Closed swainn closed 1 year ago
Hi!
Right now pdm-conda
uses pdm
solver and Conda to fetch candidates and install/remove packages, using Conda and libmamba as solvers would imply calling conda install
each time pdm
's solver breaks a dependency from the last conda solved env and could be really slow.
In my tests pdm
solver + Conda for fetching candidates ends up with the same solved packages as conda solver almost all the time, that said, I think adding conda/libmamba solver to pdm-conda
shoudn't be so difficult so I'll try to do it :).
@marco128 that makes sense. For some reason I was thinking that pdm-conda
was using the conda solver (which can be agonizingly slow). Thank you for the clarification. The pdm
solver is probably preferable, especially if it gives similar results. If you do end up implementing support, definitely include support for libmamba
, which is much more efficient than the classic conda solver.
@swainn v0.9.0 adds support to conda solver (now pdm-conda
always uses it to solve conda managed requirements). By default it uses classic solver but you can change it to libmamba
with conf solver
(as your example).
Great! Thank you.
I'd like to use pdm-conda with the libmamba solver (https://www.anaconda.com/blog/a-faster-conda-for-a-growing-community).
This is normally done by installing libmamba and then adding the
--solver libmamba
option to the install command like so:I think the ideal would be if the
solver
option could be specified in thepyproject.toml
: