* Automatically elide empty inputs and outputs to backend methods
* Refactor service generation
Despite the fact that the majority of our Protobuf service methods require
an open collection, they were not accessible with just a Collection
object. To access the methods (e.g. because we haven't gotten around to
exposing the correct API in Collection yet), you had to wrap the collection
in a Backend object, and pay a mutex-acquisition cost for each call, even
if you have exclusive access to the object.
This commit migrates the majority of service methods to the Collection, so
they can now be used directly, and improves the ergonomics a bit at the
same time.
The approach taken:
- The service generation now happens in rslib instead of anki_proto, which
avoids the need for trait constraints and associated types.
- Service methods are assumed to be collection-based by default. Instead of
implementing the service on Backend, we now implement it on Collection, which
means our methods no longer need to use self.with_col(...).
- We automatically generate methods in Backend which use self.with_col() to
delegate to the Collection method.
- For methods that are only appropriate for the backend, we add a flag in
the .proto file. The codegen uses this flag to write the method into a
BackendFooService instead of FooService, which the backend implements.
- The flag can also allows us to define separate implementations for collection
and backend, so we can e.g. skip the collection mutex in the i18n service
while also providing the service on a collection.
Previously it was Backend's responsibility to store the last progress,
and when calling routines in Collection, one had to construct and pass
in a Fn, which wasn't the most ergonomic. This PR adds the last progress
state to the collection, so that the routines no longer need a separate
progress arg, and makes some other tweaks to improve ergonomics.
ThrottlingProgressHandler has been tweaked so that it now stores the
current state, so that callers don't need to store it separately. When
a long-running routine starts, it calls col.new_progress_handler(),
which automatically initializes the data to defaults, and updates the
shared UI state, so we no longer need to manually update the state at
the start of an operation.
The backend shares the Arc<Mutex<>> with the collection, so it can get
at the current state, and so we can update the state when importing a
backup.
Other tweaks:
- The current Incrementor was awkward to use in the media check, which
uses a single incrementing value across multiple method calls, so I've
added a simpler alternative for such cases. The old incrementor method
has been kept, but implemented directly on ThrottlingProgressHandler.
- The full sync code was passing the progress handler in a complicated
way that may once have been required, but no longer is.
- On the Qt side, timers are now stopped before deletion, or they keep
running for a few seconds.
- I left the ChangeTracker using a closure, as it's used for both importing
and syncing.
Will be handy to use it in our other scripts in the future too - thanks
Rumo!
Results of benchmarking ./run before and after these crate splits:
- Touching a proto file leads to a slight increase: about +90ms
- Touching an rslib file leads to a bigger decrease, as there's less to
recompile: about -700ms
And ./ninja test is even better: about +200ms and -3800ms.
This PR replaces the existing Python-driven sync server with a new one in Rust.
The new server supports both collection and media syncing, and is compatible
with both the new protocol mentioned below, and older clients. A setting has
been added to the preferences screen to point Anki to a local server, and a
similar setting is likely to come to AnkiMobile soon.
Documentation is available here: <https://docs.ankiweb.net/sync-server.html>
In addition to the new server and refactoring, this PR also makes changes to the
sync protocol. The existing sync protocol places payloads and metadata inside a
multipart POST body, which causes a few headaches:
- Legacy clients build the request in a non-deterministic order, meaning the
entire request needs to be scanned to extract the metadata.
- Reqwest's multipart API directly writes the multipart body, without exposing
the resulting stream to us, making it harder to track the progress of the
transfer. We've been relying on a patched version of reqwest for timeouts,
which is a pain to keep up to date.
To address these issues, the metadata is now sent in a HTTP header, with the
data payload sent directly in the body. Instead of the slower gzip, we now
use zstd. The old timeout handling code has been replaced with a new implementation
that wraps the request and response body streams to track progress, allowing us
to drop the git dependencies for reqwest, hyper-timeout and tokio-io-timeout.
The main other change to the protocol is that one-way syncs no longer need to
downgrade the collection to schema 11 prior to sending.
The Rust community appear to have converged on tracing - it's used by
the Rust compiler, and receives close to 10x the number of downloads
that slog does. Its API is more ergonomic, and it does a much nicer
job with async rust.
