[vorbis-dev] some (basic?) development questions...

[Michael Smith]

At 03:25 AM 5/16/00 -0700, you wrote:
>Pardon my ignorance, as I just started playing with Ogg stuff a few days
>ago, but I'd appreciate some help if someone has time...
>
>I was trying to write a very basic comment editor for Ogg Vorbis files.  I
>think I mostly have it working, but it seems to output slightly broken
>streams.

Well - there's a coincidence. Guess what I've been doing? Mine now works,
produces correct streams, and has a suckier-than-sucky user interface. 

If you're still having troubles, and you'd like to look at my code, just
ask - it's not too horrible for the most part. 

>
>Here's a basic overview of what I'm doing; could someone tell me if I've
>got the right basic idea?
>
>Basically because the all the packets are variable-width, I (AFAIK) need
>to read out all the packets one at a time, replacing the second (comment
>header) packet with the new comments.  All the other packets can stay
>exactly the same.  I read the packets out of one ogg stream, and feed them
>into a second ogg stream, grabbing pages from the second stream as I get
>them and writing them to disk.  Then I move the new file into place where
>the old file was.
>

That's exactly correct - in almost all cases, everything other than the
second page (which includes the second packet) will remain identical, so
you'd be able to just copy the data raw from the input file to the output
after writing out the headers normally.

Unfortunately, under some circumstances the comments packet could be
sufficiently large that it gets split over two pages - then the rest of the
stream has to be decoded (not fully - you need only extract the packets
from the stream), then restreamed, because the page sequence numbers
change. It'd be possible to detect this case, and only use this (slower)
method if you needed to (my code currently doesn't).

>It's a little ugly, but I can't think of any other way to do it.
>
>If I knew the second or third header was going to fall on a page boundary,
>I could figure out how long the header pages were in the first file, then
>figure out how long the header pages were in the new stream, then expand
>or contract the original file to match and overwrite the header pages.  
>That would probably be a classier way of doing it, but since I am probably
>screwing with the page boundaries by changing the comment header, that
>won't work, and my two file approach is apparently the only way to do it.
>
>Can anyone tell me if the above assumptions are correct?  If they are I
>can show you a few lines of pseudocode and hopefully you could tell me
>what I'm doing wrong...

Well, the theory here is all correct, so it's probably just a minor bug.
The only part of the process that the libvorbis API makes somewhat hard is
actually regenerating the header packets (well, just the comment packet is
needed - but the api only allows for generating all of them). The rest is
just a matter of using the ogg layer of libvorbis to pull the packets out,
then recreate pages from them and push these out onto the disk.

If you want me to take a look at the code, you can mail it to me, and I'll
try and figure out where the problem is.

Michael

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/