AppTest format_func #8189
AppTest format_func #8189
Conversation
| @@ -347,6 +347,8 @@ def _multiselect( | |||
| multiselect_proto.value[:] = serde.serialize(widget_state.value) | |||
| multiselect_proto.set_value = True | |||
|
|
|||
| if ctx: | |||
There was a problem hiding this comment.
since I am not too familiar with the testing framework yet: does this only resolve to True in the testing-scenario? If yes, it might make sense to leave a comment or rename ctx to make it more obvious
There was a problem hiding this comment.
No. It is true any time you are running streamlit. The only time the ScriptRunContext is not available is if you are running a streamlit script directly via python (python my_app.py)
There was a problem hiding this comment.
Thanks! So this means that save_for_app_testing is executed also in non-testing flows, right? Will ctx.session_state[TESTING_KEY][k] always raise a key-error that is caught in non-testing executions then for every re-run? If yes, do we expect any performance issues because of it? My first reaction is that we would want to check for something like if IS_TEST_MODE: or similar
There was a problem hiding this comment.
It will be executed in all flows, and raise the error in non-testing executions. I don't expect it to be much of a performance cost, as exceptions are pretty similar to regular if branches in terms of perf in python.
There was a problem hiding this comment.
I think I would also feel slightly better if we could introduce some kind of global flag or config option to indicate that it is in app testing mode. We have something very similar for the development mode as well:
Having this flag could also be useful if we need to integrate other testing-related logic into Streamlit as well.
There was a problem hiding this comment.
Added a config option
…r in AppTest (#8264) ## Describe your changes Reverts a previous fix that made trigger values correct in AppTest but introduced an infinite recursion when `st.rerun()` is called conditionally on the value of a trigger widget. Instead, making trigger widgets return correct values in AppTest is handled by storing their values at runtime using the new `save_for_app_testing` functionality introduced by #8189 ## GitHub Issue Link (if applicable) #7768
passing arguments to CachedFunc.clear() todo update Reverts a previous fix that made trigger values correct in AppTest but introduced an infinite recursion when `st.rerun()` is called conditionally on the value of a trigger widget. Instead, making trigger widgets return correct values in AppTest is handled by storing their values at runtime using the new `save_for_app_testing` functionality introduced by streamlit#8189 GitHub Issue Link streamlit#8153
## Describe your changes Fixes issues with selecting items in widgets when `format_func` is used, by storing the `format_func` in session state during app tests and using it to convert the provided values into the ones stored by the protobuf. ## GitHub Issue Link (if applicable) streamlit#7679 and streamlit#8019 ## Testing Plan App tests added
…r in AppTest (streamlit#8264) ## Describe your changes Reverts a previous fix that made trigger values correct in AppTest but introduced an infinite recursion when `st.rerun()` is called conditionally on the value of a trigger widget. Instead, making trigger widgets return correct values in AppTest is handled by storing their values at runtime using the new `save_for_app_testing` functionality introduced by streamlit#8189 ## GitHub Issue Link (if applicable) streamlit#7768
Describe your changes
Fixes issues with selecting items in widgets when
format_funcis used, by storing theformat_funcin session state during app tests and using it to convert the provided values into the ones stored by the protobuf.GitHub Issue Link (if applicable)
#7679 and #8019
Testing Plan
App tests added
Contribution License Agreement
By submitting this pull request you agree that all contributions to this project are made under the Apache 2.0 license.