Specify newer versions of numpy be used #32
Conversation
Signed-off-by: David Black <[email protected]>
…n 3.10 & a specific version of numpy being pulled in by pyarrow. Signed-off-by: David Black <[email protected]>
This reverts commit c8ab431.
Signed-off-by: David Black <[email protected]>
|
This follows on from #31 . |
|
Test execution can be seen at https://github.com/dbaxa/databricks-sql-python/runs/7837710866?check_suite_focus=true (note that installing dependencies in python 3.10 is considerably faster after this change than it was for #31 -> https://github.com/dbaxa/databricks-sql-python/runs/7837393055?check_suite_focus=true ) |
…s-of-numpy-be-used # Conflicts: # poetry.lock # pyproject.toml Signed-off-by: David Black <[email protected]>
| @@ -11,7 +11,8 @@ include = ["CHANGELOG.md"] | |||
| [tool.poetry.dependencies] | |||
| python = "^3.7.1" | |||
| thrift = "^0.13.0" | |||
| pandas = "^1.3.0" | |||
| pandas = "^1.3.5" | |||
| numpy = [{version="^1.23.2", python=">=3.8"}, {version="^1.21.6", python="<3.8"}] | |||
There was a problem hiding this comment.
I am not sure I have done what I want here. I want to specify "use at least this version of numpy for python >= 3.8". But this ends up as
{':python_version < "3.8"': ['numpy>=1.21.6,<2.0.0'],
':python_version >= "3.8"': ['numpy>=1.23.2,<2.0.0']}
Which is likely fine but pandas 1.3.5 for example ends up with
numpy>=1.17.3; platform_machine!='aarch64' and platform_machine!='arm64' and python_version<'3.10'
numpy>=1.19.2; platform_machine=='aarch64' and python_version<'3.10'
numpy>=1.20.0; platform_machine=='arm64' and python_version<'3.10'
numpy>=1.21.0; python_version>='3.10'
There was a problem hiding this comment.
This can be addressed by using >=$version, but I still feel that obtaining numpy version information via pandas is better.
|
I am going to close this as it is likely better to use a minimum version of |
While pandas makes use of https://github.com/scipy/oldest-supported-numpy/blob/main/setup.cfg , we need not hold the version of
numpyin use back. By requiring a recent release - there should be no issues with regards to architectures such as arm64 on darwin or loongarch64. For python 3.7, we can use 1.21.6.