Just realised that this appears to be the same on the main site http://www.themoviedb.org/movie/top-rated?page=18 Clicking next to view page 19 of 66 throws an error.

Hi Jools,

Strange problem indeed. I've created a new ticket for this.

Thanks for reporting it.

This should be fixed. Can you give it a try now?

Seems to be working great. Cheers! :-)

I'm not sure if this is the same thing but it seems related. While the final pages on the list are now presenting there is something odd about the last two pages. Essentially a block of 7 movies at the end of the second last page, is replicated at the beginning of the last page. The structure as I write this

Page 86 of 87 [Item 1-13 of 20 seem normal] 14 Alex Rider 15... 16... 17... 18... 19... 20 Street fighter

Page 86 of 87 1 Alex Rider 2... 3... 4... 5... 6... 7 Street fighter [Item 8-12 of 12 seem normal]

I've confirmed that this isn't my code by viewing the API in the browser http://api.themoviedb.org/3/movie/top_rated?api_key=XXX&page=86 http://api.themoviedb.org/3/movie/top_rated?api_key=XXX&page=87

I was thinking that this might not just be a problem with the last pages, although the few pages before the last seem fine. I took a look at some of the earlier pages and saw it occur to on a smaller scale. The 4th pages begins with a repeat on the last item on the 3rd. I thought maybe there was some growing rounding error that cause the big mistake at the end, so I kept looking. The 7th page begins with a repeat of the last 2 of page 6. Page 9 begins with a repeat of the last 1 from page 8. Page 11 has a single repeat from the end of page 10 So there's no growth real growth pattern. And then I saw for the first time I saw a movie (Das Experiment) replicated from the end of one page into the middle of another page. It's as though each is pregenerated and they weren't generated at the same time. The suggestion of a growth pattern was probably just a result of where movies are most likely to change position.

It's as though each is pregenerated and they weren't generated at the same time. The suggestion of a growth pattern was probably just a result of where movies are most likely to change position.

That's exactly it. Combined with the fact that each page is cached for ~24 hours and depending when each one was requested we cannot guarantee items are going to be consistent throughout.

Ah, well at least it makes sense now. Given that they're cached for 24 hours, wouldn't it make more sense to just generate them all at once, every 24 hours, straight into the cache, rather than depending on user requests, which I imagine will be plentiful enough to justify the generation, but random enough make the current method give incorrect results?