sync columns); the sync engine diffs them against the last state both sides agreed on (sync_item) to find what to push.

flowchart LR
  UI[Android/iOS UI] --> G[AppGraph]
  G --> S[QuranDataService]
  G --> F[SyncAuthService]
  F --> A[Auth Module]
  S --> A
  S --> P[Persistence Module]
  S --> E[sync-v2 engine]
  P --> E
  U[Umbrella XCFramework] --> UI
Loading

Sync

  • Engine (:sync-v2). Each run fetches every change since the stored token, merges it three ways with the local changes (an edit beats a delete), and pushes what is left in batches of at most 100. An item the server keeps refusing is held, so the rest still syncs. SYNC_ENGINE.md maps how.
  • Persistence on plain tables. The repositories write plain domain tables, owned by :sync-v2's types, through one store per type, so the app and the importer follow the same rules. The tables carry no sync columns; the engine keeps its own state in sync_item and sync_meta.
  • Wiring (:sync-pipelines). SyncClient runs the engine for the signed-in account and reports each run; SyncScheduler decides when runs happen. Sync requests time out after 60 seconds without data, which catches a server that stops answering; a request as a whole has no cap, so a large first-sync page on a slow network can finish. Connecting gives up after 15 seconds on Android and the JVM, while on iOS (Ktor's Darwin engine ignores the connect timeout) the 60-second idle timeout covers connecting. Android now notices a stalled connection after 60 seconds instead of OkHttp's default 10; a timeout fails the attempt as a network error (SyncErrorKind.NETWORK), which is retried.
  • When it syncs. A run starts 30 seconds after app start (at once when the saved session is restored), 5 seconds after a local change, 2 seconds after a return to the foreground (syncOnForeground(); none within a minute of a sync that pulled the server's changes, unless one failed since), and at once for triggerSync(immediately = true) and a sign-in. A token refresh starts no run. After every run, success or not, the next one is due in 30 minutes, so sync never stops. Runs never overlap. A failed run is retried with exponential backoff: up to 4 retries after about 2, 6, 18 and 54 seconds (2 s, ×3), each moved at random by up to half of itself either way so devices that failed together don't retry together, and at most 60 seconds. A 403 that may be the token's scope pauses pushes for an hour, or until the access token is refreshed or a user signs in; automatic runs keep pulling, and triggerSync(immediately = true) pushes anyway.
  • Auth. The sync account is the ID token's sub. Signing in to a different account wipes the previous account's local data before anything is pushed. A 401 forces one token refresh and runs again; a refused refresh ends the session, while a network error, a timeout or a server error during a refresh keeps it, as does a refresh cancelled with its caller.
  • The user's profile. authState's user is the ID token's identity with the account's profile on top. The ID token repeats the claims captured when the user consented, on every refresh, so a name or photo changed on Quran.com later comes only from GET /v1/users/profile on the sync host, with the sync headers. It is fetched during the sign-in's sync, during the first sync that starts a day or more after the last fetch (an hour or more after a failed attempt; no timer of its own, so nothing while offline), at once during a manual sync (triggerSync(immediately = true); the sync never waits for it), and when the app calls QuranDataService.refreshProfile(), such as when the user closes a Quran.com account page; opening screens fetches nothing. A call during a fetch for the same session joins it. The profile is stored apart from the ID token's claims (AuthStorage), so a token refresh doesn't bring back the old name, and it is cleared with the session on logout and on an account switch. It is stored only for the session's account (sub) and only while that session lasts; a profile for another account is dropped without ending the session. A 401 forces one token refresh and tries once more; any failure keeps the stored profile and the session, and never blocks sign-in or sync. authState republishes only when the user changed. The photo is a social login's URL as it is, or for one uploaded on Quran.com, the medium (256 px) avatar on SynchronizationEnvironment.avatarsHost, built as the users backend's avatarsBuilder builds it. Each AppEnvironment has its own avatar host: production's qreflect-bucket and prelive's qreflect-bucket-prelive S3 buckets.
  • Status. QuranDataService.syncStatus shows Idle, Syncing while an attempt runs, Retrying while a failed run waits for its next attempt (when it failed, its SyncError, and when it tries again; triggerSync(immediately = true) runs at once instead), or how the latest run ended: Succeeded with its time, or Failed with its time and a SyncError once the retries run out. A run that only pulls during a push backoff keeps the failure that started it.
  • History. QuranDataService.syncHistory lists the newest 100 runs, newest first, one SyncRun per run (retries included): when it started and ended, its trigger (APP_START, LOCAL_CHANGE, FOREGROUND for syncOnForeground(), PERIODIC, MANUAL for triggerSync(immediately = true), SIGN_IN), its result (SUCCEEDED, PULLED_ONLY during a push backoff, FAILED with its SyncError, CANCELLED), its attempts, and how many items it pushed, pulled and left held. A run that found no session before it reached the server, or that a logout or reset stopped, is not recorded; one the session's end stops after it reached the server (a refused token refresh, or a retry that finds no one signed in) is recorded as CANCELLED with what it did. It holds no user data; it lives in the database (sync_run) and is cleared with the rest of the data on logout and on an account switch.
  • Held items. QuranDataService.heldItems lists the items that don't sync, as HeldSyncItems: the item's type and public id, a HeldReason (TOO_SHORT, TOO_LONG, INVALID_VERSE, RESERVED_NAME, REJECTED_BY_SERVER with the HTTP status and error code, ...), and for an item waiting on a held one (WAITING_ON_PARENT, WAITING_ON_OTHER_ITEM), that item. It updates as soon as an item is held or released: an edit that fixes a held item takes it out at once, without a sync.
  • Unsynced items. QuranDataService.unsyncedItems counts the items whose local state the server doesn't have yet, from one read, as UnsyncedItemCounts: pending, new, edited or deleted since they last synced, which the next sync sends; and held, the heldItems (held items and the items waiting on them), which don't sync until they are fixed. Logout loses them all (it wipes all local data), so an app can warn with these counts. An item counts once, in one of the two; a change the server never sees doesn't count (an edit only to a field that isn't synced, a verse removed with its collection, a delete the server would refuse, which sync undoes). It may over-count pending items before the first sync (items a sync will link to the server's equal ones count as new) and while a sent batch's answer is unknown. Like heldItems, it updates on every local write, without a sync; each write re-reads it, push planning included, so collect it only while it is shown.

