This version bumps the Realm file format version to 23. Realm files written by this version cannot be read by older versions of Realm.
Enhancements
- The Realm file is now automatically shrunk if the file size is larger than needed to store all of the data. (Core PR #5755)
- Pinning old versions (either with frozen Realms or with Realms on background threads that simply don't get refreshed) now only prevents overwriting the data needed by that version, rather than the data needed by that version and all later versions. In addition, frozen Realms no longer pin the transaction logs used to drive change notifications. This mostly eliminates the file size growth caused by pinning versions. (Core PR #5440)
- Rework how Dictionaries/Maps are stored in the Realm file. The new design uses less space and is typically significantly faster. This changes the iteration order of Maps, so any code relying on that may be broken. We continue to make no guarantees about iteration order on Maps (Core #5764).
- Improve performance of freezing Realms (Core PR #6211).
Fixed
- Fix a crash when using client reset with recovery and flexible sync with a single subscription (Core #6070, since v10.28.2)
- Encrypted Realm files could not be opened on devices with a larger page size than the one which originally wrote the file. (#8030, since v10.32.1)
- Creating multiple flexible sync subscriptions at once could hit an assertion failure if the server reported an error for any of them other than the last one (Core #6038, since v10.21.1).
Set<AnyRealmValue>andList<AnyRealmValue>considered a string and binary data containing that string encoded as UTF-8 to be equivalent. This could result in a List entry not changing type on assignment and for the client be inconsistent with the server if a string and some binary data with equivalent content was inserted from Atlas. (Core #4860 and Core #6201, since v10.8.0)- Querying for NaN on Decimal128 properties did not match any objects (Core #6182, since v10.8.0).
- When client reset with recovery is used and the recovery did not need to make any changes to the local Realm, the sync client could incorrectly think the recovery failed and report the error "A fatal error occured during client reset: 'A previous 'Recovery' mode reset from did not succeed, giving up on 'Recovery' mode to prevent a cycle'". (Core #6195, since v10.32.0)
- Writing to newly in-view objects while a flexible sync bootstrap was in progress would not synchronize those changes to the server (Core #5804, since v10.21.1).
- If a client reset with recovery or discard local was interrupted while the "fresh" realm was being downloaded, the sync client could crash with a MultpleSyncAgents exception (Core #6217, since v10.25.0).
- Sharing Realm files between a Catalyst app and Realm Studio did not properly synchronize access to the Realm file (Core #6258, since v10.0.0).
Compatibility
- Realm Studio: 13.0.2 or later.
- APIs are backwards compatible with all previous releases in the 10.x.y series.
- Carthage release for Swift is built with Xcode 14.2.
- CocoaPods: 1.10 or later.
- Xcode: 13.3-14.2.
Internal
- Upgraded realm-core from 12.13.0 to 13.4.0