reliable 0.3.3

An offline-first, secure data repository with automatic sync, retry, and encryption for Dart & Flutter.

CHANGELOG

0.3.3

• Merge branch 'main' into prod

0.3.2

• Merge branch 'main' into prod

0.3.1

• docs: remove Discord, point contributors to issues and email

0.3.0

• chore: standardize README and CI/CD workflows, remove donation links, update contact info and repository links
• chore: resolve analyzer issues

0.2.1

• Released @ 10/2026 (UTC)
• A read that got no answer at all (a ReliableNetworkException without a statusCode: offline, dropped connection, timeout, a request the adapter chose not to send) now logs at Log.alert whether or not the cache could answer, and also when the error propagates; being offline is expected, not a bug. Server error statuses, malformed answers and bugs still log at Log.err. ReliableNetworkException.statusCode documents the contract: an adapter throws a status-less exception only for "no answer", and lets a bug propagate as itself

0.2.0

• Released @ 10/2026 (UTC)
• BREAKING: fetch now throws a FormatException (naming the keys it saw) for a list answer of unknown shape, instead of silently reading it as [] and serving stale cache forever; a list entry that is not an object also throws a FormatException. {items: [...]} is now unwrapped alongside a bare array and {data: [...]}; a null body still reads as []
• A read whose answer lands after purgeCollections, clearUserData or a setActiveUser switch to another user no longer writes it into the cache (the caller still gets the answer); a purge arriving part-way through a multi-step cache write stops it, and the purge clears what landed
• SembastAdapter reopens a dead database connection (iOS Safari killing IndexedDB under a backgrounded tab) and retries the operation once, giving up for the session if the reopen fails, hangs past reopenTimeout, or is followed straight away by another dead-connection failure; new optional constructor parameters reopenTimeout and openDatabase, plus recoveryGaveUp and SembastAdapter.isDeadConnectionError
• A failed read served from cache (NETWORK_OR_ELSE_CACHE) now logs at Log.alert instead of Log.err when the request got no answer at all (offline, dropped connection, timeout); Log.err remains when nothing answered from cache, the error propagates, the server answered with an error status (ReliableNetworkException.statusCode set, e.g. 401 or 5xx), or the answer was malformed
• A good read whose cache copy cannot be saved is returned as a success, with the failed save logged as a Log.alert warning (REPO_CACHE_WRITE_FAILED) instead of failing the read

0.1.1

• Update dependencies

0.1.0

• Initial release