Modules

Module Purpose
:auth OIDC login/logout, auth state, token handling
:persistence The app database (MobileSyncDatabase) and the repositories for bookmarks/collections/notes/reading sessions, over :sync-v2's plain tables
:sync-v2 The generic sync engine (fetch, three-way merge, push) and the plain domain tables of its types; see SYNC_ENGINE.md
:sync-pipelines DI graph, the sync client and scheduler, and the QuranDataService managed data API
:umbrella iOS XCFramework export (Shared.xcframework)
:demo:android Android sample app (Compose)
:demo:common Shared demo helpers/models
:mutations-definitions Shared helpers: the server's text rules (ServerText) and the DI scope (AppScope)
:logging Routes logs, including HTTP calls, to a host app's logger (MobileSyncLogging)

Requirements

  • JDK 17
  • Android Studio (for Android demo)
  • Xcode (for iOS demo)
  • Cocoa toolchain for iOS simulator/device builds

Optional but recommended:

  • local.properties with the OAuth client ID for the Android demo app:
OAUTH_CLIENT_ID=your_client_id

Published library artifacts do not embed credentials. Android apps are OAuth public clients using PKCE and must not embed OAuth client secrets.

Quick Start

  1. Clone and enter the repo:
git clone https://github.com/quran/mobile-sync.git
cd mobile-sync
  1. Ensure Gradle can resolve dependencies:
./gradlew help
  1. Run tests:
./gradlew allTests --stacktrace --continue

Usage

Integration modes

Choose one integration mode per database.

Persistence-only mode uses the public :persistence repositories directly. Repositories preserve local integrity and mutation tracking, but consumers must not mutate SQL tables directly. Use this mode only when the database is not managed by :sync-pipelines.

Managed sync-capable mode uses :sync-pipelines through AppGraph. In this mode, all user-visible data reads and writes go through QuranDataService, and authentication goes through SyncAuthService. Do not mix direct persistence writes with the managed sync graph. If sync may be enabled later for the same app data, use managed mode from day one, even while logged out. Logged-out managed-service writes are local-first; sync triggers no-op until the user is authenticated.

