Add explanatory text on tests #124
Conversation
content/docs/guide-development.md
Outdated
| @@ -258,6 +258,13 @@ os.environ['ODA_SECRET_STORAGE'] = '/secrets_custom' | |||
|
|
|||
| It is a good practice to test the developed notebook. This allows to make sure that the code remains valid in the future. | |||
| A test is implemented as another notebook, except that name of the notebook starts with "test_". The notebook should call other notebooks and check that the output matches expectations. See an example of such a test [here](https://gitlab.renkulab.io/astronomy/mmoda/mmoda-nb2workflow-example/-/blob/master/notebooks/test_lightcurve.ipynb). | |||
| To run tests, it is necessary to add some dependencies to the basic environment: in `environment.yml`, add libmagic see [this example](https://gitlab.renkulab.io/astronomy/mmoda/jemx-expert/-/blob/main/environment.yml?ref_type=heads); in `requirements.txt` add the `dispatcher-plugin-nb2workflow` from the git repository, see the last line of [this example](https://gitlab.renkulab.io/astronomy/mmoda/jemx-expert/-/blob/main/requirements.txt?ref_type=heads). | |||
There was a problem hiding this comment.
Why libmagic here? Isn't it specific to jemx-expert?
Concerning dispatcher-plugin-nb2workflow, it helps to look on how dispatcher "understands" the workflow and possible "catch" improper annotations. But it's not needed to run a test notebook.
The downside of having it in the requirements is that it will be installed in the container, blowing its size and complicating dependencies.
There was a problem hiding this comment.
Hi Denys, thanks for this comment. We had already part of this discussion, but it is important to review our understanding.
libmagicis required by nb2workflow to run the notebook possibly through the plugin (see missing compulsory argument in nbadapter call to NB2WProductQuery from dispatcher-plugin-nb2workflow nb2workflow#199 (comment) )- without
dispatcher-plugin-nb2workflownb2workflow run did not complete (see missing compulsory argument in nbadapter call to NB2WProductQuery from dispatcher-plugin-nb2workflow nb2workflow#199 (comment) )
I would be happy if this could simplify, but this is what we concluded trying to run a test on my notebooks.
There was a problem hiding this comment.
- OK, this dependency problems need to be resolved in the package Possibly dependencies problem with
libmagicdispatcher-plugin-nb2workflow#124
Having some dependencies pinned in the repo is sometimes necessary, but I wouldn't recommend putting this into docs. In this case, it's a kind of hotfix.
There was a problem hiding this comment.
I now better understand this libmagic issue. libmagic itself is not a python package. Therefore, it's not possible to ensure its presence in the container via python package dependencies, only with conda (or modifying Dockerfile and installing with system package manager). Then the advise to install it makes sense. But it must be formulated the way it's obvious that it's only needed if there is a specific problem with dispatcher plugin.
And I still think dispatcher plugin is not strictly required there
|
All correct about libmagic, you summarized the issue very clearly. |
|
I verified that it runs also without plugin, as expected, so I modified the documentation accordingly. However, I think that it is important to keep this explanation available. It allows the developer to track possible issues. As a matter of fact, we discovered bugs and features in the plugin, which is used to render. output of workflows in the dispatcher. |
Co-authored-by: Denys Savchenko <56398430+dsavchenko@users.noreply.github.com>
|
@volodymyrss anything to comment? I |
based on current experience, we improve docs on testing