This PR fixes an issue where social annotation point markers were not showing up for some pages, specifically identified in the chemistry course.
The root cause of the issue was due to the ids_added attribute on the publication not taking into account pages which may not have been processed as part of the publication process when ids were fixed, leading to the possibility that certain pages added in after the fact may not have been processed to ensure these ids exist.
The fix here is to remove the ids_added attribute from the publication record and instead add it to the revision record, and run the ensure_ids logic on every publish. Once a revision has been processed and has this flag set, it no longer needs to be processed going forward because the system guarantees at that point that these ids exists at the authoring level.
The risk here is medium since it involves adding an attribute to the revisions table, but special care was used to ensure that this attribute is also included in the revision_from_previous logic and there is a decent amount of unit test coverage around revisions. We will want to spend some extra time in QA testing the publishing process, authoring then republishing to ensure everything is still working correctly here.
This is an important fix for social annotation feature, but also in general to close this gap where the content is assumed to have the required ids based on this flag, however in the current state that assumption is not always correct.
https://eliterate.atlassian.net/browse/NG23-228
This PR fixes an issue where social annotation point markers were not showing up for some pages, specifically identified in the chemistry course.
The root cause of the issue was due to the
ids_addedattribute on the publication not taking into account pages which may not have been processed as part of the publication process when ids were fixed, leading to the possibility that certain pages added in after the fact may not have been processed to ensure these ids exist.The fix here is to remove the
ids_addedattribute from the publication record and instead add it to the revision record, and run theensure_idslogic on every publish. Once a revision has been processed and has this flag set, it no longer needs to be processed going forward because the system guarantees at that point that these ids exists at the authoring level.The risk here is medium since it involves adding an attribute to the revisions table, but special care was used to ensure that this attribute is also included in the revision_from_previous logic and there is a decent amount of unit test coverage around revisions. We will want to spend some extra time in QA testing the publishing process, authoring then republishing to ensure everything is still working correctly here.
This is an important fix for social annotation feature, but also in general to close this gap where the content is assumed to have the required ids based on this flag, however in the current state that assumption is not always correct.