AppGraph exposes only managed facades: quranDataService: QuranDataService and authService: SyncAuthService. Raw write-capable persistence repositories are no longer exposed from :sync-pipelines AppGraph; they remain public in :persistence for persistence-only consumers. The raw :auth AuthService is hidden from AppGraph and wrapped by SyncAuthService.

This repository is still in active development. Schema changes are made directly in .sq files; .sqm migrations are not currently maintained.

1. Initialize managed app graph (Android/Kotlin)

Configure the Android app manifest placeholders for the login and post-logout redirect URIs. The :auth artifact contributes com.quran.shared.auth.android.SafeOidcRedirectActivity as the exported browser callback entry point and keeps the upstream OIDC HandleRedirectActivity non-exported.

android {
    defaultConfig {
        manifestPlaceholders["oidcRedirectScheme"] = "com.quran.oauth"
        manifestPlaceholders["oidcRedirectHost"] = "callback"
        manifestPlaceholders["oidcPostLogoutRedirectScheme"] = "com.quran.oauth"
        manifestPlaceholders["oidcPostLogoutRedirectHost"] = "callback"
    }
}

Do not declare or export org.publicvalue.multiplatform.oidc.appsupport.HandleRedirectActivity in the app manifest. If you override AuthConfig.redirectUri or AuthConfig.postLogoutRedirectUri on Android, keep the corresponding manifest placeholders aligned with those URIs.

Graph construction does not require an Android Activity. Build the graph from application context and runtime configuration. Optionally route Mobile Sync's logs to the app's logger with MobileSyncLogging.configure(sink, minSeverity); it must come before SharedDependencyGraph.init, because HTTP clients read the severity when they are created. At DEBUG, HTTP calls also log their headers and bodies, which hold user content such as note text. configure replaces Kermit's process-wide logger configuration, so if the app or another library also logs with Kermit, those logs reach the sink too. Without configure, logs go to the platform log, including a summary line for each HTTP request and response.

import com.quran.shared.logging.MobileSyncLogSeverity
import com.quran.shared.logging.MobileSyncLogging
import com.quran.shared.persistence.DriverFactory
import com.quran.shared.pipeline.AppEnvironment
import com.quran.shared.pipeline.di.SharedDependencyGraph
import com.quran.shared.pipeline.storage.createMobileSyncStorage

// Optional.
MobileSyncLogging.configure(
    sink = { severity, tag, message -> appLog(severity, tag, message) },
    minSeverity = MobileSyncLogSeverity.INFO
)

val graph = SharedDependencyGraph.init(
    driverFactory = DriverFactory(context = applicationContext),
    storage = createMobileSyncStorage(applicationContext),
    appEnvironment = AppEnvironment.PRELIVE,
    clientId = appClientId
)

val authService = graph.authService
val quranDataService = graph.quranDataService

If clientId is blank, the same initializer creates a managed local-only graph. QuranDataService continues to read and write local data, while SyncAuthService.isAuthenticationConfigured is false and sign-in fails with an authentication-not-configured error. This lets open-source builds run without bundled OAuth credentials.

Register the Android browser auth flow only before interactive sign-in. The registered Activity handles browser launch and redirect delivery; it is not needed for graph construction or local data operations.

import com.quran.shared.auth.di.AuthFlowFactoryProvider
import org.publicvalue.multiplatform.oidc.appsupport.AndroidCodeAuthFlowFactory

if (graph.authService.isAuthenticationConfigured) {
    val authFactory = AndroidCodeAuthFlowFactory(useWebView = false)
    authFactory.registerActivity(activity)
    AuthFlowFactoryProvider.initialize(authFactory)
}

// Later, from the login UI:
graph.authService.login()

2. Initialize managed app graph (iOS/Swift)

As on Android, the optional MobileSyncLogging.configure(sink:minSeverity:) call must come before SharedDependencyGraph.doInit, because HTTP clients read the severity when they are created.

