What I was going to do (but didn't get to due to school and time constraints..) was peek around.. 5mb of data at a time, check if the sequence for AIPrivateData is found, and if so, return. Of course this'd mean that we could peek over the entire file... So, also not an optimal thing, however thanks to Adobe being adobe.. we gotta do it..

What I was going to do (but didn't get to due to school and time constraints..) was peek around.. 5mb of data at a time, check if the sequence for AIPrivateData is found, and if so, return. Of course this'd mean that we could peek over the entire file... So, also not an optimal thing

Sounds good @vladfrangu.
Peeking in a loop with tokenizer will result in an infinite loop as it will not increment the stream position with the data read. But if you know it is PDF, it perfectly fine to 'consume' all the data you need with read instead of peek.

, however thanks to Adobe being adobe.. we gotta do it..
Sounds good @vladfrangu.

Yeah the embedded metadata looks amazingly bad. Why bothering to put that in if it is always the same?

Peeking in a loop with tokenizer will result in an infinite loop as it will not increment the stream position with the data read. But if you know it is PDF, it perfectly fine to 'consume' all the data you need with read instead of peek.

I'm aware peek doesn't advance the position (hence it's name), so you'd skip as many bytes as you peek if you check in a loop, although you can probably just.. read it and skip the peek-ignore step

Also, pro tip: you can skip the entire first metadata part if you parse the header (it includes a length of bytes you can skip to get past the whole xml in pdf ordeal)

I'm aware peek doesn't advance the position (hence it's name), so you'd skip as many bytes as you peek if you check in a loop, although you can probably just.. read it and skip the peek-ignore step

Exactly.

Also, pro tip: you can skip the entire first metadata part if you parse the header (it includes a length of bytes you can skip to get past the whole xml in pdf ordeal)

Bingo, en then use tokenizer.ignore(nrOfbytesToSkip), which will prevent reading the data of the underlying media supports it.

Processing the "PDF blocks" is easier said then done. The COS ("Carousel" Object Structure) format which PDF is based on, requires semi text line oriented processing. Which is complex and tends to cross the binary format scope boundaries.

Can you fix the merge conflict?

Is there any progress on this? Any help needed?

Can you fix the merge conflict?

Done.