new python version support?

[brainfo]

The env is obsolete and hard to set up (install).
would be nice if this package gets updated as I and many users really like the semi-linear methods.

While it's possible to have a env for quasildr, I'd prefer at least to fix the python version for one research project.

[jzthree]

Can you tell me more about what's your environment and where did you encounter issues?
I tried it on a new conda env (in linux) and I also found pip install quasildr failed due to a nmslib installation problem, but I can install it after running pip install nmslib==2.0.6. I can update the dependencies if it solves the problem.

[brainfo]

Can you tell me more about what's your environment and where did you encounter issues? I tried it on a new conda env (in linux) and I also found pip install quasildr failed due to a nmslib installation problem, but I can install it after running pip install nmslib==2.0.6. I can update the dependencies if it solves the problem.

Hi,
The issue, in essence, came from numpy 1.18.1 is with python <= 3.8 and I am with higher version, python 3.12.8.

I can install quasidr with nmslib 1.8 installed. The issue comes from numpy deprecations such as np.int in quasildr/external/neighbors/. and numpy 1.18.1 is with python <= 3.8.

Thanks for help testing, supposedly I should try python 3.8 + quasildr 0.2.2 + nmslib 2.0.6 + numpy 1.18.1, right?

Followed are my previous trials:

In general I was using scanpy 1.10.3 but I could have separate scanpy 1.4.3 from pypi for quasidr's numpy version requirement. For now we can see the scanpy and scverse community hardly maintain the versions compatible with numpy 1.18.1 so there could be unresolved issues trying to work with this version. I personally have no experience in terms of if it worths to upgrade the numpy dependency. And with this,

the conflicts are from ipykernel, pip, jupyterlab, ... these basics are with python > 3.8; these are just bases, rather I don't use any of these set-ups, just have separate env for quasildr task script. ie, not incoparate quasildr in this project env.

[project]
authors = ["brainfo"]
channels = ["conda-forge"]
description = "pregmet snRNAseq study"
name = "10x"
platforms = ["linux-64"]
version = "0.1.0"

[tasks]

[dependencies]
ipykernel = ">=6.29.5,<7"
pip = ">=24.3.1,<25"
jupyterlab = ">=4.3.3,<5"
pixi-kernel = ">=0.5.2,<0.6"
openpyxl = ">=3.1.5,<4"

[feature.general.dependencies]
scanpy = "==1.10.3"
upsetplot = ">=0.9.0,<0.10"
python-igraph = ">=0.11.8,<0.12"
leidenalg = ">=0.10.2,<0.11"
anndata = ">=0.11.1,<0.12"

[feature.quasidr.dependencies]
plotly = ">=5.24.1,<6"
numpy = "==1.18.1"

[feature.quasidr.pypi-dependencies]
scanpy = "==1.4.3"
quasildr = "==0.2.2"
nmslib = "==1.8"

[environments]
general = ["general"]
quasildr = ["quasidr"]

Happy python's year.

[jzthree]

ok i think we can simplify the dependencies, scanpy is not a dependency unless you need to run the notebooks. externals.neighbors seem to need some update, but it's not needed if you use the default method that uses nmslib.

before we can finish maintaining the repo, the graphdr and structdr scripts https://github.com/jzhoulab/quasildr/blob/master/quasildr/graphdr.py and https://github.com/jzhoulab/quasildr/blob/master/quasildr/structdr.py are relatively self-contained and you can even just use that without installing

[brainfo]

ok i think we can simplify the dependencies, scanpy is not a dependency unless you need to run the notebooks. externals.neighbors seem to need some update, but it's not needed if you use the default method that uses nmslib.

before we can finish maintaining the repo, the graphdr and structdr scripts https://github.com/jzhoulab/quasildr/blob/master/quasildr/graphdr.py and https://github.com/jzhoulab/quasildr/blob/master/quasildr/structdr.py are relatively self-contained and you can even just use that without installing

OK yes.

I realized that I had an earlier project incl. quasildr. I checked that conda env, it's compatible with python 3.10.8 + numpy 1.23.5.

sorry for the confusion.