Re: [Ipodlinux-devel] iTunesDB import

[Jens T. <jl...@sh...>]

Hi Terje,

On Sun, Apr 11, 2004 at 03:43:31PM +0200, Terje Wiesener wrote:
> Hi
> 
> Seems we are working on the same thing, basically =)
> 
> I just finished my version 0.11 yesterday, downloadable from 
> http://orz.dyndns.org/stash/parsedb0.11.tar.gz ,
> which parses the playlists as well as the tracks.
> 
> I would be very interested in the performance of your program, when run on 
> the ipod. My port parses my 1600 track db with output to /dev/null in 85 
> milliseconds on my 1800MHz Athlon, but uses a whopping 30 seconds on the 
> ipod.
I have not yet put linux on the ipod. And I will likely not find the
time until the week after next week. I would certainly like to know
if/how well it runs on the ipod. There might be endianess problems in
two places. But that should be fixable easyly.

One way to optimize for ipod use it to only parse the information the
ipod uses or is going to use (most likely Artist, Album, Track Name,
Track Number, Volume Adjust, Playcount, Rating, Genre). That would save
us the UTF conversion etc. I have been thinking about adding that. A
rough estimate is that it would increase performance by about 

Bernard: I guess using mmap is not an option on the ipod since it does
not have a full scale MMU?

> I have tried to build your program, but the ./configure script fails with:
> 	checking how to run the C++ preprocessor... /lib/cpp
> 	configure: error: C++ preprocessor "/lib/cpp" fails sanity check
> I guess I'm missing some package.
This is the first time I have done an autotools package. So it is not
unlikely something is broken in that regard. What version of the
autotools programs are you using? (I am using autoconf 2.59, automake
1.7.9)
Here it says:
	checking how to run the C++ preprocessor... g++ -E
You might also want to try setting the C++ preprocessor manually by
specifying CXXCPP (see configure --help).


> Anyway, good work, I'll probably discontinue work on the port if your 
> library turns out to be more efficient =)
Thanks.
I think a native implementation is preferable. (That's why I ended up
coding it.) It should be quite fast. There also is still room for
improvement (by inlining certain functions that are only used once
etc.). But for now I want the implementation to be correct and clean in
the first place.

Adding playlist support should be easy.

Perhaps we can join our efforts.

If the project is appreciated by the other developers I will put it in
CVS on sf.org.

> Terje
> 
> PS: Funny that we both ended up using the same UTF conversion functions =)
Well, I have actually taken a look on your port. And than figures that
the implementation by unicode.org should be the most reliable. Also the
api did fit perfectly. 

Thank you very much for your feedback!

Jens


-- 
Jens Taprogge