Implementing ReadSeeker

[tab1293]

It would be useful if the Reader struct implemented the ReadSeeker interface. This would come in handy when trying to serve RAR files over HTTP with http.ServeContent. This is useful if the content that is archived is a video file; providing a ReadSeeker would allow players such as VLC to seek in the video.

[f4n4t]

I would also like this feature.
Im using rardecode to retrieve information for media files in combination with MediaInfoLib.
It would be realy useful to have the ability to seek on rar files.

[KyleSanderson]

@nwaples any issue if I try and take this on? I'm seeing the same problem and it's causing my filesystem to read 4GB to get a 16KB block after that boundary.

[nwaples]

Are these files stored compressed?
If they are I dont see a way to do it. If a file is split into blocks (and multiple files) you can seek to the next block. The file block header stores the Unpacked size of the whole file, and the Packed size of that block. There is no way to know what the size of the unpacked information for that block is, so you dont know what the index is for the beginning of that block. If you need to seek in the file you don't know what block to go to and how far in.
If the files are stored uncompressed then the packed size is the same as the unpacked size, so it would then be possible.
I could probably do Seek, but only for uncompressed files (unless I missed something).

[KyleSanderson]

Are these files stored compressed? If they are I dont see a way to do it. If a file is split into blocks (and multiple files) you can seek to the next block. The file block header stores the Unpacked size of the whole file, and the Packed size of that block. There is no way to know what the size of the unpacked information for that block is, so you dont know what the index is for the beginning of that block. If you need to seek in the file you don't know what block to go to and how far in. If the files are stored uncompressed then the packed size is the same as the unpacked size, so it would then be possible. I could probably do Seek, but only for uncompressed files (unless I missed something).

No they're not compressed - rar is just used for splitting.

[nwaples]

I have an initial implementation of this on my bytereader branch. It supports seeking on uncompressed files. The checksum has to be ignored though, with the either OpenFSCheck or SkipCheck option. It wont work with the Reader's, only in RarFS and File.Open().
I have not tested it much, so there may be bugs.

[javi11]

Hey, nice work here :). I was trying your branch of bytereader, you didn´t exported the methods yet no?

[nwaples]

You need to use OpenFS to open a filesystem implementation of the files in the rar archive.
Then when you open a file you need to do a check against io.Seeker (or ReadSeeker, ReadSeekCloser), just like any other fs.FS implementation.

[javi11]

Thanks its working :)

[MunifTanjim]

Opened #55 with the API inspiration taken from https://github.com/muja/unrar.rs