* pyi_generator: avoid fully qualified names for reflex. types
The previous method added quite a large line diff to the pyi files, so for now,
lets just add the `Breakpoints` import to the default set along with Var,
EventHandler, Style and others.
* pyi_generator: fix generated files using ruff
also generate updated pyi files in reflex/experimental namespace
* remove black from dev dependencies
* poetry.lock: relock deps
* Update all pyi files with new formatting
* pre-commit fixes
* pyi_generator: fix __init__.pyi files again
* implement basic behavior to handle both radix and css
* add all components in theme, AND change pyi to include path of reflex types
* order type annotations in pyi_generator (#3585)
* order type annotations in pyi_generator
* run pyi_generator
* classvars should not be backend vars (#3578)
* classvars should not be backend vars
* cleanup RESERVED_BACKEND_VAR_NAMES
* order type annotations in pyi_generator (#3585)
* order type annotations in pyi_generator
* run pyi_generator
* do not validate non-cached var deps (#3576)
* do not validate non-cached var deps
* further improve Exceptions for misconfigured var deps
* order type annotations in pyi_generator (#3585)
* order type annotations in pyi_generator
* run pyi_generator
* fix git merge remains
* define function globally
* use more modern annotations
* only trigger when direct value and not state
* simplify unions
* accidentally removed avatar oop
---------
Co-authored-by: benedikt-bartscher <31854409+benedikt-bartscher@users.noreply.github.com>
* add computed backend vars
* finish computed backend vars, add tests
* fix token for AppHarness with redis state manager
* fix timing issues
* add unit tests for computed backend vars
* automagically mark cvs with _ prefix as backend var
* fully migrate backend computed vars
* rename is_backend_variable to is_backend_base_variable
* add integration test for implicit backend cv, adjust comments
* replace expensive backend var check at runtime
* keep stuff together
* simplify backend var check method, consistent naming, improve test typing
* fix: do not convert properties to cvs
* add test for property
* fix cached_properties with _ prefix in state cls
* Automatic authentication for backend on Github Codespaces
When running reflex on Github codespaces, the port forwarding mechanism
requires authentication, which happens automatically when first accessing the
port via HTTPS; however since the backend connects over the WSS protocol
instead, it gets an access error with no way to redirect to Github's
authentication servers to get the port open.
This PR adds an automatic redirection mechanism to a backend route when there
is a connection error accessing the frontend. After the backend route loads, it
redirects back to the frontend, but now it can connect to the backend via
websocket because the port forward is authenticated.
* manually update .pyi file 🫨
* Dont escape backticks in JS string interpolation
* add unit tests
* Fix darglint
* add a note to re-visit after new Var API is implemented
* tests should have a good meaning
* upgrade to latest ruff
* try to fix dep review
* try to fix dep review (2)
* upgrade black
* upgrade black (2)
* update allowed dependencies
* update allowed dependencies (2)
* update allowed dependencies (3)
* wait between interim and final in yield test
* remove previous commit, increase delay between yield
* forgot to save on the time.sleep(1) removal
* fix integration (maybe?)
* fix pyi?
* what even is going on
* what is realityi?
* test another fix for app harness
* try to wait even longer?
* force uvloop to be optional
* downpin fastapi < 0.111, remove changes to test
In pydantic 1.10.16 (hopefully the last version), they added the `pydantic.v1`
namespace to make transitioning to the v2 package easier without changing code.
However, Reflex was depending on pydantic v1 NOT having `pydantic.v1` to skip
patching names from `v1` into the top level `pydantic` (normally we do this to
force sqlmodel to use v1, even when the v2 package is installed).
Unbreak CI
* Make client_state work without global refs (local only)
* client_state: if the default is str, mark _var_is_string=True
Ensure that a string default is not rendered literally
* add `to_int` as a Var operation
* Allow an event handler lambda to return a Var in some cases
If an lambda is passed to an event trigger and it returns a single Var, then
treat it like the Var was directly passed for the event trigger.
This allows ClientState.set_var to be used within a lambda.
