set directory and load in one call? about rtorrent

[rakshasa]

This is already possible with the load commands, just append any commands to call;

load_start="foo.torrent",print=$d.hash

[blaenk]

@rakshasa: Beautiful, thanks :) I've traced around the source a bit but I can't figure out how rtorrent expects the example call you gave in xml. I am taking a look at rpc/parse_command.cc at the moment. Before, I saw in command_events.cc that apply_load takes the first argument as the torrent's path, and subsequent arguments as 'commands'. What I can't seem to figure out is how rtorrent expects those commands to be written. I have tried, for example, this:

<?xml version="1.0" encoding="UTF-8"?>
<methodCall>
<methodName>load</methodName>
<params>
<param><value><string>thetorrent.torrent</string></value></param>
<param><value><string>print=$d.get_hash</string></value></param>
</params>
</methodCall>

The torrent loads (shows up in rtorrent), but rtorrent tells me Command on torrent creation failed: Could not find '='., the error message issued in parse_command.cc, so I'm clearly not forming my xmlrpc method call's extra commands correctly. Unfortunately my C++ expertise isn't quite at your level to make sense of the command parser. I will keep trying to figure it out, though I would surely appreciate any further help you can give.

I just came across this thread on the old issue tracker that helps a lot with other questions, but unfortunately doesn't go into the way the xml should be formed.

[rakshasa]

Should be 'print=$d.get_hash=' as there's a second command embedded.

[blaenk]

@rakshasa: Wow! Indeed, I can't believe I kept trying and spending all that time for such a minor mistake that I should know by now. I guess it's time to take a break. Once again, thanks for your time in helping me with such a petty issue, who knows how much longer I would've spent trying this.

Just fixed that typo and indeed everything works beautifully. Now I can encapsulate all of the necessary logic into one call and rtorrent will automatically take care of calling it once the asynchronous torrent loading is done. Nice :)

Keep up the awesome work.

[blaenk]

@rakshasa: Just a quick question, is there a difference between calling load and load_start aside from (my impression that) rtorrent starts the torrent once it's loaded with load_start? I tried doing load with extra calls as detailed in this issue. For example, a d.set_directory which did work (showed up in rtorrent to the directory I set), but I also wanted to finish off the extra commands with a d.start= to start the torrent, but it didn't seem to have any effect. I think I saw in the view that it was closed, inactive, so I did a d.open= before the start and the interesting thing is that the torrent then does show up as [OPEN], but still inactive. I thought maybe it was a bug that prevented the last command from being executed, but adding an extra d.start= at the end had no effect.

For the bigger picture:

<?xml version="1.0" encoding="UTF-8"?>
<methodCall>
<methodName>load</methodName>
<params>
<param><value><string>thetorrent.torrent</string></value></param>
<param><value><string>d.set_directory=/some/dir</string></value></param>
<param><value><string>d.open=</string></value></param>
<param><value><string>d.start=</string></value></param>
</params>
</methodCall>

I do notice that the way core/download_factory.cc does it is by doing d.state.set with a parameter of 1, but is this only an internal command? Because I don't see it with system.listMethods. I kind of figured those mapped to d.start et al.

Would appreciate your insight. Thanks!

[rakshasa]

The download factory does some black magic (read; hacks that need to be cleaned up) in order to start the torrent, and as such use load_start and not d.start.

The d.state.set command is internal, and should not be called by the user at all.