Kotlin default arguments are not exported to Swift, so pass every argument, such as triggerSync(immediately: false).

import Shared

final class AppContainer {
    static let shared = AppContainer()
    static var graph: AppGraph { shared.graph }

    let graph: AppGraph

    private init() {
        // Optional; `AppLogSink` conforms to `MobileSyncLogSink`.
        MobileSyncLogging.shared.configure(sink: AppLogSink(), minSeverity: .info)
        Shared.AuthFlowFactoryProvider.shared.doInitialize()
        let driverFactory = DriverFactory()
        let storage = AppleMobileSyncStorageFactory.shared.create()
        graph = SharedDependencyGraph.shared.doInit(
            driverFactory: driverFactory,
            storage: storage,
            appEnvironment: AppEnvironment.prelive,
            clientId: appClientId,
            clientSecret: appClientSecret
        )
    }
}

Advanced override for custom endpoints remains available:

import com.quran.shared.auth.model.AuthEnvironment
import com.quran.shared.auth.model.AuthConfig
import com.quran.shared.pipeline.SynchronizationEnvironment

val graph = SharedDependencyGraph.init(
    driverFactory = DriverFactory(context = applicationContext),
    storage = createMobileSyncStorage(applicationContext),
    environment = SynchronizationEnvironment(
        endPointURL = "https://custom-sync.example.com/auth"
    ),
    authConfig = AuthConfig(
        environment = AuthEnvironment.PRELIVE,
        clientId = appClientId
    )
)

Android backup exclusions

Sync and session lifecycle metadata is stored in the shared user_data.db database next to the data it describes (the sync token, account and outbox in sync_meta, per-item sync state in sync_item). Auth metadata and OAuth tokens stay outside the database, so a lost or recreated database does not sign the user out. Both use stable DataStore file names so apps can exclude them from backup or device transfer:

<exclude domain="file" path="datastore/quran_mobile_sync_auth.preferences_pb"/>
<exclude domain="file" path="datastore/org.publicvalue.multiplatform.oidc.tokenstore.preferences_pb"/>

Add those entries to both full-backup-content and Android 12+ data-extraction-rules when app backup is enabled.

3. Use managed services

Core API examples:

  • Observe: authState, syncStatus, syncHistory, heldItems, unsyncedItems, bookmarks, collectionsWithBookmarks, notes
  • Mutations: addBookmark, deleteBookmark, addCollection, deleteCollection, addNote, deleteNote
  • Trigger sync: triggerSync() after a burst of edits, triggerSync(immediately = true) for a user's "Sync now", or syncOnForeground() each time the app returns from the background
  • Profile: refreshProfile() when the user may have just edited their name or photo on Quran.com, such as when a Quran.com account page the app opened closes
  • Auth: use SyncAuthService.login() (it always sends prompt=login, so the provider asks for credentials again), logout(), and clearError()

Reading bookmarks are three slots (ReadingBookmarkSlot), each an ayah, a page, or empty. Page bookmarks work across mushafs: tell the library which Mushaf the app shows, and it does the page mapping.

  • Mushaf.MADANI_1405 and Mushaf.MADANI_1440, the 1405 and 1440 Madani prints, have pages 1 to 604.
  • Mushaf.INDOPAK_15_LINES has app pages 2 to 611: apps skip a cover page, so app page N is the server's page N - 1.
  • Mushaf.SHEMERLY, the Shemerly print, has pages 1 to 521. The server has no Shemerly mushaf yet (#355), so a Shemerly page is stored and synced as the 1440 page holding its verses (mushafId 1) and mapped back when read. Seven Shemerly pages (9, 364, 379, 422, 495, 500 and 516) read back as a neighbouring page. Once the server has a Shemerly id, add it to ServerMushaf (521 pages), map it to the Shemerly layout in PageLayout.of, and send Mushaf.SHEMERLY as it; apps don't change.
  • Mushaf.WARSH and Mushaf.QALOUN, the Warsh and Qaloun riwayat, have pages 1 to 604. They number verses differently, so their pages are never mapped: a page is stored with its own mushafId (20 or 22), reads unchanged in the same mushaf, and reads as EmptyReadingBookmark in any other. Riwayah app builds don't sync. If one did, today's backend rejects these ids, so the change is held and never reaches the server; a backend that accepts them but keeps only the page group would relabel the page as a 1440 page until it keeps the mushaf id (#354).
  • setPageReadingBookmark(slot, page, mushaf) throws IllegalArgumentException for a page the mushaf doesn't have.
  • A page is stored and synced with the mushaf it was set in (mushafId, such as 2 for MADANI_1405). Today's backend keeps only a page group and returns its stand-in mushaf (5 for the 604-page group, read as the 1440 layout), so after a sync a 1405 page reads as a 1440 one.
  • readingBookmarks(mushaf), setPageReadingBookmark, and renameReadingBookmark take the app's mushaf and return PageReadingBookmarks as pages of it. A page set in a mushaf of the same layout keeps its number; another becomes the page holding its middle verse. A page of a mushaf with no layout here (one this library doesn't know, or a riwayah such as Warsh, which numbers verses differently) in any other mushaf, or a page outside its mushaf's pages, reads as EmptyReadingBookmark; its location is kept, and a rename sends it back unchanged.
  • Ayah bookmarks need no mapping: every Hafs mushaf numbers verses the same way, so setAyahReadingBookmark and clearReadingBookmark take no mushaf. An ayah bookmark is taken as a Hafs (sura, ayah), in riwayah builds too, which don't sync.