* wip type transforming serializers
* old python sucks
* typing fixups
* Expose the `to` parameter on `rx.serializer` for type conversion
Serializers can also return a tuple of `(serialized_value, type)`, if both ways
are specified, then the returned value MUST match the `to` parameter.
When initializing a new rx.Var, if `_var_is_string` is not specified and the serializer returns a `str` type, then mark `_var_is_string=True` to indicate that the Var should be treated like a string literal.
Include datetime, color, types, and paths as "serializing to str" type.
Avoid other changes at this point to reduce fallout from this change:
Notably, the `serialize_str` function does NOT cast to `str`, which
would cause existing code to treat all Var initialized with a str as a
str literal even though this was NOT the default before.
Update test cases to accomodate these changes.
* Raise deprecation warning for rx.Var.create with string literal
In the future, we will treat strings as string literals in the JS code. To get
a Var that is not treated like a string, pass _var_is_string=False.
This will allow our serializers to automatically identify cast string literals
with less special cases (and the special cases need to be explicitly
identified).
* Add test case for mismatched serialized types
* fix old python
* Remove serializer returning a tuple feature
Simplify the logic; instead of making a wrapper function that returns
a tuple, just save the type conversions in a separate global.
* Reset the LRU cache when adding new serializers
---------
Co-authored-by: Masen Furer <m_github@0x26.net>
* Addresses an issue with pydantic v2 models as Vars
It looks like there's an issue with state vars
which are pydantic v2 models... Here's a
reproducible test case:
```python
import reflex as rx
from pydantic import BaseModel
from reflex.utils.serializers import serializer
class User(BaseModel):
has_image: bool = False
@serializer
def serialize_user(user: User) -> dict:
return user.dict()
class State(rx.State):
user: User = None
def index() -> rx.Component:
return rx.container(
rx.cond(State.user,
rx.text(State.user.has_image),
rx.text("No user"))
)
app = rx.App()
app.add_page(index)
```
This app works only with pydantic <2 installed:
```bash
reflex-test $ reflex run
...
AttributeError: 'FieldInfo' object has no attribute 'outer_type_'
reflex-test $ pip install pydantic==1.10.15
─────────────────────────────────── Starting Reflex App ───────────────────────────────────
Compiling: ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 13/13 0:00:00
─────────────────────────────────────── App Running ───────────────────────────────────────
App running at: http://localhost:3000
```
Looks like this is caused by `outer_type_` no
[longer existing][1] in pydantic v2. I'm guessing
this was introduced back in [v0.4.6][2].
1: https://github.com/pydantic/pydantic/discussions/7217
2: 86526cba51
This change explicitly ignores pydantic v2 models in
`get_attribute_access_type`, rather than trying to treat
them as v1 models.
* ruff formatting
---------
Co-authored-by: Masen Furer <m_github@0x26.net>
* [REF-2977] [REF-2982] Merge layout prop and automatic darkmode
* Expose `template` prop in rx.plotly
* Set default `template` to a color_mode_cond that responds to dark mode
* Merge user-provided `layout` and `template` props into the serialized
plot data
* pyi_generator: avoid affecting the HTML Template component
* Raise warning when importing rx.plotly without plotly installed
* Remove PlotlyLib component and consolidate imports [fixup]
* pyi_generator: do not generate kwargs for event trigger props
event triggers are handled separately
* Implement event handlers for Plotly
* py38 compat: from __future__ import annotations
* init fix
* Update reflex/components/plotly/plotly.py
Co-authored-by: Masen Furer <m_github@0x26.net>
* plotly: treat `data` as a `dict`-type Var in _render
this allows the data to be passed directly as a figure or from a state var
* removed width height prop as they are no longer needed
* updated
* reverted some of the changes
* fixed unit tests
* regen pyi
---------
Co-authored-by: Hongyu Yao <hongyuyao@hongyus-mbp-3.lan>
Co-authored-by: Masen Furer <m_github@0x26.net>
Co-authored-by: Hongyu Yao <hongyuyao@Hongyus-MacBook-Pro-3.local>
* Improve import times
* add lazy loading to rx.el
* add lazy loading to reflex core components
* minor refactor
* Get imports working with reflex web
* get imports to work with all reflex examples
* refactor to define imports only in the root.
