Update dependencies #6253
Update dependencies #6253
Conversation
There was a problem hiding this comment.
To make it easier to understand the consequences of this change:
dependencies = [
"param >=2.0,<3.0",
"numpy >=1.0",
"pyviz_comms >=2.1",
"panel >=1.0",
"colorcet",
"packaging",
"pandas >=0.20.0",
]
recommended = ["ipython >=5.4.0", "notebook", "matplotlib >=3", "bokeh >=3.1", "plotly >=4.0"]
Compared with what's on conda-forge:
run:
- python >=3.9
- param >=1.12.0,<3.0
- panel >=1.0
- numpy >=1.0
- pandas >=0.20.0
- matplotlib-base >=3.0
- bokeh >=3.1
- colorcet
- pyviz_comms >=0.7.4
- packaging
And defaults:
run:
- python
- colorcet
- numpy >=1.0
- packaging
- panel >=1.0
- pandas >=0.20.0
- param >=1.12.0,<3.0
- pyviz_comms >=0.7.4
- matplotlib-base >=3
- I'd vote for aligning the runtime dependencies defined on these three channels.
- I care less about the
pyvizchannel, in fact I hope we could retire it in some way, onlypyviz/label/devseems useful these days. We should certainly remove references to the-c pyvizchannel in the docs. Yet, removing/updating therecommendedpackages is a breaking change for users of this channel, so I'd motivate a change there by the goal of aligning the deps across the channels, rather than the intent to gain a few seconds on the CI.
Users are also affected by this. We pull in a big dependency that we don't use. I just noticed it is because of our CI. For alignment, I would propose these as required dependencies. bokeh >=3.1
colorcet
numpy >=1.0
packaging
pandas >=0.20.0
panel >=1.0
param >=2.0,<3.0
pyviz_comms >=2.1Maybe even bump the lower pin of pandas and numpy to something released in the last 3-5 years, or even remove them as they are so low it doesn't matter. (numpy 1.0 was released 25th October 2006, pandas 0.20.0 was released 5th May 2017) |
| "numpy >=1.21", | ||
| "packaging", | ||
| "pandas >=0.20.0", | ||
| "pandas >=1.3", |
There was a problem hiding this comment.
Pandas and numpy versions were changed to the first minor version with a micro release after Python 3.10 was released.
There was a problem hiding this comment.
Pandas and Numpy of these versions were released 3 years ago in June/July, this looks fine to me.
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #6253 +/- ##
==========================================
+ Coverage 88.48% 88.49% +0.01%
==========================================
Files 323 323
Lines 67640 67599 -41
==========================================
- Hits 59848 59825 -23
+ Misses 7792 7774 -18 ☔ View full report in Codecov by Sentry. |
| "numpy >=1.21", | ||
| "packaging", | ||
| "pandas >=0.20.0", | ||
| "pandas >=1.3", |
There was a problem hiding this comment.
Pandas and Numpy of these versions were released 3 years ago in June/July, this looks fine to me.
|
This pull request has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs. |
Updates required dependencies:
Numpy 1.0 --> 1.21
Pandas 0.20.0 --> 1.3
Bokeh (moved up from recommended)
Updates recommended dependencies:
Removed notebook and IPython.
Moved Bokeh up to required.