Migrate all tests to use Flask #111
Conversation
Tests were failing on mac osx sproadically. Switching to a different web server seems to have solved the issuse. Added vscode files for devleopment.
pyproject.toml
Outdated
| @@ -32,10 +32,21 @@ dependencies = [ | |||
| [tool.hatch.version] | |||
| path = "src/kagglehub/__init__.py" | |||
|
|
|||
| [project.optional-dependencies] | |||
| dev = [ | |||
There was a problem hiding this comment.
As discussed yesterday, is this block still needed? Probably best to standardize on always using hatch.
There was a problem hiding this comment.
sg. Hatch works natively, I will remove this.
|
|
||
|
|
||
| @contextmanager | ||
| def create_env() -> Generator[Any, Any, Any]: |
There was a problem hiding this comment.
Why is the type Generator[Any, Any, Any]? At line #78, you don't yield anything.
There was a problem hiding this comment.
This was the suggested type from mypy. See attached screenshot.
Generators can be used for bidirectional communication. From my understanding, the way the contextmanager works is by using generators to stop execution at the first yield statement. After, the with statement body of the users function completes. Then, the program resumes execution from yield statement.
01 @contextmanager
02 def my_custom_open():
03 try:
04 f = open("a/b/c/d.txt", "wb")
05 yield f
06. finally:
07 f.close()
08
09 def foo():
10 with my_custom_open() as f:
11 chunk1 = f.read(1024)
12 chunk2 = f.read(1024)For ex, in the above example, if the user executes the foo function on line 09, the code execution for my_custom_open is advanced to line 05. After, line 11 and 12 are resumed and completed. Finally, the execution resumes at line 05 to completion.
There was a problem hiding this comment.
Sorry, I wasn't clear. I am familiar with how contextmanager works, I setup these test harness methods.
My suggestion was around the exact type for Generator.
You have to specify three types: the yield, send and return type.
Given you yield Nothing, the yield type should be None instead of Any. Same for the send and return typs, we don't send or return values into the generator.
| class TestColabCacheModelDownload(BaseTestCase): | ||
| @classmethod | ||
| def setUpClass(cls): # noqa: ANN102 | ||
| # TODO: the colab environment has flaky behavior if set in test. It also can't be set in the `tests/__init__.py` |
There was a problem hiding this comment.
What's flaky? Is it the server not responsive or not using the right address? Given it is simply calling a local http server, I wouldn't expect any flakiness.
There was a problem hiding this comment.
I documented this issue in this b/337257114
| @app.route("/api/v1/models/<org_slug>/<model_slug>/create/instance", methods=["POST"]) | ||
| def model_instance_create_instance(org_slug: str, model_slug: str) -> ResponseReturnValue: | ||
| post_data = request.get_json() | ||
| # TODO: error here |
There was a problem hiding this comment.
I added this while refactoring. Forgot to take this out.
b/337248131