* lint
* deadcode remove
* update poetry deps
* unit tests fix
* app_harness fix
* app_harness fix
* pyi file generate
* pyi file generate
* sort pyi order
* fix pyi
* fix docker ci
* rework pyi-generator
* generate pyi for __init__ files
* test pyright
* test pyright ci
* partial pyright fix
* more pyright fix
* pyright fix
* fix pyi_generator
* add rx.serializer and others
* add future annotation import which fixes container CI, then also load recharts lazily
* add new pyi files
* pyright fix
* minor fixes for reflex-web and flexdown
* forward references for py38
* ruff fix
* pyi fix
* unit tests fix
* reduce coverage to 68%
* reduce coverage to 67%
* reduce coverage to 66%as a workaround to coverage's rounding issue
* reduce coverage to 66%as a workaround to coverage's rounding issue
* exclude lazy_loader dependency review checks.
* its lazy-loader
* Add docstrings and regenerate pyi files
* add link
* address Pr comments
* CI fix
* partially address PR comments.
* edit docstrings and fix integration tests
* fix typo in docstring
* pyi fix
* Get `action` and `cancel` working for rx.toast
Respect defaults set in ToastProvider toast_options when firing a toast with
it's own ToastProps set.
* Update reflex/components/sonner/toast.py
Co-authored-by: Thomas Brandého <thomas.brandeho@gmail.com>
* Move queueEvent formatting into rx.utils.format module
Implement on_auto_close and on_dismiss callbacks inside ToastProps
* Update rx.call_script to use new format.format_queue_events
Replace duplicate logic in rx.call_script for handling the callback function.
* Move PropsBase to reflex.components.props
This base class will be exposed via rx._x.PropsBase and can be shared by other
wrapped components that need to pass a JS object full of extra props.
---------
Co-authored-by: Thomas Brandého <thomas.brandeho@gmail.com>
* Throw Warning for Projects Created in OneDrive on Windows
* precommit
* remove dead code
* REFLEX_USE_NPM escape hatch to opt out of bun
In some unsupported environments, we need to just not use bun. Further
investigation needed.
---------
Co-authored-by: Masen Furer <m_github@0x26.net>
* upgrade to latest pip for in_docker_test_script.sh
* Bump gunicorn to 22.0.0 (security)
Changelog: https://docs.gunicorn.org/en/stable/news.html#id1
use utime to notify workers liveness
migrate setup to pyproject.toml
fix numerous security vulnerabilities in HTTP parser (closing some request smuggling vectors)
parsing additional requests is no longer attempted past unsupported request framing
on HTTP versions < 1.1 support for chunked transfer is refused (only used in exploits)
requests conflicting configured or passed SCRIPT_NAME now produce a verbose error
Trailer fields are no longer inspected for headers indicating secure scheme
support Python 3.12
** Breaking changes **
minimum version is Python 3.7
the limitations on valid characters in the HTTP method have been bounded to Internet Standards
requests specifying unsupported transfer coding (order) are refused by default (rare)
HTTP methods are no longer casefolded by default (IANA method registry contains none affected)
HTTP methods containing the number sign (#) are no longer accepted by default (rare)
HTTP versions < 1.0 or >= 2.0 are no longer accepted by default (rare, only HTTP/1.1 is supported)
HTTP versions consisting of multiple digits or containing a prefix/suffix are no longer accepted
HTTP header field names Gunicorn cannot safely map to variables are silently dropped, as in other software
HTTP headers with empty field name are refused by default (no legitimate use cases, used in exploits)
requests with both Transfer-Encoding and Content-Length are refused by default (such a message might indicate an attempt to perform request smuggling)
empty transfer codings are no longer permitted (reportedly seen with really old & broken proxies)
** SECURITY **
fix CVE-2024-1135
* Remove TYPE_CHECKING guard for pydantic v1 imports
Retain TYPE_CHECKING guard in v1 fallback to force pyright into pydantic.v1 namespace
* Run unit tests with pydantic v1 now that v2 is installed via poetry
