Fix ci testing-datalad install for python 3.12 + git-annex (#1667)

* fix ci testing

* test si way of installing datalad

* Revert "test si way of installing datalad"

This reverts commit a6463be.

* another attempt to get datalad to work

* oops

* add comments

* from warning to ci error (Heberto feedback)

Author: zm711

.github/workflows/io-test.yml

@@ -76,10 +76,25 @@ jobs:
         # restore-key hits should result in `cache-hit` == 'false'
         #if: steps.cache-conda-env.outputs.cache-hit != 'true'
         run: |
-          conda install datalad pip numpy=${{ matrix.numpy-version }} -c conda-forge
+          conda install pip numpy=${{ matrix.numpy-version }} -c conda-forge
         # this command is for updating cache. We are resting removal.
         # conda env update --name neo-test-env-${{ matrix.python-version }} --file environment_testing.yml --prune
 
+      - name: Install git-annex
+        # this is the trick from the spikeinterface repo for getting git-annex to work with datalad
+        shell: bash
+        run: |
+          pip install datalad-installer
+          wget https://downloads.kitenet.net/git-annex/linux/current/git-annex-standalone-amd64.tar.gz
+          mkdir /home/runner/work/installation
+          mv git-annex-standalone-amd64.tar.gz /home/runner/work/installation/
+          workdir=$(pwd)
+          cd /home/runner/work/installation
+          tar xvzf git-annex-standalone-amd64.tar.gz
+          echo "$(pwd)/git-annex.linux" >> $GITHUB_PATH
+          cd $workdir
+          git config --global filter.annex.process "git-annex filter-process"  # recommended for efficiency
+
       - name: Configure git
         run: |
           git config --global user.email "neo_ci@fake_mail.com"

neo/test/rawiotest/common_rawio_test.py

@@ -19,19 +19,27 @@
 
 import logging
 import unittest
+import importlib.util
+import warnings
+import os
 
 from neo.utils.datasets import download_dataset, get_local_testing_data_folder, default_testing_repo
 
 from neo.test.rawiotest.tools import can_use_network
 
 from neo.test.rawiotest import rawio_compliance as compliance
 
-try:
-    import datalad
-
+datalad_spec = importlib.util.find_spec("datalad")
+if datalad_spec is not None:
     HAVE_DATALAD = True
-except:
+else:
     HAVE_DATALAD = False
+    # pytest skip doesn't explain why we are skipping. 
+    # raise error if in CI to prevent tests from spuriously skipping and appearing
+    # as passing.
+    if os.environ.get("GITHUB_ACTIONS") == 'true':
+        raise RuntimeError("Datalad is required for running the CI.")
+    
 
 # url_for_tests = "https://portal.g-node.org/neo/" #This is the old place
 repo_for_test = default_testing_repo

neo/test/rawiotest/tools.py

@@ -4,6 +4,8 @@
 
 import logging
 import os
+import importlib.util
+
 
 logger = logging.getLogger("neo.test")
 
@@ -13,14 +15,18 @@ def can_use_network():
     Return True if network access is allowed
     """
 
+    # env variable for local dev
     if os.environ.get("NEO_TESTS_NO_NETWORK", False):
         return False
-    try:
-        import datalad
-
+    
+    # check for datalad presence
+    datalad_spec = importlib.util.find_spec("datalad")
+    if datalad_spec is not None:
         HAVE_DATALAD = True
-    except:
+    else:
         HAVE_DATALAD = False
+
     if not HAVE_DATALAD:
         return False
+    
     return True

pyproject.toml

@@ -49,7 +49,7 @@ test = [
     # "dhn_med_py<2.0", # ci failing with 2.0 test future version when stable
     "pytest",
     "pytest-cov",
-    # datalad   # this dependency is covered by conda (environment_testing.yml)-- maybe moving straight to ci
+    "datalad",
     "scipy>=1.0.0",
     "pyedflib",
     "h5py",