move py.typed to correct places #403
Conversation
wyattscarpenter
requested review from
rcypher-databricks,
yunbodeng-db,
andrefurlan-db,
jackyhu-db,
benc-db and
kravets-levko
as code owners
https://peps.python.org/pep-0561/ says 'For namespace packages (see PEP 420), the py.typed file should be in the submodules of the namespace, to avoid conflicts and for clarity.'. Previously, when I added the py.typed file to this project, #382 , I was unaware this was a namespace package (although, curiously, it seems I had done it right initially and then changed to the wrong way). As PEP 561 warns us, this does create conflicts; other libraries in the databricks namespace package (such as, in my case, databricks-vectorsearch) are then treated as though they are typed, which they are not. This commit moves the py.typed file to the correct places, the submodule folders, fixing that problem. Signed-off-by: wyattscarpenter <[email protected]>
I think this might fix the CI code-quality checks failure, but unfortunately I can't replicate that failure locally and the error message is unhelpful Signed-off-by: wyattscarpenter <[email protected]>
…led (no mypy cache directory)'; see python/mypy#10768 (comment) Signed-off-by: wyattscarpenter <[email protected]>
Signed-off-by: wyattscarpenter <[email protected]>
|
Hi @wyattscarpenter! CI now reports this error: |
Fixes the problem by cding and supplying a flag to mypy (that mypy needs this flag is seemingly fixed/changed in later versions of mypy; but that's another pr altogether...). Also fixes a type error that was somehow in the arguments of the program (?!) (I guess this is because you guys are still using implicit optional) --------- Signed-off-by: wyattscarpenter <[email protected]>
|
@kravets-levko the "Source file found twice under different module names" error is, ultimately, caused by mypy not dealing with the namespaces correctly. It's either a historical bug or behavior they decided to change later, since it actually works fine without workarounds or flags on the latest version of mypy: https://github.com/wyattscarpenter/databricks-sql-python/tree/upgrade_mypy (I'm thinking I'll do a pr from that branch later, since the version upgrade touches a number of other files and therefore might be inconvenient to review at the same time as this pr; but I could also make squash it into a commit in this pr if you would prefer.) |
Return the old result_links default, make the type optional, & I'm pretty sure the original problem is that add_file_links can't take a None, so these statements should be in the body of the if-statement that ensures it is not None Signed-off-by: wyattscarpenter <[email protected]>
"self.download_manager is unconditionally used later, so must be created. Looks this part of code is totally not covered with tests 🤔" Co-authored-by: Levko Kravets <[email protected]> Signed-off-by: wyattscarpenter <[email protected]>
|
Thank you @wyattscarpenter! |
|
Thanks! |
https://peps.python.org/pep-0561/ says 'For namespace packages (see PEP 420), the py.typed file should be in the submodules of the namespace, to avoid conflicts and for clarity.'. Previously, when I added the py.typed file to this project (#382) I was unaware this was a namespace package, so I added the py.typed file to the top-level folder (the namespace). As PEP 561 warns us, this does create conflicts; other libraries in the databricks namespace package (such as, in my case, databricks-vectorsearch) are then treated as though they are typed, which they are not, leading to unnecessary type errors in those other packages. This commit moves the py.typed file to the correct places, the submodule folders, fixing that problem.