pitrou commented on code in PR #41904:
URL: https://github.com/apache/arrow/pull/41904#discussion_r1654733971
##########
docker-compose.yml:
##########
@@ -1221,6 +1222,50 @@ services:
volumes: *conda-volumes
command: *python-conda-command
+ conda-python-no-numpy:
+ # Usage:
+ # docker-compose build conda
+ # docker-compose build conda-cpp
+ # docker-compose build conda-python
+ # docker-compose build conda-python-no-numpy
+ # docker-compose run --rm conda-python-no-numpy
+ image: ${REPO}:${ARCH}-conda-python-${PYTHON}-no-numpy
+ build:
+ context: .
+ dockerfile: ci/docker/conda-python-pandas.dockerfile
+ cache_from:
+ - ${REPO}:${ARCH}-conda-python-${PYTHON}-pandas-${PANDAS}
+ args:
+ repo: ${REPO}
+ arch: ${ARCH}
+ python: ${PYTHON}
+ numpy: ${NUMPY}
+ pandas: ${PANDAS}
+ shm_size: *shm-size
+ environment:
+ <<: [*common, *ccache, *sccache]
+ PARQUET_REQUIRE_ENCRYPTION: # inherit
+ # The without numpy mark is only used to select tests to be run
+ # when numpy is not installed but those tests will also run
+ # if numpy is present.
Review Comment:
I understand the concerns but I think the alternative solution that's been
chosen (run a small subset of the tests) is much worse. Why wouldn't we
exercise reading CSV and JSON files, for example?
I also find concerning that this is hidden somewhere in the
`docker-compose.yml` (many developers will never go look there) instead of
declared in plain sight at the top of the corresponding test files.
For example, if you don't want to run `test_csv.py` when NumPy is absent
(assuming there's a good reason for it), you could add an explicit skip at the
top of that file. The maintenance burden would be small enough.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
