Closes#469.
- Methods defined in `ZcashRustBackendWelding` protocol which are doing
some IO operations are now async.
- `Initializer` no longer have methods to get balance. Only
`Synchronizer` now have these methods. These methods are newly async.
And `Initilializer` doesn't have alternative APIs.
- There is lot of changes in tests but those are mostly adding `await`.
No logic is changed.
Closes#876.
- Add `testMultipleSynchronizersCanRunAtOnce()` test which creates
multiple instances of the `SDKSynchronizer` and then executes sync in
parallel.
- To achieve this it's required to have multiple instances of the
lightwalletd in the darkside mode. Because of this new test is in the
new target `AliasDarksideTests`. It's little bit harder to run this
test than to run regular darkside tests. So it has it's own target.
- Fix bug when all the instances of the `SDKSynchronizer` used same data
DB even with different aliases.
- Add script to run multiple instances of the lightwalletd in
darkside mode. And add script to stop these instances and clean after
those.
- Update `DarksideWalletService` a little so the endpoint can be passed
to it. Without this it would always use default endpoint when created
with service.
- Small cleanup in `TestCoordinator`.
Closes#484.
- `TransactionRepository` has async API. `Synchronizer` and alternative
APIs are updated accordingly.
These methods and properties in the `Synchronizer` are async now:
- `cancelSpend(transaction:)`
- All the variants of the `getMemos(for:)` method.
- All the variants fo the `getRecipients(for:)` method.
- `allConfirmedTransactions(from:limit:)`
- `pendingTransactions`
- `clearedTransactions`
- `sentTransactions`
- `receivedTransactions`
Closes#209.
[#845] Introduce ZcashSynchronizerAlias enum
Closes#845.
[#852] SDKSynchronizer using queues label based on the alias
Closes#852.
[#847] Remove posibility to use DatabaseStorageManager as singleton
Closes#847.
[#850] Remove synchronizerConnectionStateChanged notification
Closes#850.
[#855] Add check if the Alias is already used
Closes#855
- Added `UsedAliasesChecker` utility which is used to register aliases
that are in use.
- `prepare()` and `wipe()` methods now check if the current alias can't
be used and if not then `InitializerError.aliasAlreadyInUse` is
thrown/emitted.
- Some public methods that could cause harm if used with the Alias that
is already in use now throw `SynchronizerError.notPrepared`. Thanks to
this the client app is forced to call `prepare()` first. And
`prepare()` does check for the Alias.
- Added tests for new conditions.
[#849] Make InternalSyncProgress aware of the Alias
Closes#849.
[#853] Only instance with default Alias migrates legacy cache DB
Closes#853.
[#851] Apply the Alias to the URLs
Closes#851.
- `Initializer` now updates paths according to alias before paths are
used anywhere in the SDK.
- Paths update can fail. It would be incovenient for the client apps to
handle errors thrown from `Initiliazer` constructor. So the error is
then handled in `SDKSynchronizer.prepare()` or `SDKSynchronizer.wipe()`.
[#846] Stop using SDKMetrics as singleton (#862)
- metrics are not longer a singleton
- tests fixed
- metrics outside init of the synchronizer
[#848] Make logger aware of the alias
- logger is now an instance passed throughout the sdk instead of a static proxy
[#848] Make logger aware of the alias (#868)
- comments addressed
[#848] Make logger aware of the alias (#868)
- returning protocol back
Fix typos
[#856] Add possibility to test multiple synchronizers in the sample app
Closes#856.
- Added `alias` property to `Synchronizer`.
- Added `SyncBlocksListViewController` which provides UI to use multiple
synchronizers at once.
[#209] Add changelog
- Add changelog for #209.
- Overall improve readability of the rendered changelog. Tickets
references are now prefixed with `###` instead of `- `.
Fix compilation
Data DB name is derived from timestamp in TemporaryDbBuilder. Timestamp
is converted to seconds before it's used to generate Data DB name. So
if the test is super fast and can run twice in one seconds it fails.
Solution is to always remove data DB at the end of the test.