To make this change, we no longer pass around explicit loggers, and rely
on a globally-registered one. The log file location has been changed
from one in each profile folder to a single one in the base folder. This
will remain empty for most users, since only errors are logged by default,
but may be useful for debugging future changes.
* Run cargo +nightly fmt
* Latest prost-build includes clippy workaround
* Tweak Rust protobuf imports
- Avoid use of stringify!(), as JetBrains editors get confused by it
- Stop merging all protobuf symbols into a single namespace
* Remove some unnecessary qualifications
Found via IntelliJ lint
* Migrate some asserts to assert_eq/ne
* Remove mention of node_modules exclusion
This no longer seems to be necessary after migrating away from Bazel,
and excluding it means TS/Svelte files can't be edited properly.
This reverts commit 09cb8b3cf6.
Overhead on larger folders/slower devices is more than I originally
anticipated, and can run into multiple seconds. This seems to be
particularly egregious on mobile, which I presume is due to sandboxing
overhead.
(for upgrading users, please see the notes at the bottom)
Bazel brought a lot of nice things to the table, such as rebuilds based on
content changes instead of modification times, caching of build products,
detection of incorrect build rules via a sandbox, and so on. Rewriting the build
in Bazel was also an opportunity to improve on the Makefile-based build we had
prior, which was pretty poor: most dependencies were external or not pinned, and
the build graph was poorly defined and mostly serialized. It was not uncommon
for fresh checkouts to fail due to floating dependencies, or for things to break
when trying to switch to an older commit.
For day-to-day development, I think Bazel served us reasonably well - we could
generally switch between branches while being confident that builds would be
correct and reasonably fast, and not require full rebuilds (except on Windows,
where the lack of a sandbox and the TS rules would cause build breakages when TS
files were renamed/removed).
Bazel achieves that reliability by defining rules for each programming language
that define how source files should be turned into outputs. For the rules to
work with Bazel's sandboxing approach, they often have to reimplement or
partially bypass the standard tools that each programming language provides. The
Rust rules call Rust's compiler directly for example, instead of using Cargo,
and the Python rules extract each PyPi package into a separate folder that gets
added to sys.path.
These separate language rules allow proper declaration of inputs and outputs,
and offer some advantages such as caching of build products and fine-grained
dependency installation. But they also bring some downsides:
- The rules don't always support use-cases/platforms that the standard language
tools do, meaning they need to be patched to be used. I've had to contribute a
number of patches to the Rust, Python and JS rules to unblock various issues.
- The dependencies we use with each language sometimes make assumptions that do
not hold in Bazel, meaning they either need to be pinned or patched, or the
language rules need to be adjusted to accommodate them.
I was hopeful that after the initial setup work, things would be relatively
smooth-sailing. Unfortunately, that has not proved to be the case. Things
frequently broke when dependencies or the language rules were updated, and I
began to get frustrated at the amount of Anki development time I was instead
spending on build system upkeep. It's now about 2 years since switching to
Bazel, and I think it's time to cut losses, and switch to something else that's
a better fit.
The new build system is based on a small build tool called Ninja, and some
custom Rust code in build/. This means that to build Anki, Bazel is no longer
required, but Ninja and Rust need to be installed on your system. Python and
Node toolchains are automatically downloaded like in Bazel.
This new build system should result in faster builds in some cases:
- Because we're using cargo to build now, Rust builds are able to take advantage
of pipelining and incremental debug builds, which we didn't have with Bazel.
It's also easier to override the default linker on Linux/macOS, which can
further improve speeds.
- External Rust crates are now built with opt=1, which improves performance
of debug builds.
- Esbuild is now used to transpile TypeScript, instead of invoking the TypeScript
compiler. This results in faster builds, by deferring typechecking to test/check
time, and by allowing more work to happen in parallel.
As an example of the differences, when testing with the mold linker on Linux,
adding a new message to tags.proto (which triggers a recompile of the bulk of
the Rust and TypeScript code) results in a compile that goes from about 22s on
Bazel to about 7s in the new system. With the standard linker, it's about 9s.
