#vogon

I second this motion

(from @catchymemes, couldn't re-blog because it wouldn't let me add video).

So I decided I did want to deal with the PDF problem.

I didn't resolve the large memory footprint problem yet, but I think today's work puts me closer to a solution. I've written a new custom PDF reader on top of the PDF.js library from Mozilla. This reader is much more simple than the full one built into Firefox browsers, but it utilizes the same underlying rendering libraries so it should be relatively accurate (for desktop users). The big benefit this brings is that it's a match for the existing CBZ reader controls and behavior, unifying the application and bringing in additional convenience features like history tracking and auto link to next issue/book when you reach the end of the current one.

The history tracking on the PDFs is a big win, in my opinion, as I have a lot of comics in PDF form that I can now throw into series and read.

Word of warning. PDFs bigger than a few MB can completely overload mobile browsers, resulting in poor rendering with completely missing elements, or a full browser crash. I'm betting the limit has a lot to do with your device's RAM, but I don't have enough familiarity to know for sure. These limitations are almost the same as the limits for the Mozilla written reader, so I think it has more to do with storing that much binary data in a browser session than how I am navigating around their APIs.

As far as addressing that problem, I suspect that if I can get TCPDF to read the meta data (i.e. the number of pages), I can use the library to write single page PDF files to memory and serve those instead of the full file. I'm already serving the PDF as a base64 encoded binary glob rather than from the file system directly. It doesn't solve the problem of in-document links, but since I provide the full Mozilla reader as a backup (accessible by clicking the new "Switch Reader" button), you can always swap readers if you need that.

As noted earlier today, the install script is still not working correctly so unless you have a working install already you'll have to wait until I get my VMs set up to be able to test and fix that. Good news on that front, the new media server builds much much faster than the old PI, so fast iteration should be very doable.