How to manage shared code

[NVolcz]

I have a bunch of Azure Python Function Apps that share some code and is not sure how I am sure how this is supposed work. The shared code is proprietary and cannot be uploaded to PyPi. My issue is similar to: Azure/azure-functions-python-worker#219

This is how I would like the folder structure to look like:

AzureFunctions
|_ FunctionApp1
    |_ requirements.txt
    |_ HttpTrigger1
        |_ __init__.py
        |_ function.json
|_FunctionApp2
    |_ requirements.txt
    |_ HttpTrigger1
        |_ __init__.py
        |_ function.json
|_ SharedCode
    |_ setup.py
    |_ requirements.txt
    |_ src
        |_ __init__.py
        |_ common1.py
        |_ common2.py
    |_ tests
        |_ test1.py
        |_ test2.py

I have tried to use the file: protocol in FunctionApp1 and FunctionApp2's requirements.txt:
file:../SharedCode
Most tooling understands this except for pip freeze. Skipping the file: part also seems works well with other tooling but neither does not work well with func. Remote build and --build-native-deps cannot build these function apps since they do not get access to the SharedCode folder, only the functions individual folders. I have therefor tried to build locally. It seems like the first problem with this lies with the packapp script: https://github.com/Azure/azure-functions-core-tools/blob/dev/tools/python/packapp/__main__.py#L73-L79
This SO thread seems related: https://stackoverflow.com/questions/11147667/is-there-a-way-to-list-pip-dependencies-requirements
pip download outputs this:

Processing <PATH>/AzureFunctions/SharedCode
    Link is a directory, ignoring download_dir

What should I do?

[zsolt-halo]

Did you take a look at this?

It works for me. Not very clear as the comments say so but once you figure out the example it will work.

 __app__
 | - my_first_function
 | | - __init__.py
 | | - function.json
 | | - example.py
 | - my_second_function
 | | - __init__.py
 | | - function.json
 | - shared_code
 | | - my_first_helper_function.py
 | | - my_second_helper_function.py
 | - host.json
 | - requirements.txt
 | - Dockerfile
 tests

and the following imports will work from within any of the __init.py__ -s

from . import example #(explicit relative)  # importing from local folder
from ..shared_code import my_first_helper_function #(explicit relative) # 1 folder up
from __app__ import shared_code #(absolute) # using the special function app var __app__ which means the root of your project (the folder which contains your whole azure functions project)
import __app__.shared_code #(absolute) # same as above but exporting everything from shared_code

[NVolcz]

The difference between function apps and functions is confusing to say the least (and there are no such things on the other platforms). The above example "works" but is not ideal. We wanted separate function apps for separate managed identities. Also, how many functions can fit in a functions app? We currently have four but are planning to add many more and without the possibility to share code between functions apps all with end up in the same function app. In an ideal world I would not like to manage the function app at all.

[maximeLeurent]

Hello,

I faced the same problem of folder structure.
I managed to make it work by using simlink and -r inside requirements.txt

Here the Folder Structure I use. (Almost what you propose but without test in SharedCode, and setup.py)

AzureFunctions
|_ FunctionApp1
    |_ requirements.txt
    | SharedCode(simlink)
    |_ HttpTrigger1
        |_ __init__.py
        |_ function.json
|_FunctionApp2
    |_ requirements.txt
    | SharedCode(simlink)
    |_ HttpTrigger1
        |_ __init__.py
        |_ function.json
|_ SharedCode
    |_ requirements.txt
    |_ __init__.py
    |_ common1.py
    |_ common2.py

I created the simlink with :

cd FunctionApp1
ln -s ../SharedCode/ SharedCode

requirements.txt in FunctionApp1 look like

lib_for_FunctionApp1
-r ./SharedCode/requirements.txt

[apawast]

@NVolcz are you still experiencing this issue?

[NVolcz]

We gave up on trying to use Python with Azure Functions after not receiving any help on this issue.
Have you changed anything?