Some other changes of note:
- Our Rust workspace now uses cargo-hakari to ensure all packages agree on
available features, preventing unnecessary rebuilds.
- pylib/anki is now a PEP420 implicit namespace, avoiding the need to merge
source files and generated files into a single folder for running. By telling
VSCode about the extra search path, code completion now works with generated
files without needing to symlink them into the source folder.
- qt/aqt can't use PEP420 as it's difficult to get rid of aqt/__init__.py.
Instead, the generated files are now placed in a separate _aqt package that's
added to the path.
- ts/lib is now exposed as @tslib, so the source code and generated code can be
provided under the same namespace without a merging step.
- MyPy and PyLint are now invoked once for the entire codebase.
- dprint will be used to format TypeScript/json files in the future instead of
the slower prettier (currently turned off to avoid causing conflicts). It can
automatically defer to prettier when formatting Svelte files.
- svelte-check is now used for typechecking our Svelte code, which revealed a
few typing issues that went undetected with the old system.
- The Jest unit tests now work on Windows as well.
If you're upgrading from Bazel, updated usage instructions are in docs/development.md and docs/build.md. A summary of the changes:
- please remove node_modules and .bazel
- install rustup (https://rustup.rs/)
- install rsync if not already installed (on windows, use pacman - see docs/windows.md)
- install Ninja (unzip from https://github.com/ninja-build/ninja/releases/tag/v1.11.1 and
place on your path, or from your distro/homebrew if it's 1.10+)
- update .vscode/settings.json from .vscode.dist
* Add crate snafu
* Replace all inline structs in AnkiError
* Derive Snafu on AnkiError
* Use snafu for card type errors
* Use snafu whatever error for InvalidInput
* Use snafu for NotFoundError and improve message
* Use snafu for FileIoError to attach context
Remove IoError.
Add some context-attaching helpers to replace code returning bare
io::Errors.
* Add more context-attaching io helpers
* Add message, context and backtrace to new snafus
* Utilize error context and backtrace on frontend
* Rename LocalizedError -> BackendError.
* Remove DocumentedError.
* Have all backend exceptions inherit BackendError.
* Rename localized(_description) -> message
* Remove accidentally committed experimental trait
* invalid_input_context -> ok_or_invalid
* ensure_valid_input! -> require!
* Always return `Err` from `invalid_input!`
Instead of a Result to unwrap, the macro accepts a source error now.
* new_tempfile_in_parent -> new_tempfile_in_parent_of
* ok_or_not_found -> or_not_found
* ok_or_invalid -> or_invalid
* Add crate convert_case
* Use unqualified lowercase type name
* Remove uses of snafu::ensure
* Allow public construction of InvalidInputErrors (dae)
Needed to port the AnkiDroid changes.
* Make into_protobuf() public (dae)
Also required for AnkiDroid. Not sure why it worked previously - possible
bug in older Rust version?
* Refactor MediaManager transactions
* Add media entries when importing
* Fix legacy apkg import not setting csum
Also test for this in the roundtrip test.
* Avoid reallocating MediaCopier's buffer
* Make sha1 optional (dae)
* Add apkg export on backend
* Filter out missing media-paths at write time
* Make TagMatcher::new() infallible
* Gather export data instead of copying directly
* Revert changes to rslib/src/tags/
* Reuse filename_is_safe/check_filename_safe()
* Accept func to produce MediaIter in export_apkg()
* Only store file folder once in MediaIter
* Use temporary tables for gathering
export_apkg() now accepts a search instead of a deck id. Decks are
gathered according to the matched notes' cards.
* Use schedule_as_new() to reset cards
* ExportData → ExchangeData
* Ignore ascii case when filtering system tags
* search_notes_cards_into_table →
search_cards_of_notes_into_table
* Start on apkg importing on backend
* Fix due dates in days for apkg export
* Refactor import-export/package
- Move media and meta code into appropriate modules.
- Normalize/check for normalization when deserializing media entries.
* Add SafeMediaEntry for deserialized MediaEntries
* Prepare media based on checksums
- Ensure all existing media files are hashed.
- Hash incoming files during preparation to detect conflicts.
- Uniquify names of conflicting files with hash (not notetype id).
- Mark media files as used while importing notes.