This change introduces two new protocols: `ClosureSynchronizer` and
`CombineSynchronizer`. These two protocols define API that doesn't use
`async`. So the client can choose exactly which API it wants to use.
This change also introduces two new objects: `ClosureSDKSynchronizer`
and `CombineSDKSynchronizer`. These two implement the respective protocols
mentioned above. Both are structures. Neither of these two keeps any state.
Thanks to this each is very cheap to create. And usage of these two isn't
mutually exclusive. So devs can really choose the best SDK API for each
part of the client app.
[#831] Use async inside of the SDKSynchronizer
- In general lot of methods inside the `SDKSynchronizer` and
`CompactBlockProcessoer` which weren't async are now async. And other
changes are made because of this change.
- `CompactBlockProcessor` no longer uses Combine to communicate with
`SDKSynchronizer`. Reason for this is that Combine doesn't play great
with async. Closure passed to `sink` isn't async.
- Because of this and because of how our tests work (receiving signals
from CBP directly) `CompactBlockProcessor` must be able to handle more
event closures. Not just one. So it now has `eventClosures`
dictionary. It's little bit strange but it works fine.
- `SyncStatus` inside the `SDKSynchronizer` was previously protected by
lock. Now it's protected by simple actor wrapper.
- Changes in tests are minimal. Changes were mady only because
`CompactBlockProcessor` changes from Combine to closures.
[#831] Add tests for ClosureSDKSynchronizer
- Added tests are testing in general if the `ClosuresSDKSynchronizer` is correctly
calling `Synchronizer` and if the values are correctly returned.
- `ClosuresSDKSynchronizer` doesn't contain any logic but it is public
API and we should be sure that it works correctly.
[#831] Add tests for CombineSDKSynchronizer
[#831] Add changelog
Closes#724
- All the notifications are gone. Only
`synchronizerConnectionStateChanged` stayed because it's used
internally. Let's deal with this one in another task.
- Synchronizer has now two new publishers (`stateStream` and
(`eventStream`) which are used to notify the client app about
what is going on. These publishers replace notifications.
- There is also new property `latestState` which can be used to get the
SDK state in synchronous manner.
- `SDKSynchronizer.status` is no longer public. It is used internally to
refresh `latestState` and emit new values from `stateStream.
- When `SDKSynchronizer.status` is update `notify()` function is
triggered. And this function is now responsible for generating new
snapshot of `SynchronizerState` and updating `latestState` and
`stateStream`.
Closes#826
- `viewingKeys` and `walletBirthday` are removed from `Initializer`
constuctor. These parameters are moved to `SDKSynchronizer.prepare`
function.
- Constructor of the `SDKSynchronizer` no longer throws exception.
- Any value emitted from `lastState` stream before `SDKSynchronizer.prepare`
is called has `latestScannedHeight` set to 0.
- `Initializer.initialize` function isn't public anymore. To initialize
SDK call `SDKSynchronizer.prepare` instead.
- Real birthday is now known after the `SDKSynchronizer.prepare` is
called. But this function isn't async and therefore it's hard to
pass birthday from here to `CompactBlockProcessor` in synchronous
manner. We must be sure that it's safe to call start() after prepare
function finishes.
For this reason `CompactBlockProcessor` has `walletBirthdayProvider`
instead of `walletBirthday`. And this new parameter is closure which
reads birthday from `Initiliazer` on each call. Thanks to this
`CompactBlockProcessor` has always the right birthday.
- Some cleanup is done in `TestCoordinator`.
contains no public API changes.
Adds `func shouldClearBlockCacheAndUpdateInternalState() -> BlockHeight?` to `SyncRanges`
so that the compact block processor can advert internal states that are not consistent and
recover from such state.
Closes#821
Closes#816
- `SDKSynchronizer.rewind(policy:)` can be now called anytime.
- If the sync is in progress then the sync is first stopped and then rewind is executed.
- Rewind now returns AnyPublisher which completes or fails when wipe is done.
- `AgterSyncHooksManager` is used to execute rewind after sync process
is stopped.
Closes#801
- `SDKSynchronizer.wipe()` can be now called anytime.
