• If a user creates a new series inline then unchecks "add works to series” the new series does not appear in the submit recap, even if you re-chech and select the new series
• When selecting a series with existing items, in some cases the relationship order is not the same between existing series items and new items added: ("X contains Y" vs. “Y is part of X”). To reproduce, select existing series, uncheck and re-check "add works”, then select the series again. https://usercontent.irccloud-cdn.com/file/Lzf37...
What I don't currently understand is that it seems to be working most of the time (relationship will be added to the series and to the work). This needs more testing to figure out in what circumstances the relationship doesn't get added.
Probably some more complicated interaction
test.bb seems to updated with latest commits, and i can see some issues with series section.
I'm redeploying now just to be sure
I had a doubt, but it looks like it was a separate issue with my deployment setup.
monkey: same feeling, most of the time it automatically modify rel set on both entities. and this is i believe how normal entity editor works that is submitting single rel only on series.
Yep. Somehow I did something that must have meant the Work wasn't part of the entities being processed.
Given an infinity of monkeys you can break any code ;p
You know, you can't forget that at this point we're solving bugs for isolated parts of the system, which means the rest is working well :)
but still do you think it is fine if PR don't get merge before final code submission?, i think we should test this thoroughly before merging but not sure if it is possible before deadline.
So since we are working off of a feature branch, we can keep working on the branch until we're happy with the result. The GSoC project doesn't require specifically that PRs be merged, just that the work was done
So don't worry too much about the deadline, you've been doing good work and we are not going to rush to the finish line :)
Of course I would immensely appreciate your continued help with polishing the project once we are past the deadline
also for the editor icon, do you know why there's no sparkles around magic wand on test.bb? or have any particular icon in mind to replace that?
Not sure about the missing sparkles. My guess is that we used the wrong accentuation when we performed the magic ritual, but ¯\_(ツ)_/¯
I think your first choice (the book icon) was the way to go
There must be a fallback mechanism though so that importing faMagic still worked
Ah no, no fallback mechanism needed
Yep, now its fixed. thanks!
Shubh: I think I have an idea of what happened with the missing work-series relationship. I believe it is a timing-related conflict between the revisions.
For the work in question (3ba0d95c-e9b3-4ecd-b121-40acf830c361), have a look at the revisions, and then for each revision look at the relationship set id. The latest revision has a relSetId lower than the revision that came before, which seems odd.
Looking at the relationship sets, the relationship #38951 is the Work-Series relationship, so it was added, but then straight after that a new relationship set was created that overwrote it (without containing the just-added relationship) https://usercontent.irccloud-cdn.com/file/RWsMK...
Not sure if I'm making sense; let me know and I can write it up a bit clearer with more details
Conclusion: at some point in the process we should probably refresh the entity data (or maybe only the sets) to fetch newly-added data before continuing with the next step.
As to finding where exactly that is happening…that's another story
Do we concurrently process entities, or are they processed one after the other?
Ahh it might be just that
we do use Promise.all which is known to cause issues, let me make it sequential
I think that might just be trick ! Although since I can't seem to be able to replicate the issue in any reliable manner, we'll probably have to wait and see.
But at least we're now off the trail of "it didn't create the relationship for the other entity" which was just a red herring from my part
monkey: any idea why error like `ErrorCtor [CustomError]: EmptyResponse` from bookshelf would occur, i've seen it several times and it seems very ambiguous since it doesn't cause any issue most of the time but sometime it do, not sure why.
Sequential calling seems to fixed the issue, also other issues related to works not added also fixed.
one thing left is that currently we just merge the old state with new state on server for old entities (https://github.com/tr1ten/bookbrainz-site/blob/...), this causes some issues for attributes like rels which couldn't directly can be merge. so do you think overriding rels (passing old rels along new ones) would be suitable here?