Add a few model dependent questions to parse flow.cylc in jinja2 - #832
Add a few model dependent questions to parse flow.cylc in jinja2#832metdyn wants to merge 3 commits into
Conversation
before templating suite_str for flow.cylc, in workflow case:
{% for i in range( 1, models[model_component]['ensemble_num_chunks'] + 1 ) %}
[[RunJediEdaControlPertExecutable_chunk{{i}}-{{model_component}}]]
{% endfor %}
|
I'm not against merging this if it solves the issue, but I'd like to maybe propose an alternative solution. I believe the issue you're encountering is that trying to use a task question in the runtime section of (A) Either as a comment: (B) Or better yet, as a parent task. This has additional benefits such as being able to track the tasks execution as a group in cylc: Admittedly, these strategies will not work if your control logic depends on tasks being in an adversarial relationship with each other. For example if you needed to have a case like: |
|
@mranst, Let's not use tricks to bypass the scientific need. A simple fix like this will solve the problem. BTW, thanks for the workaround you provided. I am sure case (A) will not work, case (B) will bypass the issue but leave scientific confusions to others, because the task |
Description
This PR addresses issue #831, which is required for EDA ControlPert framework.
It solves the issue in flow.cylc in case of complex rendering.
The code here replaces self.experiment_dict by render_dict, where the latter contains the former dictionary plus two key-value pairs involving ensemble_num_chunks amd ensemble_num_members needed by EDA.
Dependencies
Impact