Add customizations for use in HEB 115 #3
Conversation
| userGroups = OodSupport::User.new.groups.sort_by(&:id).map(&:name).flat_map{ |str| str.scan(/^canvas(\d+)-\d+/) }.flatten | ||
| enabledGroups = [ | ||
| "135510", # HUIT Open OnDemand Testing | ||
| "139860" # HEB 115: Investigating the Human Genome |
There was a problem hiding this comment.
Nice and clear pattern for adding sub apps to other courses!
There was a problem hiding this comment.
Agreed! This is a nice pattern that we can use for other sub-apps as well. Do we think that it's clear from the context that enabledGroups and userGroups refer to lists of Canvas Course IDs, not POSIX group names? I'm mainly asking because as we propagate these changes to other apps, it may be helpful to quickly locate places that are operating specifically on Canvas Course IDs (like here), compared to places that are operating on POSIX group names (or IDs).
There was a problem hiding this comment.
Maybe a note is warranted, since there's some nuance to it. It's the Canvas Course IDs, which are used because they're present in the grouper group names which are translated into POSIX group names.
There was a problem hiding this comment.
Awesome work @jaguillette 🚀 Really cool to see sub apps used this way!
|
Note that I've pushed this out in prod, but haven't set up a PR to add it to IAC yet, pending approval on this PR. Just wanted to get the functionality in front of the instructor. |
There was a problem hiding this comment.
The sub-app approach that you outlined and implemented @jaguillette looks like a good way to go for customizing apps per-course. Thanks for the detailed description, including the caveat about managing app configurations on multiple portal. We'll have to remember that as we move forward with that. Great work! 👍
| # @note Used to define the submitted cluster, title, description, and | ||
| # hard-coded/user-defined attributes that make up this Batch Connect app. | ||
| <%- | ||
| userGroups = OodSupport::User.new.groups.sort_by(&:id).map(&:name).flat_map{ |str| str.scan(/^canvas(\d+)-\d+/) }.flatten |
There was a problem hiding this comment.
Great use of OodSupport to get the user groups (I wasn't previously aware of this helper library). One thing that jumps out at me as being slightly unclear, and this may just be my lack of familiarity with Ruby, is the use of flat_map. I wonder if that could be replaced with a standard map and then just flatten once at the end. To me that reads a little clearer and maybe less redundant, although the result will be exactly the same.
There was a problem hiding this comment.
I have no strong opinions about Ruby syntax, so I'm happy to change it if it makes things clearer. Any Ruby code I'm writing, I'm doing with significant help from Stack Overflow and GPT.
| userGroups = OodSupport::User.new.groups.sort_by(&:id).map(&:name).flat_map{ |str| str.scan(/^canvas(\d+)-\d+/) }.flatten | ||
| enabledGroups = [ | ||
| "135510", # HUIT Open OnDemand Testing | ||
| "139860" # HEB 115: Investigating the Human Genome |
There was a problem hiding this comment.
Agreed! This is a nice pattern that we can use for other sub-apps as well. Do we think that it's clear from the context that enabledGroups and userGroups refer to lists of Canvas Course IDs, not POSIX group names? I'm mainly asking because as we propagate these changes to other apps, it may be helpful to quickly locate places that are operating specifically on Canvas Course IDs (like here), compared to places that are operating on POSIX group names (or IDs).
This PR sets up the spack app with a custom version just for HEB 115, which is in fact the only version available. It sets a pattern for making apps available only to certain classes, and for setting up sub-apps with slightly different configurations.
Sub-App Definition
The sub-app is configured in
local/humanEvolutionaryBiology-115.yml.erb. This setup is described in the OOD docs on sub-apps. I initially thought that the only way to change how the app shows up is to change the name of the yml file, but Issue 3005 in the OOD repo indicates that sub-apps can now override the metadata of their parent app. That means that we can override the title of the app with a top-leveltitle:attribute in the form.yml variant that sets up the sub-app.In this case, the sub-app hard codes the Spack installation and environment to use. There is an existing Spack environment for HEB 115, and I've added Jupyter Lab and the usual suspects for Python to that environment, so that the tools configured for the course are available by default. The definition for this environment has been added to the spack environments folder.
Restricted Access
The other part of that
yml.erbfile is the determination of who to show the app to. This is done with a Ruby function available in OnDemand to grab group names and then some chained commands to turn that into a list of Canvas IDs for courses that the user is enrolled in. This is compared to a list of Canvas IDs that for courses the app is supposed to be enabled for. If there's overlap between the two lists, then the app is displayed. If there isn't then it isn't. The way this visibility is changed is by setting theclusterattribute on the form. If it's set to"*", then the app will be displayed, as it is enabled for any cluster available. To disable the app,clusteris set todisable-this-app. Since there is no cluster by that name, the app is not displayed. Theclusterattribute is set based on the Ruby functions described.Future Enhancement
An enhancement for the app would be to reference custom metadata attached to the cluster definition file. As written, if an app configured this way needs to be made available to a different class, then the app configuration must be changed on the portal node by pulling from GitHub. This isn't too hard in our current setup, with one portal node. However, with multiple portal nodes, we would need a mechanism to update the app configurations on all of the running portal nodes. Instead, we could hook into the existing cluster config update method, ensuring that changes propagate once the cluster definition file in S3 is updated.