- Finally copy used media.
* Handle encoding in `replace_media_refs()`
* Add trait to keep down cow boilerplate
* Add notetypes immediately instaed of preparing
* Move target_col into Context
* Add notes immediately instaed of preparing
* Note id, not guid of conflicting notes
* Add import_decks()
* decks_configs → deck_configs
* Add import_deck_configs()
* Add import_cards(), import_revlog()
* Use dyn instead of generic for media_fn
Otherwise, would have to pass None with type annotation in the default
case.
* Fix signature of import_apkg()
* Fix search_cards_of_notes_into_table()
* Test new functions in text.rs
* Add roundtrip test for apkg (stub)
* Keep source id of imported cards (or skip)
* Keep source ids of imported revlog (or skip)
* Try to keep source ids of imported notes
* Make adding notetype with id undoable
* Wrap apkg import in transaction
* Keep source ids of imported deck configs (or skip)
* Handle card due dates and original due/did
* Fix importing cards/revlog
Card ids are manually uniquified.
* Factor out card importing
* Refactor card and revlog importing
* Factor out card importing
Also handle missing parents .
* Factor out note importing
* Factor out media importing
* Maybe upgrade scheduler of apkg
* Fix parent deck gathering
* Unconditionally import static media
* Fix deck importing edge cases
Test those edge cases, and add some global test helpers.
* Test note importing
* Let import_apkg() take a progress func
* Expand roundtrip apkg test
* Use fat pointer to avoid propogating generics
* Fix progress_fn type
* Expose apkg export/import on backend
* Return note log when importing apkg
* Fix archived collection name on apkg import
* Add CollectionOpWithBackendProgress
* Fix wrong Interrupted Exception being checked
* Add ClosedCollectionOp
* Add note ids to log and strip HTML
* Update progress when checking incoming media too
* Conditionally enable new importing in GUI
* Fix all_checksums() for media import
Entries of deleted files are nulled, not removed.
* Make apkg exporting on backend abortable
* Return number of notes imported from apkg
* Fix exception printing for QueryOp as well
* Add QueryOpWithBackendProgress
Also support backend exporting progress.
* Expose new apkg and colpkg exporting
* Open transaction in insert_data()
Was slowing down exporting by several orders of magnitude.
* Handle zstd-compressed apkg
* Add legacy arg to ExportAnkiPackage
Currently not exposed on the frontend
* Remove unused import in proto file
* Add symlink for typechecking of import_export_pb2
* Avoid kwargs in pb message creation, so typechecking is not lost
Protobuf's behaviour is rather subtle and I had to dig through the docs
to figure it out: set a field on a submessage to automatically assign
the submessage to the parent, or call SetInParent() to persist a default
version of the field you specified.
* Avoid re-exporting protobuf msgs we only use internally
* Stop after one test failure
mypy often fails much faster than pylint
* Avoid an extra allocation when extracting media checksums
* Update progress after prepare_media() finishes
Otherwise the bulk of the import ends up being shown as "Checked: 0"
in the progress window.
* Show progress of note imports
Note import is the slowest part, so showing progress here makes the UI
feel more responsive.
* Reset filtered decks at import time
Before this change, filtered decks exported with scheduling remained
filtered on import, and maybe_remove_from_filtered_deck() moved cards
into them as their home deck, leading to errors during review.
We may still want to provide a way to preserve filtered decks on import,
but to do that we'll need to ensure we don't rewrite the home decks of
cards, and we'll need to ensure the home decks are included as part of
the import (or give an error if they're not).
https://github.com/ankitects/anki/pull/1743/files#r839346423
* Fix a corner-case where due dates were shifted by a day
This issue existed in the old Python code as well. We need to include
the user's UTC offset in the exported file, or days_elapsed falls back
on the v1 cutoff calculation, which may be a day earlier or later than
the v2 calculation.
* Log conflicting note in remapped nt case
* take_fields() → into_fields()
* Alias `[u8; 20]` with `Sha1Hash`
* Truncate logged fields
* Rework apkg note import tests
- Use macros for more helpful errors.
- Split monolith into unit tests.
- Fix some unknown error with the previous test along the way.
