Reconcile filename and title with regards to search

[bent0b0x]

Problem
When we render videos in the upload tool, if a video is lacking a title we show its filename as a fallback. This is okay, but is a tiny problem when we go to implement video search/discovery.

Let's say we want to be able to search on videos that did not have a custom title saved; this means we want to search on filename, because that is effectively the video's title. This means we need to index filename and enable it for keyword search.

However, if this is the case even if a video has a title we will also search on filename (unless we intentionally do not do this). Filename might be entirely irrelevant to the video contents, and I do not think we should search on it if the video has a title. The filename is not exposed to the UI if the uploader entered a title, so we could get search results that aren't clear as to why they have been returned by our query.

Potential Solutions

• in paratii-lib, only search on filename if title is not populated for a particular video
• have the FE save the filename as title on video upload, meaning that we do not have to index filename for search at all and can always just look at title, whatever its value may be.

cc @geckoslair @jellegerbrandy

[jellegerbrandy]

I thiknk we just need to idnex filename and that's it. (Or are we alreyad doing so, @geckoslair ?)

Search is often fuzzy - and expected to be so, i think. I.e stemming is standard (search for "working" and it will find "work " and "worked" as well), if you find synonyms (like google does a bit) you will not be confused (search for "work" and find a "job").

What will happen here is that you have a file called "Ben's video.mpg" and upload it under the title "Pretty Kitty", you will still find it when you search for "Ben's video", even if you do not see the filename anywhere. For me, that is cool.

[bent0b0x]

Yes that is already being done, but I don’t think we should :)

We’ll have fuzzy search but since we don’t surface the filename (nor do I think we should) the results could look wrong. And usually or often the filename is just nothing meaningful/not relevant, so why would anyone even want to search based on the filename? Just think of searching by filename on YouTube; you would never do it.

But not a huge deal, we can leave the behavior as is.