- If the sync is in progress then the sync is first stopped and then
wipe is executed.
- Wipe now returns AnyPublisher which completes or fails when wipe is
done.
- Majority of wipe's work is to delete files. That is only operation
that can throw error during wipe. This operation should succeed every
time. If it fails that something is seriously wrong. When this happens
the SDK can happen in inconsistent state. There is no recovery for
now. Only way how to fix this is to reinstall the app in the device.
- Added hooks mechanism. This is implemented in `AfterSyncHooksManager`
and it is used by `CompactBlockProcessor`. It's just more organized
way how to track what should happen when the sync process is canceled.
Closes#795
- Added structure `SaplingParamsSourceURL` which holds source URLs for
sapling-spend file and sapling-output file. URLs are ultimately passed
to `SaplingParameterDownloader`.
- This is done so it's possible to pass different URLs when running
tests.
- sapling-spend file and sapling-output file are in tests bundle. And
when running tests these files are loaded from bundle and not from
network.
Closes#778
- .swiftlint.yml is now used to lint only the code of the SDK.
- .swiftlint_tests.yml is now used to lint only the code of the tests.
This config disables rules that were previously disabled in the code
in lot of tests.
Closes#764
- All notification that were previously sent from CompactBlockProcessor are now gone.
- `CompactBlockProcessor` now provides `eventStream` to communicate with `SDKSynchronizer`.
- Added `synchronizerStoredUTXOs` notification.
- Tests are updated to not use notifications anymore. Some tests there
weren't working before are now fixed.
- Added `CompactBlockProcessorEventHandler` utility class. Previously
expectations were subscribing to notifications. This class replaces
this functionality. It subscribes to events from
`CompactBlockProcessor` and fullfill expectations.
Closes#771
- Renamed `cacheDirectory` param in `FSCompactBlockRepository`
constructor to `fsBlockDbRoot`.
- Pass same paths to `FSCompactBlockRepository` and `FSMetadataStore`
from tests.
This implements `getRecipients()` function which retrieves the possible
recipients from a sent transaction. These can either be addresses or
internal accounts depending on the transaction being a shielding tx
or a regular outgoing transaction.
Closes#767
Other changes:
Fix version of zcash-light-client-ffi to 0.1.1
Enhance error reporting on a test make Mock comply with protocol
Update CHANGELOG.md
Closes#746
- `LightWalletGRPCService` is no longer public. `LightWalletService` is no longer public.
- `LightWalletGRPCService` shouldn't be used dicrectly. Use `LightWalletServiceFactory` to create instance of the service.
- Moved sending of `blockProcessorConnectivityStateChanged` notification to `Initilizer`.
- All the errors from GRPC are mapped to `LightWalletServiceError`. Handling of `GRPCStatus` in the SDK is no longer required.
- `Service` directory (that contains GRPC code) is moved to `Modules/Service`. We can put more code to `Modules/` if we decide
to modularize more code.
Closes#743
- Added new protocol `BlockValidator` and `BlockValidatorImpl` class. All the code related to blocks validation is extracted to `BlockValidatorImpl`.
- Fixed `testSingleDownloadAndScan()` test.
- Changed how validation error are handled. Previously validation error were handled directly next to validation code. And when some validation
errors happened method from `CompactBlockProcessor` were called. Now validation errors are handled in `CompactBlockProcessor.processNewBlocks`
in a same way as any other error that can occur during sync process.
Closes https://github.com/zcash/ZcashLightClientKit/issues/697
Closes https://github.com/zcash/ZcashLightClientKit/issues/720
Closes https://github.com/zcash/ZcashLightClientKit/issues/587
Closes https://github.com/zcash/ZcashLightClientKit/issues/667
Closes https://github.com/zcash/ZcashLightClientKit/issues/443
Closes https://github.com/zcash/ZcashLightClientKit/issues/754
- [#790] Fix ShieldFundsTests
Closes#790
Removes comments on `ShieldFundsTests` since those issues have been fixed
Depends on zcash-light-client-ffi changes that adopt newer versions of
librustzcash crates `zcash_primitives 0.10`, `zcash_client_backend 0.7`,
`zcash_proofs 0.10`, `zcash_client_sqlite 0.5.0`.