Import local data through QuranDataService.importData(PersistenceImportData(...)). Each ImportCollectionAyahBookmark.collectionName must name an ImportCollection in the same batch. Names match exactly, except that any case or surrounding-whitespace variant of Favorites resolves to the default collection; entries sharing a name describe one destination. Use readingBookmarks with ImportReadingBookmark.Ayah or .Page, each specifying a ReadingBookmarkSlot (TEAL, ORANGE, or RED), lastUpdated, and an optional name; a page also names its Mushaf. A page the mushaf doesn't have is skipped with a warning; with deleteExisting = true its slot is cleared like an omitted slot. Each slot may appear once. An imported slot replaces its location, name (including clearing it when null), and modification time. Merge imports leave omitted slots unchanged; deleteExisting = true clears their locations while preserving slot identities and names for sync. Imports are atomic and report the number of imported reading bookmarks in PersistenceImportResult.readingBookmarksImported. An import writes through the repositories' rules: text is trimmed like the server, each note gets a creation time of its own (nudged by a millisecond on a clash), a verse keeps one highlight color, and only the newest 20 reading sessions are kept. Matching against notes already on the device and sync linking share one note identity (noteIdentity): the set of verse ranges (order and repeats ignored) and the body trimmed like the server, with each run of whitespace inside it (spaces, tabs, newlines) counted as one space. Tracked history still matches notes by the import fingerprint, which also ignores whitespace runs. Times don't count. An imported note equal to a note already on the device, archived ones included, is skipped and counted in matched. The next sync pushes the rest like any other local change, except that an imported note is linked, instead of pushed, to an equal note another device synced that this device hasn't fetched yet (the same legacy notes imported on two devices before signing in). With trackHistory = true, an import records a fingerprint of each item it handled, and later tracked imports skip those items, even after the user edits or deletes what they brought in. Logout keeps this history, so re-running a legacy import afterwards can't bring back deleted items or add the legacy data to another account.

Lifecycle note:

  • QuranDataService is app-scoped. Initialize once via SharedDependencyGraph.init(...).
  • Prefer SyncAuthService.logout() for managed apps. QuranDataService.logout() remains available for callers that route auth and data lifecycle through one facade.
  • Writes and imports don't wait for an interactive sign-in (SyncAuthService.login()), however long the browser stays open. A logout doesn't wait for a sign-in in progress: it ends it, so the login() call that opened the browser throws CancellationException, and another login() made while it ran, which waits on it, throws AuthenticationFailedException. A login() waiting on another sign-in throws AuthenticationCancelledException when that sign-in is cancelled, say because the call that opened the browser was cancelled. A sign-in started while a logout runs fails with SessionResetInProgressException.
  • Do not clear app-scoped services from UI/view-model teardown.

4. App environment selection

The published artifact can target either environment at runtime:

  • AppEnvironment.PRELIVE / AppEnvironment.prelive
  • AppEnvironment.PRODUCTION / AppEnvironment.production

AppEnvironment keeps auth and sync aligned by default:

  • PRELIVE -> https://prelive-oauth2.quran.foundation and https://apis-prelive.quran.foundation/auth
  • PRODUCTION -> https://oauth2.quran.foundation and https://apis.quran.foundation/auth

buildkonfig.flavor now only controls the default fallback used when the app does not pass an explicit app environment.

