[cmccarthy1]

Hey!

We’re intending to add numpy 2.0 as soon as possible but don’t yet have a definitive timeline on when it will be released. Currently we’re researching the changes that are going to be needed for it to try provide a more hardened timeline.

[cmccarthy1]

Hey,

The Insights Core Libraries are shipped with PyKX directly but should only be executable within a Linux x86 environment. If you set

export PYKX_DEBUG_INSIGHTS_LIBRARIES=true

that will provide information on exactly why the library wasn’t loaded. If possible providing the output (redacted if needed) of

python -c "import pykx;print(pykx.util.debug_environment())"

will provide us with information which should help us debug further.

[cmccarthy1]

Hi,

This issue should be resolved with PyKX 2.4.0 (release notes at https://code.kx.com/pykx/2.4/release-notes/changelog.html#beta-features).

All the best,

Conor

code.kx.com

PyKX - PyKX

PyKX Documentation

[cmccarthy1]

Hi  it appears that the issue is in the logic on the server side `.u.upd` function, unfortunately I haven’t been able to replicate it locally with the equivalent table you are looking to use. Could you provide some information as to the expected inputs and logic of your upd function?

[cmccarthy1]

Hi ,

If possible could you elaborate a little more on your use case and the issue youre running into?

[cmccarthy1]

Hi,

This should work, could you provide the error that you’re seeing when attempting to initialise the connection?

There are multiple IPC connection types with PyKX, my suggestion would be that for TLS enabled IPC you make use of the SecureQConnection class documented here

If this does not work a minimal code example would be extremely helpful.

[cmccarthy1]

Hey,

There isn’t presently a way for multi-dimensional arrays to be returned in a different representation by default, if the returned data is a ‘rectangular’ data representation you can call np.stack on the data post processing to flatten it’s representation if needed

>>> import pykx as kx 
>>> import numpy as np 
>>> arr = kx.toq(np.random.rand(3,2))
>>> np.stack(arr.np()) 
array([[0.98669419, 0.47121857], 
[0.27329073, 0.15091775], 
[0.30553385, 0.32638803]])

 

[cmccarthy1]

Yes this can be done as follows

// Move into namespace d .ns fn1:{} fn2:{} // Move out of namespace d .

[cmccarthy1]

---

following the instruction guide using pip.

---

By this do you mean using conda? Installing Python from source is probably more accurate, if you installed Python via conda this issue is much less common. Our suggestion is that you install everything within a conda environment.

[cmccarthy1]

Hi,

Could you please provide some more information about you’re environment?

– What version of Python are you running?

– How did you install embedPy and JupyterQ?

The FAQ section of embedPy outlines some potential reasons for this problem https://code.kx.com/q/ml/embedpy/faq/#installing-embedpy-on-system-with-python-installed-from-source

[cmccarthy1]

Hi,

I don’t know if this is still an issue but, in step 7 your installation indicates that QHOME could not be found? can you run

export QHOME=/path/to/q/folder

and see if that fixes the issue?

All the best,

Conor

[cmccarthy1]

Apologies, I don’t see the information I’m expecting. Being more explicit could you start your q session running the following and provide the full output.

PYKX_DEBUG_INSIGHTS_LIBRARIES=true q pykx.q

[cmccarthy1]

Thanks. With the environment variable set (I see you have set it in the process) could you try run the \l pykx.q and return what the output in the q session is now?