Annotations on articles from nprEd not grouped properly on activity pages

[klemay]

This came from https://hypothesis.zendesk.com/agent/tickets/2309 (if unable to view that, try this: https://hypothes-is.slack.com/files/U8D0XFS72/F9QP8LLN5/ticket2309.pdf), which is a long ticket that contains a few other issues, so probably most helpful to stick to summary below:

Steps to reproduce

1. Annotate any article from NPR's education section (https://www.npr.org/sections/ed), i.e. this one: https://www.npr.org/sections/ed/2018/03/13/591858922/how-school-walkouts-test-student-rights-and-school-responsibilities
2. Annotate a second article from this section, i.e. this one: https://www.npr.org/sections/ed/2018/03/12/577134090/transgender-teachers-long-isolated-are-finding-strength-in-numbers
3. Annotate an NPR article, not from the education section, i.e. this one: https://www.npr.org/sections/thetwo-way/2018/03/15/594014437/fiu-pedestrian-bridge-collapses-in-miami-days-after-being-installed
4. Go to your activity page to see these annotations. (or check out mine: https://hypothes.is/users/klemay)

Expected behaviour

You should see three cards, with some variation on the article titles:

1. How School Walkouts Test Student Rights And School Responsibilities
2. Transgender Teachers, Long Isolated, Are Finding Strength In Numbers
3. Pedestrian Bridge Collapse Death Toll Rises To 6 In Miami-Dade County

Actual behaviour

You'll see two cards:

1. The one regarding the third, non-education article will be categorized correctly.
2. The second card will have a title that has nothing to do with either education article: Fake Or Real? How To Self-Check The News And Get The Facts

Browser/system information

Latest version of Chrome for MacOS

Additional details

• Even though the articles are incorrectly grouped under this random title, the hyp.is links for individual annotations work properly. The "Visit annotations in context" and "URL" sections of the card will inherit from the 2nd education article you annotated (weirdly, not the one that matches the title given to the group. See attached screenshot.
• Product backlog #127 would provide a workaround
• Screenshot attached sums this up pretty well:

[hmstepanek]

Just my first impression is that it appears that some code is not picking up the title correctly when the annotation was created. Sounds like Sean has indicated something similar in #127 . The activity pages simply display whatever is returned from the es query so I'm inclined to think the problem originates from the query results, which then means that we have to answer the question how it got in there in the first place. That being said, I think this is outside of activity pages and in that annotation creation logic.

# Fetch results.
results = query.execute(self.request, q, page_size=page_size)

return {
             'search_results': results,

[ajpeddakotla]

This is really helpful. Thanks @hmstepanek! Question though: This is likely to impact other annotations/sites as well - do we have any idea of how big of a problem this might be?

[hmstepanek]

I can’t say without taking a closer look and root causing what the problem is exactly.

[judell]

It's even worse. Here's the first annotation @klemay referenced, which I found by search for user:klemay, any:"fake or real"

The URL on that card -- not the Visit annotations in context link but rather the URL link -- is none of the three above, it's https://www.npr.org/templates/story/story.php?storyId=4989625.

Also, the URL you find when you Google for the title on the card is: https://www.npr.org/sections/alltechconsidered/2016/12/05/503581220/fake-or-real-how-to-self-check-the-news-and-get-the-facts which is yet another none-of-the-above.

Well, if NPR is a PJAX site now, as it appears that it is, then all bets are off and we should expect craziness like this until we deploy https://github.com/hypothesis/client/pull/684.

Once it is deployed, of course, we'll only prevent future messes. I'm not sure there's even in principle a way to unwind/redo what's already done.

[mattdricker]

Issue reported by user in https://app.hubspot.com/contacts/6291320/ticket/244887417/

[blefuscu]

Any news on this? The mess-up between the title and the annotation content is quite annoying and still not fixed.

[robertknight]

No news to report. This issue will be updated as and when someone is able to look into it.