feat: update code to pandas 2 compatibility #132
Conversation
| description = "A projection package" | ||
| authors = ["Octopize <help@octopize.io>"] | ||
| license = "Apache-2.0" | ||
|
|
||
| [tool.poetry.dependencies] | ||
| python = ">=3.9,<3.12" | ||
| pandas = "^1.3" | ||
| pandas = ">=1.5.3" |
There was a problem hiding this comment.
This allows a package depending on saiph to work with pandas v2.
The rest of the code has been updated to work with pandas 2, and pandas 1.
There was a problem hiding this comment.
Moreover, this allows us to not break the API/core if we don't update to pandas straight away.
| @@ -249,6 +249,7 @@ def get_variable_correlation( | |||
| df_quali = pd.get_dummies( | |||
| df[model.original_categorical].astype("category"), | |||
| prefix_sep=DUMMIES_SEPARATOR, | |||
| dtype=np.uint8, | |||
There was a problem hiding this comment.
pandas v1 had default np.uint8, pandas v2 has it as bool.
The latter might break the missing values handling that add a _True and _False string suffix.
| series = pd.read_json( | ||
| data, orient="index", typ="series", dtype=json_dict["dtype"] | ||
| StringIO(data), orient="index", typ="series", dtype=json_dict["dtype"] | ||
| ) |
There was a problem hiding this comment.
Passing a string (data) to pd.read_json is deprecated in pandas v2/
| [[package]] | ||
| name = "pandas" | ||
| version = "1.5.3" | ||
| version = "2.2.2" |
There was a problem hiding this comment.
Here, pandas is at v2. poetry.lock is what is being used when developing locally. It is not put on pypi for the dependent package to install.
| description = "A projection package" | ||
| authors = ["Octopize <help@octopize.io>"] | ||
| license = "Apache-2.0" | ||
|
|
||
| [tool.poetry.dependencies] | ||
| python = ">=3.9,<3.12" | ||
| pandas = "^1.3" | ||
| pandas = ">=1.5.3" |
There was a problem hiding this comment.
Because saiph is a dependency and it's a pain when dependencies have hard constraints on another package.
Also, this allows us to update to pandas at a later stage in the API, while still benefiting from new features in saiph.
I've tested that this update works with the API and core with
pandas v1.5.3.Next step is to update pandas to v2 in API, but at least
saiphstill works withpandas 1.5.3if we don't update straight away, and is now compatible withpandas 2.2.2.