Also allows wallets to define a shielding_threshold and will set
foundations to customize minimum confirmations for balances, spends
and shielding operations.
**Test Bootstrapping**
- `ZcashCompactBlockDescriptor`: struct that holds functions to
describe blocks as filenames and compare those filenames
`ZcashCompactBlockDescriptor.live` has the actual implementation
but it can be replaced by mocks if needed on Tests
main implementations are held under `FSCompactBlockRepository.filenameDescription` and `FSCompactBlockRepository.filenameComparison` on a separate extention
`DirectoryListingProviders` provide two default implementations of listing a
directory deterministically. `FileManager` does not define a sorting and needs to be done in-memory by calling `.sorted()` on the resulting collection. If this
is a big toll on performance it can be changed to a POSIX implementation but
this is good for now.
`ZcashCompactBlockDescriptor` adds a `height` helper function to
turn a filename into the height of the block stored.
Implemented `func latestHeight() throws -> BlockHeight ` that
returns the blockheight by querying the cache directory in a sorted
fashion and getting the last value and turning the filename into a
`BlockHeight`
Added `Meta` struct to ZcashCompactBlock.
Tests implemented:
- `filterBlockFiles`
- `testClearTheCache`
- `testLatestHeightEmptyCacheThrows`
- `testLatestHeightEmptyCacheThrowsAsync`
- `testRewindEmptyCacheDoesNothing`
- `testRewindEmptyCacheDoesNothingAsync`
- `testWhenBlockIsStoredItFollowsTheDescribedFormat`
- `testWhenBlockIsStoredItFollowsTheFilenameConvention`
- `testGetLatestHeight`
- `testRewindDeletesTheRightBlocks` test
- `testPerformanceExample` test. This isn't a real performance test because the API doesn't work with async/await yet
adopts `shield_funds` shielding threshold parameter
Implements `initBlockMetadataDb` and fix tests
Renames dbCache parameter to `fsBlockDbRoot`. Builds but tests don't pass.
Removes cacheDb uses from code. Testing utilities still persist.
Added needed information in MIGRATING and CHANGELOG.
Added helper to perform deletion of legacy db and creation a the
new file system backed cache.
Renames parameters and changes code where needed.
Network Constants turned into `enum` with static methods.
DeletelastDownloadedBlock helper from initializer
Removes CompactBlockStorage and CompactBlockEntity.
Implements `latestCachedBlockHeight` on rustbackend.
*Replaces dependencies on ZcashRustWelding with `FSMetadataStore`*
This allows the tests to not depend in a particular implementation
of either the MockRustBackend of or ZcashRustBackend. Also provides
a way to test errors properly and switch implementations of critical
areas like `writeBlocks`.
Closes#740
`CompactBlockDownloading` protocol is renamed to `BlockDownloaderService`. `CompactBlockDownloader` is renamed to `BlockDownloaderServiceImpl`.
What is the motivation for this rename?
Goal of this change is to extract code from `CompactBlockProcessor` which is related to downloading. So naturaly it should be called "downloader".
But `CompactBlockDownloader` was already taken. After analysis of `CompactBlockDownloading` and `CompactBlockDownloader` I found out that
`CompactBlockDownloader` doesn't download blocks exactly. It's more like abstraction over `LightWalletService`. And it does very similar job to
`LightWalletService`. That is why I decided to rename it to "service".
What is the motivation for protoco/impl naming scheme?
First `BlockDownloaderService` protocol doesn't describe capability in my eyes. It describe what something is. Also `BlockDownloaderServicing` sounds
strange. I agree that add suffix "Impl" to class name isn't super nice. But it is acceptable. And it makes sense to use nicer type name
(e.g. `BlockDownloaderService`) for protocol because it's used on much more places in the code.
- Added `BlockDownloader` protocol and `BlockDownloaderImpl` class. The code related to downloading was exctracted to this class.
Michal Fousek
Francisco Gindre
Lukas Korba
Francisco Gindre