* Update JS deps
* Update semver-compat Rust deps
* Update some semver-incompat Rust deps
- hyper/axum held back because reqwests is not ready
- rusqlite held back due to burn-rs incompat version
- wiremock held back due to compile issue
* pylint wants changes to our _rsbridge.pyi
* Update Python deps
Also solves a security warning in orjson
Reformat with latest black
The 0.14.12 release appears to have broken "-A duplicate". Fix by
updating our checks to use the latest release/format.
Also update iana-time-zone, which was yanked, and ignore safemem,
which is only used when bundling.
* Pack FSRS data into card.data
* Update FSRS card data when preset or weights change
+ Show FSRS stats in card stats
* Show a warning when there's a limited review history
* Add some translations; tweak UI
* Fix default requested retention
* Add browser columns, fix calculation of R
* Property searches
eg prop:d>0.1
* Integrate FSRS into reviewer
* Warn about long learning steps
* Hide minimum interval when FSRS is on
* Don't apply interval multiplier to FSRS intervals
* Expose memory state to Python
* Don't set memory state on new cards
* Port Jarret's new tests; add some helpers to make tests more compact
https://github.com/open-spaced-repetition/fsrs-rs/pull/64
* Fix learning cards not being given memory state
* Require update to v3 scheduler
* Don't exclude single learning step when calculating memory state
* Use relearning step when learning steps unavailable
* Update docstring
* fix single_card_revlog_to_items (#2656)
* not need check the review_kind for unique_dates
* add email address to CONTRIBUTORS
* fix last first learn & keep early review
* cargo fmt
* cargo clippy --fix
* Add Jarrett to about screen
* Fix fsrs_memory_state being initialized to default in get_card()
* Set initial memory state on graduate
* Update to latest FSRS
* Fix experiment.log being empty
* Fix broken colpkg imports
Introduced by "Update FSRS card data when preset or weights change"
* Update memory state during (re)learning; use FSRS for graduating intervals
* Reset memory state when cards are manually rescheduled as new
* Add difficulty graph; hide eases when FSRS enabled
* Add retrievability graph
* Derive memory_state from revlog when it's missing and shouldn't be
---------
Co-authored-by: Jarrett Ye <jarrett.ye@outlook.com>
* Support searching for deck configs by name
* Integrate FSRS optimizer into Anki
* Hack in a rough implementation of evaluate_weights()
* Interrupt calculation if user closes dialog
* Fix interrupted error check
* log_loss/rmse
* Update to latest fsrs commit; add progress info to weight evaluation
* Fix progress not appearing when pretrain takes a while
* Update to latest commit
The approach in #2542 unfortunately introduced a regression, as whilst
it ensured that duplicate keys are removed when downgrading, it no longer
prevented the duplicates from being removed when converting to a legacy
Schema11 object. This resulted in things like backend.get_notetype_legacy()
returning duplicate keys, and could break syncing:
https://forums.ankiweb.net/t/windows-desktop-sync-error/33128
As syncing and schema11 object usage is quite common compared to downgrading,
the extra Value deserialization seemed a bit expensive, so I've switched
back to explicitly removing the problem keys. To ensure we don't forget to
add new keys in the future, I've added some new tests that should alert us
whenever a newly-added key is missing from the reserved list.
Also fix minilints declaring a stamp it wasn't creating. The same
approach is necessary with archives now too, as it no longer executes
under a standard "runner run".
For now, rustls is hard-coded - we could pass the desired TLS impl in
from the ./ninja script, but the runner is not recompiled frequently
anyway.
Workspace deps were introduced in Rust 1.64. They don't cover all the
cases that Hakari did unfortunately, but they are simpler to maintain,
and they avoid a couple of issues that Hakari had:
- It sometimes made updating dependencies harder due to the locked versions,
so you had to disable Hakari, do the updates, and then re-generate (
e.g. 943dddf28f)
- The current Hakari config was breaking AnkiDroid's build, as it was
stopping a cross-compile from functioning correctly.
- Dropped the protobuf extensions in favor of explicitly listing out
methods in both services if we want to implement both, as it's clearer.
- Move Service/Method wrappers into a separate crate that the various
clients can import, to easily get at the list of backend services and
their correct indices and comments.
I'd been thinking it might be useful for a future API service, but
I think that's better implemented with more codegen, so we have a
statically-typed interface.
* 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.
This is of limited usefulness at the moment, as it doesn't help consumers
of the public API.
Also removed detached comments from the included comments.
Provides better visibility into what the build is currently doing.
Motivated by slow node.js downloads making the build appear stuck.
You can test this out by running ./tools/install-n2 then building
normally. Please report any problems, and 'cargo uninstall n2' to get
back to the old behaviour. It works on Windows, but prints a new line
each second instead of redrawing the same area.
A couple of changes were required for compatibility:
- n2 doesn't resolve $variable names inside other variables, so the
resolution needs to be done by our build generator.
- Our inputs and outputs in build.ninja need to be listed in a deterministic
order to avoid unwanted rebuilds. I've made a few other tweaks so the
build file should now be fully-deterministic.
* Fix .no-reduce-motion missing from graphs spinner, and not being honored
* Begin migration from protobuf.js -> protobuf-es
Motivation:
- Protobuf-es has a nicer API: messages are represented as classes, and
fields which should exist are not marked as nullable.
- As it uses modules, only the proto messages we actually use get included
in our bundle output. Protobuf.js put everything in a namespace, which
prevented tree-shaking, and made it awkward to access inner messages.
