Flex-objects data file ignored

I have two custom data files I am using with flex-objects. I have verified several ways that the files are well formed, syntactically correct JSON. When I update these files, one of them always disappears. The object counter for that file in the Grav admin UI goes to zero, and none of the data appears. The other file is fine, and always shows up.

The thing is: the missing file has always eventually appeared. I’ll do SOMETHING or something will happen and it will just show up the way it was always supposed to. But I have no clue what makes it vanish or reappear. Clearing the various caches does nothing. Enabling and disabling caching does nothing. I got it to show up by restarting Apache on my local machine once, but I haven’t been able to reproduce that, and anyway, it’s not like I can do that when the site’s up on a public webhost.

This has just happened again and I’m freaking out, partly because this board seems kind of like a ghost town sometimes and I don’t that the answer exists anyplace. Has anybody else encountered this? Any ideas for a potential fix?

There’s nothing like posting to a public forum to make the answer show up. I am TENTATIVELY chalking this up to line breaks in the file being Windows and not Linux. At any rate, when I converted the line endings in Notepad++ from CR-LF to just LF the data appeared again. Still weird, since I don’t have to do this kind of thing ever for the other file. Hoping this turns out to be the reproducible fix. :crossed_fingers:

OK, update: Linebreaks was not the fix. I also tried removing whitespace to reduce file size. Nothing. I can sometimes get the data to appear by creating a single record through the admin UI, copying my real data into the generated file, and refreshing the admin page.

Nothing that reliably either reproduces or fixes this issue.

Next question: do the record ids have to be unique to the FILE, or unique GLOBALLY? I’ve been generating IDs that were unique in the file but the same from file-to-file.

Does flex-objects cache these IDs somewhere? This still seems like some kind of caching problem, although I have no clear evidence.