(Was failing after 969484de4388d225c9f17d94534b3ba0094c3568.)
* Fix sorting of imported decks
Also adjust the test, so it fails without the patch. It was only passing
before, because the parent deck happened to come before the
inconsistently capitalised child alphabetically. But we want all parent
decks to be imported before their child decks, so their children can
adopt their capitalisation.
* target[_id]s → existing_card[_id]s
* export_collection_extracting_media() → ...
export_into_collection_file()
* target_already_exists→card_ordinal_already_exists
* Add search_cards_of_notes_into_table.sql
* Imrove type of apkg export selector/limit
* Remove redundant call to mod_schema()
* Parent tooltips to mw
* Fix a crash when truncating note text
String::truncate() is a bit of a footgun, and I've hit this before
too :-)
* Remove ExportLimit in favour of separate classes
* Remove OpWithBackendProgress and ClosedCollectionOp
Backend progress logic is now in ProgressManager. QueryOp can be used
for running on closed collection.
Also fix aborting of colpkg exports, which slipped through in #1817.
* Tidy up import log
* Avoid QDialog.exec()
* Default to excluding scheuling for deck list deck
* Use IncrementalProgress in whole import_export code
* Compare checksums when importing colpkgs
* Avoid registering changes if hashes are not needed
* ImportProgress::Collection → ImportProgress::File
* Make downgrading apkgs depend on meta version
* Generalise IncrementableProgress
And use it in entire import_export code instead.
* Fix type complexity lint
* Take count_map for IncrementableProgress::get_inner
* Replace import/export env with Shift click
* Accept all args from update() for backend progress
* Pass fields of ProgressUpdate explicitly
* Move update_interval into IncrementableProgress
* Outsource incrementing into Incrementor
* Mutate ProgressUpdate in progress_update callback
* Switch import/export legacy toggle to profile setting
Shift would have been nice, but the existing shortcuts complicate things.
If the user triggers an import with ctrl+shift+i, shift is unlikely to
have been released by the time our code runs, meaning the user accidentally
triggers the new code. We could potentially wait a while before bringing
up the dialog, but then we're forced to guess at how long it will take the
user to release the key.
One alternative would be to use alt instead of shift, but then we need to
trigger our shortcut when that key is pressed as well, and it could
potentially cause a conflict with an add-on that already uses that
combination.
* Show extension in export dialog
* Continue to provide separate options for schema 11+18 colpkg export
* Default to colpkg export when using File>Export
* Improve appearance of combo boxes when switching between apkg/colpkg
+ Deal with long deck names
* Convert newlines to spaces when showing fields from import
Ensures each imported note appears on a separate line
* Don't separate total note count from the other summary lines
This may come down to personal preference, but I feel the other counts
are equally as important, and separating them feels like it makes it
a bit easier to ignore them.
* Fix 'deck not normal' error when importing a filtered deck for the 2nd time
* Fix [Identical] being shown on first import
* Revert "Continue to provide separate options for schema 11+18 colpkg export"
This reverts commit 8f0b2c175f.
Will use a different approach
* Move legacy support into a separate exporter option; add to apkg export
* Adjust 'too new' message to also apply to .apkg import case
* Show a better message when attempting to import new apkg into old code
Previously the user could end seeing a message like:
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xb5 in position 1: invalid start byte
Unfortunately we can't retroactively fix this for older clients.
* Hide legacy support option in older exporting screen
* Reflect change from paths to fnames in type & name
* Make imported decks normal at once
Then skip special casing in update_deck(). Also skip updating
description if new one is empty.
Co-authored-by: Damien Elmes <gpg@ankiweb.net>
* Write media files in chunks
* Test media file writing
* Add iter `ReadDirFiles`
* Remove ImportMediaError, fail fatally instead
Partially reverts commit f8ed4d89ba.
* Compare hashes of media files to be restored
* Improve `MediaCopier::copy()`
* Restore media files atomically with tempfile
* Make downgrade flag an enum
* Remove SchemaVersion::Latest in favour of Option
* Remove sha1 comparison again
* Remove unnecessary repr(u8) (dae)
The old Python code was only checking for NFC encoding, but we should
check for other issues like special filenames on windows (eg con.mp3)
- On export, the user is told to use Check Media if their media has
invalid filenames.