Default fallback (gradle.properties):

buildkonfig.flavor=debug

Override for release-like builds:

./gradlew :auth:build -Pbuildkonfig.flavor=release

Run the Demos

Android demo

  1. Open project in Android Studio.
  2. < 8240 li>Run demo/android app module.

CLI build example:

./gradlew :demo:android:assembleDebug

iOS demo

  1. Open: demo/apple/QuranSyncDemo/QuranSyncDemo.xcodeproj
  2. Select QuranSyncDemo scheme.
  3. Run on an iOS Simulator.

The Xcode project includes a Run Script phase that calls:

./gradlew :umbrella:embedAndSignAppleFrameworkForXcode

CLI build example:

xcodebuild -project demo/apple/QuranSyncDemo/QuranSyncDemo.xcodeproj \
  -scheme QuranSyncDemo \
  -configuration Debug \
  -destination 'generic/platform=iOS Simulator' \
  build CODE_SIGNING_ALLOWED=NO

Build, Test, and Release Commands

Build all:

./gradlew build

Run all tests:

./gradlew allTests --stacktrace --continue

Run the end-to-end sync integration tests (devices against a fake sync server; see AGENTS.md):

./gradlew :sync-pipelines:jvmTest --tests 'com.quran.shared.pipeline.integration.*'

Build iOS XCFramework:

./gradlew :umbrella:assembleSharedXCFramework

Build Android demo compile target:

./gradlew :demo:android:compileDebugKotlin

Build iOS KMP target:

./gradlew :sync-pipelines:compileKotlinIosSimulatorArm64

Regenerating the page tables

Page reading bookmarks are mapped across mushafs with persistence/.../readingbookmark/mapping/PageLayoutTables.kt: per-page verse ranges for the five Hafs layouts, and each mushaf ID's page count and layout. It is generated from Quran.com's mushaf-mapping bundle (src/utils/mushafMapping/client_bundle.v1.json in quran/quran.com-fe) and, for the Shemerly layout the bundle lacks, scripts/data/shemerly-pages.json; never edit it by hand. To regenerate it from a newer bundle:

  1. Download the bundle at the quran.com-fe commit you want:
    gh api "repos/quran/quran.com-fe/contents/src/utils/mushafMapping/client_bundle.v1.json?ref=<commit>" \
      -H "Accept: application/vnd.github.raw" > /tmp/client_bundle.v1.json
  2. Set SOURCE_COMMIT in scripts/generate-page-layouts.py to that commit. The file header records it and the bundle's revision.
  3. Run the generator, which overwrites PageLayoutTables.kt:
    python3 scripts/generate-page-layouts.py /tmp/client_bundle.v1.json
    It finds each layout by its exact list of mushaf IDs (LAYOUTS in the script). If the bundle moved a mushaf between page groups, the script fails: update LAYOUTS.
  4. Run the tests:
    ./gradlew :persistence:allTests
    PageLayoutTest checks the tables' page counts and coverage, BundleMushafPageCountsTest checks ServerMushaf against the bundle, and ReadingBookmarkPageMapperTest pins known mappings, so a changed layout shows up as a failing test. Mushaf IDs the bundle lacks (such as INDOPAK, 3) are mapped by hand in PageLayout.of.

The Shemerly layout comes from scripts/data/shemerly-pages.json: the [sura, ayah] each of its 521 pages starts with, exported from the Android app's ShemerlyDataSource (suraForPageArray and ayahForPageArray; the file records the source repository, path and commit). The generator ends each page one verse before the next starts and checks the layout like the bundle's. To update it, replace pageStarts (and source) in that file and run the generator as above.

Releases

The manual Release workflow bumps the version and tags it, attaches the iOS XCFramework to a GitHub Release, and publishes the Android/JVM artifacts to the GitHub Pages Maven repository on the gh-pages branch. Android apps resolve it with:

maven("https://quran.github.io/mobile-sync/maven") {
    content { includeGroup("com.quran") }
}

Maven Central releases stay manual and infrequent. Publish them from a release tag so both repositories serve the same build for a version.

Publish the GitHub Pages artifacts to a local folder to inspect them:

./gradlew publishToGitHubPages -PgithubPagesMavenDir=/tmp/mobile-sync-maven

Related Repositories

About

Quran Mobile Sync Engine

Resources

Stars

2 stars

Watchers

4 watching

Forks

Releases

Packages

Contributors

Languages

0