- ./run after touching a proto file drops from about 8s to 6s on my machine. The tradeoff
is slower decoding/encoding (#2043), but that was mainly a concern for the
graphs page, and was unblocked by
37151213cd
Approach/notes:
- We generate the new protobuf-es interface in addition to existing
protobuf.js interface, so we can migrate a module at a time, starting
with the graphs module.
- rslib:proto now generates RPC methods for TS in addition to the Python
interface. The input-arg-unrolling behaviour of the Python generation is
not required here, as we declare the input arg as a PlainMessage<T>, which
marks it as requiring all fields to be provided.
- i64 is represented as bigint in protobuf-es. We were using a patch to
protobuf.js to get it to output Javascript numbers instead of long.js
types, but now that our supported browser versions support bigint, it's
probably worth biting the bullet and migrating to bigint use. Our IDs
fit comfortably within MAX_SAFE_INTEGER, but that may not hold for future
fields we add.
- Oneofs are handled differently in protobuf-es, and are going to need
some refactoring.
Other notable changes:
- Added a --mkdir arg to our build runner, so we can create a dir easily
during the build on Windows.
- Simplified the preference handling code, by wrapping the preferences
in an outer store, instead of a separate store for each individual
preference. This means a change to one preference will trigger a redraw
of all components that depend on the preference store, but the redrawing
is cheap after moving the data processing to Rust, and it makes the code
easier to follow.
- Drop async(Reactive).ts in favour of more explicit handling with await
blocks/updating.
- Renamed add_inputs_to_group() -> add_dependency(), and fixed it not adding
dependencies to parent groups. Renamed add() -> add_action() for clarity.
* Remove a couple of unused proto imports
* Migrate card info
* Migrate congrats, image occlusion, and tag editor
+ Fix imports for multi-word proto files.
* Migrate change-notetype
* Migrate deck options
* Bump target to es2020; simplify ts lib list
Have used caniuse.com to confirm Chromium 77, iOS 14.5 and the Chrome
on Android support the full es2017-es2020 features.
* Migrate import-csv
* Migrate i18n and fix missing output types in .js
* Migrate custom scheduling, and remove protobuf.js
To mostly maintain our old API contract, we make use of protobuf-es's
ability to convert to JSON, which follows the same format as protobuf.js
did. It doesn't cover all case: users who were previously changing the
variant of a type will need to update their code, as assigning to a new
variant no longer automatically removes the old one, which will cause an
error when we try to convert back from JSON. But I suspect the large majority
of users are adjusting the current variant rather than creating a new one,
and this saves us having to write proxy wrappers, so it seems like a
reasonable compromise.
One other change I made at the same time was to rename value->kind for
the oneofs in our custom study protos, as 'value' was easily confused
with the 'case/value' output that protobuf-es has.
With protobuf.js codegen removed, touching a proto file and invoking
./run drops from about 8s to 6s.
This closes#2043.
* Allow tree-shaking on protobuf types
* Display backend error messages in our ts alert()
* Make sourcemap generation opt-in for ts-run
Considerably slows down build, and not used most of the time.
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.
Due to the orphan rule, this meant removing our usages of impl ProtoStruct,
or converting them to a trait when they were used commonly.
rslib now directly references anki_proto and anki_i18n, instead of
'pub use'-ing them, and we can put the generated files back in OUT_DIR.
A couple of motivations for this:
- genbackend.py was somewhat messy, and difficult to change with the
lack of types. The mobile clients used it as a base for their generation,
so improving it will make life easier for them too, once they're ported.
- It will make it easier to write a .ts generator in the future
- We currently implement a bunch of helper methods on protobuf types
which don't allow us to compile the protobuf types until we compile
the Anki crate. If we change this in the future, we will be able to
do more of the compilation up-front.
We no longer need to record the services in the proto file, as we can
extract the service order from the compiled protos. Support for map types
has also been added.
* Migrate check_copyright to Rust
* Add a new lint to check accidental usages of /// in ts/svelte comments
* Fix a bunch of incorrect jdoc comments
* Move contributor check into minilints
Will allow users to detect the issue locally with './ninja check'
before pushing to CI.
* Make Cargo.toml consistent with other crates
* bump walkdir from 2.3.2 to 2.3.3
Build was failing on windows due to walkdir matching . as any folder starting with . instead of as the cwd.
* change name in CONTRIBUTORS
* change name back in CONTRIBUTORS
* Implement TTS using windows crate
* Use API calls instead of SSML
* Properly stop player in case of TTS error
* Add context to WindowsErrors
* Validate available voices
* Remove TTS text from synthesize error
* Limit maximum buffer size
* Make validation optional and list it in tts filter
* We no longer need the winrt module (dae)
* Use a separate request object so the meaning of the bool is clear (dae)
* Slightly shorten runtime error message (dae)
The default message appears to clip slightly.
* Alternate buffer implementation (dae)
* Use array instead of vec
* Drop the max buffer size to 128k (dae)
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.
* Remove deprecated `and_hms()`
* Update chrono
* Update licenses and fix script
* Remove deprecated Date struct
* Remove chrono pin
* Skip format check on .vscode
Was failing for no reason.
* Replace deprecated chrono functions
* Add cargo-deny to update-licenses & pin versions (dae)
* Remove time 0.1 dependency (dae)
We don't need to wait for chrono 0.5; it was provided behind a legacy
feature flag.
(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
