Cannot run tests on nightly

How to reproduce the behaviour

Download spacy nightly and try to run the tests after installing the requirements. Pytest gives the following complaint.

_______________________________ ERROR collecting spacy/tests/regression/test_issue5230.py ________________________________
spacy/tests/regression/test_issue5230.py:89: in <module>
    [nlp(), vectors(), custom_pipe(), tagger(), entity_linker()],
spacy/tests/regression/test_issue5230.py:61: in tagger
    tagger = nlp.add_pipe("tagger")
spacy/language.py:687: in add_pipe
    pipe_component = self.create_pipe(
spacy/language.py:585: in create_pipe
    resolved, filled = registry.resolve(cfg, validate=validate)
../../anaconda3/envs/project/lib/python3.8/site-packages/thinc/config.py:542: in resolve
    filled, _, resolved = cls._fill(
../../anaconda3/envs/project/lib/python3.8/site-packages/thinc/config.py:631: in _fill
    filled[key], validation[v_key], final[key] = cls._fill(
../../anaconda3/envs/project/lib/python3.8/site-packages/thinc/config.py:692: in _fill
    result = schema.parse_obj(validation)
pydantic/main.py:455: in pydantic.main.BaseModel.parse_obj
    ???
pydantic/main.py:344: in pydantic.main.BaseModel.__init__
    ???
pydantic/main.py:874: in pydantic.main.validate_model
    ???
E   pydantic.errors.ConfigError: field "nlp" not yet prepared so type is still a ForwardRef, you might need to call ArgModel.update_forward_refs().
================================================ short test summary info =================================================
ERROR spacy/tests/regression/test_issue5230.py - pydantic.errors.ConfigError: field "nlp" not yet prepared so type is s...
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Interrupted: 1 error during collection !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Your Environment

• spaCy version: 3.0.0a7
• Platform: Linux-4.15.0-112-generic-x86_64-with-glibc2.10
• Python version: 3.8.5

➜  spaCy git:(5fdd6b460) pip show thinc
Name: thinc
Version: 8.0.0a27
Summary: A refreshing functional take on deep learning, compatible with your favorite libraries
Home-page: https://github.com/explosion/thinc
Author: Explosion
Author-email: [e-mail]
License: MIT
Location: /home/user/anaconda3/envs/project/lib/python3.8/site-packages
Requires: numpy, murmurhash, srsly, setuptools, preshed, wasabi, pydantic, catalogue, cymem
Required-by: spacy-nightly

[tiangolo]

Thanks for the patience @svlandeg and @forgetso !

First, yeah, I have seen that error several times in other projects using Pydantic 😬 😅

Nevertheless, I started from a clean slate and I couldn’t reproduce the error in spaCy 😕

…But it took me a bit of work to actually start from a “clean slate”, and I think the error could be coming from one of the several “caches” interacting in this.

@forgetso could you try this?:

// Make sure to pull the latest changes
$ git pull upstream develop

// Clean any built Cython modules/binaries
$ make clean

// Create a new venv (just to be sure)
$ python3.8 -m venv env-2020-10-15

// Activate it
$ source env-2020-10-15/bin/activate

// Update pip and the building requirements
$ python -m pip install --upgrade pip setuptools wheel

// Install the spaCy requirements, making sure to upgrade
// The upgrade here is important even in a new environment as pip could use its local cache and use an old version of some package
$ python -m pip install --upgrade -r requirements.txt

// Then install spaCy in that venv, that will trigger the Cython build, so it will take a bit
$ python -m pip install -e .

// Activate the venv again to make sure you are using the pytest from the venv
$ source env-2020-10-15/bin/activate

// Then run pytest again
$ pytest spacy/tests/regression

[forgetso]

@tiangolo thanks. Updating setup tools and requirements seems to have done the trick.