[Fantia] Post-processor events running at unintuitive times?

[shrublet]

I believe since the Fantia extractor was refactored (f845298), the post and post-after events now trigger for each content section/entry instead of the entire post as a whole. While I don't inherently think it's a bad change, I'm unsure now how I'd trigger a post-processor when post-after would typically trigger. If count were added as a keyword, comparing and using filter with num and count would do the trick, and for post, just num suffices, but I still think it's somewhat unintuitive, especially relative to how other extractors work. I did end up just adding this in the for loop in items if there's no better alternative:

post["count"] = sum(
    len(self._process_content(post, content))
    for content in post["_data"]["post_contents"]
) + 1

Hope you have some insight or thoughts on this.

Edit: Whoops can't get count from _get_post_contents because it inserts the thumbnail into the list when it's called. Changed to just add +1 manually to post_contents to account for that.

Edit 2: Also ended up adding a count and num value for content after the loop yields the directory as I was having some edge case issues where comparing num to count didn't sufficiently achieve what I needed it to.

[mikf]

I tried something that would potentially solve your problem with content_num == content_count as filter, but it doesn't work due to how post-after is implemented and how objects get reused. It currently triggers for the last content section as well as the second-to-last one … (833dce1)

(man, the code has become such a mess over time)

[shrublet]

I tried something that would potentially solve your problem with content_num == content_count as filter, but it doesn't work due to how post-after is implemented and how objects get reused

It's good to know it wasn't just a me problem! I mentioned in my second edit I had to put post["content_num"] += 1 after yield Message.Directory, post as if I put it where it logically makes sense, I would end up with the result you're talking about where it overcounts. It's not a solution, but it's a workaround that's good enough (for me) for now as I don't necessarily need the actual number. At the very least, it's a pretty minor concern all things considered, and I at least have something that works on my end.