- On import, legacy packages will be transparently normalized. Since we're
doing the checks on export as well, any invalid names in a v3 package
are an error.
* Canonify import of i18n module
Should always be imported as `tr`, or `tr2` if there is a name collision
(Svelte).
* Add helper for garbage collecting ftl strings
Also add a serializer for ftl asts.
* Add helper for filter-mapping `DirEntry`s
* Fix `i18n_helpers/BUILD.bazel`
* run cargo-raze
* Refactor `garbage_collection.rs`
- Improve helper for file iterating
- Remove unused terms as well
- Fix issue with checking for nested messages by switching to a regex-
based approach (which runs before deleting)
- Some more refactorings and lint fixes
* Fix lints in `serialize.rs`
* Write json pretty and sorted
* Update `serialize.rs` and fix header
* Fix doc and remove `dbg!`
* Add binaries for ftl garbage collection
Also relax type constraints and strip debug tests.
* add rust_binary targets for i18n helpers (dae)
* add scripts to update desktop usage/garbage collect (dae)
Since we've already diverged from 2.1.49, we won't gain anything
from generating a stable json just yet. But once 2.1.50 is released,
we should run 'ftl/update-desktop-usage.sh stable'.
* add keys from AnkiMobile (dae)
* Mention caveats in `remove-unused.sh`
In the old HTML editor, filenames were % escaped before feeding them to
beautifulsoup, causing bare ampersands to be left alone. The new HTML
editor reads content from the DOM, where a bare ampersand has been
transformed into an &, and that gets saved back into the field,
so the media check now needs to deal with it for images as well.
https://forums.ankiweb.net/t/causing-problems-with-image-names/12171
- tokio 1.0
- updated reqwest, thanks to Rumo
- other minor dep updates
the reqwest build file has been split into two, as it was awkward
to manually update the combined file, and the platform gate is now
on the target in rslib/
So, this is fun. Apparently "DeckId" is considered preferable to the
"DeckID" were were using until now, and the latest clippy will start
warning about it. We could of course disable the warning, but probably
better to bite the bullet and switch to the naming that's generally
considered best.
Instead of generating a fluent.proto file with a giant enum, create
a .json file representing the translations that downstream consumers
can use for code generation.
This enables the generation of a separate method for each translation,
with a docstring that shows the actual text, and any required arguments
listed in the function signature.
The codebase is still using the old enum for now; updating it will need
to come in future commits, and the old enum will need to be kept
around, as add-ons are referencing it.
Other changes:
- move translation code into a separate crate
- store the translations on a per-file/module basis, which will allow
us to avoid sending 1000+ strings on each JS page load in the future
- drop the undocumented support for external .ftl files, that we weren't
using
- duplicate strings in translation files are now checked for at build
time
- fix i18n test failing when run outside Bazel
- drop slog dependency in i18n module
- Introduced a new transact() method that wraps the return value
in a separate struct that describes the changes that were made.
- Changes are now gathered from the undo log, so we don't need to
guess at what was changed - eg if update_note() is called with identical
note contents, no changes are returned. Card changes will only be set
if cards were actually generated by the update_note() call, and tag
will only be set if a new tag was added.
- mw.perform_op() has been updated to expect the op to return the changes,
or a structure with the changes in it, and it will use them to fire the
change hook, instead of fetching the changes from undo_status(), so there
is no risk of race conditions.
- the various calls to mw.perform_op() have been split into separate
files like card_ops.py. Aside from making the code cleaner, this works
around a rather annoying issue with mypy. Because we run it with
no_strict_optional, mypy is happy to accept an operation that returns None,
despite the type signature saying it requires changes to be returned.
Turning no_strict_optional on for the whole codebase is not practical
at the moment, but we can enable it for individual files.
Still todo:
- The cursor keeps moving back to the start of a field when typing -
we need to ignore the refresh hook when we are the initiator.
- The busy cursor icon should probably be delayed a few hundreds ms.
- Still need to think about a nicer way of handling saveNow()
- op_made_changes(), op_affects_study_queue() might be better embedded
as properties in the object instead
