Awesome thanks

@User can you DM me your full stack trace / error log?

+ steps to reproduce

guys.. I just notice that a custom data catalog I built this week, only works in my host machine but inside the docker the kedro cli cant find the project_root.extras.datasets.mycustom_dataset.MyCustomDataset I am extendind the AbstractDataSet with _load, _describe and _save In catalogs file, I set a path like this project_root.extras.datasets.mycustom_dataset.MyCustomDataset Am I missing something?

It will be a working directory issue, the class path isn't accessible from wherever you are calling Kedro from. If you can run

pwd

from inside the container that may help

o dont get it.. pwd inside docker give the root of my project.. and kedro run in my host machine is on root of my project either :/

behind the scenes all we're doing is

importlib.import_module(project_root.extras.datasets.mycustom_dataset.MyCustomDataset)

so that's all we need to get working outside of Kedro. Do you have

__init__.py

files in every directory?

yep

src does not

do i need to register the extras.datasets in somefile? i followed the doc, I watched the dataengineerone videos about custom data catalog, he is doing exactly the samething i am.. very confusing..

Let me DM you

@User can you start a repl in your docker container anf try importing it there. The solution may become more aparant, and easier to iterate on.

python
from project_root.extras.datasets.mycustom_dataset import MyCustomDataset

of course

give me a minute

i am entering a meeting right now

sorry

i am back.. let me test it

wow!! hahahaha.. i forgot to add trino dependency in my requirements.. and from repl o could see exactly this error! thank you @User @User i should be able to see that error in kedro run right? the error i was facing was so generic :/

^ Good point - let me look into that

Thanks! Fix is in scope for next sprint

Quick question, while I know its a bit different, is it possible to load data from the catalog into a pytest unit test?

@WolVez very possible - however many people avoid using real data in unit tests

The best way to do so is to either create a session or DataCatalog object as a fixture

Hi everyone. I have a very basic question regarding the installation and starting a new Kedro project. Does

pip install kedro

needs to be executed for every new Kedro project within a dedicated Python virtual environment? In other words, for every new project, are the following steps correct to get started? 1 - Create a new Python virtual environment. For instance, with

conda

, with the command:

conda create --name <environment_name> python=3.7 -y

2 - Install Kedro (i.e. execute

pip install kedro

) 3 - Create a new Kedro project (e.g. without starter, with the command

kedro new

) I hope my question makes sense. Thank you in advance!

Hi @julien-e yes this is the right way to do things. New venv new kedro - it makes things easier to manage long term when there are different python or kedro versions.

So your process is correct

Between steps 1 and 2 remember to activate your new env

Brilliant, thank you so much for your super quick answer @User 🙇

I need some public kedro project with a git history to try

kedro-diff

on anyone who can share one, it would be greatly appreciated.