Data Science

Machine learning & data science for beginners and experts alike.
JeffA
Alteryx
Alteryx

Q: Where can I download the tool?

A: The tool is currently available as an attachment to this blog post or through this link.

 

Q: What version of Alteryx Designer do I need to run the Jupyter Flow tool?

A: Jupyter Flow is compatible with Alteryx Designer 2020.4 and up.

 

Q: Is there a step by step guide to help me get started with Jupyter Flow?

A: Yes. The guide is available here.

 

Q: Where are the help pages?

A: The help pages are always accessible from any tool's configuration panel. However, here is the link!

 

Q: What other Jupyter kernels are supported?

A: Any kernel supported by Jupyter should work. However, only Python and Julia have been tested and confirmed to work. Keep in mind that for now Python is the only language which can read and write to Alteryx.

 

Q: Can I use Conda environments instead of pip VENVs?

A: Yes, but... Ensure that all of the packages you want installed to your tool are in the `site-packages` folder you provide. If Conda installs a package outside of the folder you provide, it will not be picked up by Jupyter Flow. Also, there is sometimes a "DLL load failed" error, which seems to be a symptom more common in Conda environments. If you run into this error, you can try creating a new Conda environment and installing everything via pip and ensuring pywin32==300 is installed instead of 301. If that does not work, the simplest known solution is to use pip venvs to build your Jupyter Flow environments.

 

Q: I have a DLL load failed error, how do I fix it?

A: There seems to be an issue with pywin32 version 301. Try installing pywin32==300 in your virtual environment. In many cases, this will fix the issue. If it does not, and you're using Conda for environment management, see the question about using conda environments instead of pip venvs. The pywin32 issue is known and can be seen on GitHub here.

 

Q: I have a NoSuchKernel error, how do I fix it?

A: See the current workaround for this issue here. Hopefully we will have a permanent fix for this issue soon.

 

Q: I can't seem to debug my _post_processed notebook. Why not?

A: In order to debug your _post_processed notebook, you must enable the back up data cache advanced option in the Jupyter Flow config pane before running the notebook, and you must have pandas and pyarrow installed in the environment from which you're running your _post_processed notebook.

 

Q: Do I have to install Jupyter in my Jupyter Flow environment in order to use Jupyter Flow?

A: No. Jupyter is only required to open, modify, and and run your Jupyter notebooks. Because interacting with your notebooks requires all of your dependencies to be installed in the environment running Jupyter, it is often faster to install Jupyter in your Jupyter Flow environment and run Jupyter from there. However, ideally you would have one environment for modifying and debugging your Jupyter Flow notebooks and one environment for running them. Similar to development vs production environments. Your development environment should also contain pyarrow and pandas for debugging purposes.

 

Q: Why is my tool taking so long to build the environment?

A: The time it takes to build a Jupyter Flow environment varies greatly depending on how many packages are in your packages folder. It is advised that you specify an environment containing only the packages you need, as opposed to the common kitchen sink environments we all have lying around.

 

Q: I have a file does not exist error when I run my notebook. What's going on?

A: A common cause of a file does not exist error is that long paths are not enabled on Windows. Follow the Microsft Docs to enable long paths and try again.

 

Q: When I input an empty folder into the "Packages" field, sometimes my notebook still runs, even though it has dependencies. Why?

A: Default packages are installed in Jupyter Flow tool. This includes pandas, jupyter, pyarrow, and others. Therefore, if your packages folder is empty, you will still have the base packages available. All of the default packages are visible under "Tools\JupyterFlow_venv\Lib\site-packages," where "Tools" is the directory in which Jupyter Flow is installed on your machine. When your packages folder does contain one of these packages, the version in your packages folder overrides the default version.

 

Q: What's with the lunar rover tool icon?

A: The tool leverages Jupyter from a distance, as if it were on one of Jupiter's moons, roving around. Plus the tool's original codename was "Moonbuggy."

 

 

Banner image by Beate Bachman