trakt.tv

[jusjoken]

Have you considered adding a match list the user maintains like BMT does that forces a match between a local title and a trakt item?

k

[tmiranda]

Originally Posted by jusjoken

Have you considered adding a match list the user maintains like BMT does that forces a match between a local title and a trakt item?

k

Yes, I just coded it this morning as a matter of fact. It will be editable by the user as well via a .properties file.

The whole "fuzzy matching" issue is a general headache. It makes you appreciate the human brain all that much more. We can just glance at something and implicitly know it's the same thing yet it takes endless logic to program a computer to do it and no matter how long you work at it, it will never be as good as a human.

[Skirge01]

Tom, I know you're already on 0.08, but I'm seeing some strange stuff in 0.06. When I wipe Trakt, it seems that I always have to wipe a second time in order to clear everything out. Have you seen that at all? I'm going to keep an eye on this during the rest of my test cases to see if there's any commonalities on what's remaining.

[tmiranda]

Originally Posted by Skirge01

Tom, I know you're already on 0.08, but I'm seeing some strange stuff in 0.06. When I wipe Trakt, it seems that I always have to wipe a second time in order to clear everything out. Have you seen that at all? I'm going to keep an eye on this during the rest of my test cases to see if there's any commonalities on what's remaining.

No, I haven't seen that. I'm not too concerned either because that's only something in there for testing.

I'll do a complete wipe tonight and see what happens on my system.

You should move off of 0.06 right away and go to 0.07. I hate to see you waste time testing something that is obsolete.

[Skirge01]

Originally Posted by tmiranda

No, I haven't seen that. I'm not too concerned either because that's only something in there for testing.

I'll do a complete wipe tonight and see what happens on my system.

You should move off of 0.06 right away and go to 0.07. I hate to see you waste time testing something that is obsolete.

Yeah, I wasn't concerned about it, either. More curiosity.

As for the testing, it's a bit of an OCD thing. I finished, though, and am on 0.07 now.

[tmiranda]

I did some checking on the "complete wipe" issue. I think what is happening is that Trakt actually takes some time to delete the history even after the API returns. If you run the wipe and then immediately look at the history it will still contain watched shows. If you wait a few minutes the number of watched shows decreases. It eventually gets to zero. (You may have to exit the dashboard and re-enter.)

[Skirge01]

Completed testing on 0.07 and I have some unexpected results. Note that every setting not shown in columns B, C, D, or E are set to 'NO' for every run. I colored everything that matched perfectly.



I was expecting the number of shows/episodes sync'd and movies sync'd to match, regardless of the multiple watches settings. In theory, the number of watches shouldn't have an impact on whether or not the show is matched, assuming the settings are similar (e.g. closest match).

For example, shouldn't rows 16, 20, and 24 have the same numbers in columns F, H, and J?

Side note: To me, it would seem that a setting of "One*" might be in conflict with setting "Only sync shows that have RealWatchedEndTime" to "Yes". It wouldn't have surprised me if that combination wound up with nothing matching.

[tmiranda]

Originally Posted by Skirge01

Completed testing on 0.07 and I have some unexpected results. Note that every setting not shown in columns B, C, D, or E are set to 'NO' for every run. I colored everything that matched perfectly.



I was expecting the number of shows/episodes sync'd and movies sync'd to match, regardless of the multiple watches settings. In theory, the number of watches shouldn't have an impact on whether or not the show is matched, assuming the settings are similar (e.g. closest match).

For example, shouldn't rows 16, 20, and 24 have the same numbers in columns F, H, and J?

Side note: To me, it would seem that a setting of "One*" might be in conflict with setting "Only sync shows that have RealWatchedEndTime" to "Yes". It wouldn't have surprised me if that combination wound up with nothing matching.

Wow, awesome testing. I need to study these results in more detail but here is some initial feedback.

Rows 16, 20, and 24 probably will NOT have the same values in columns F, H, and J. Let's just look at column H to keep it simple. Row 16 should have the biggest number because it includes literally everything. (i.e. episodes manually set watched and episodes that were really watched) Row 20 should have the smallest because it will ONLY sync episodes that have a RealWatchedEndTime. Row 24 should be somewhere in between because if it does not find an episode with a RealWatchedEndTime it will look for one that was manually set watched.

From your comment on the side note I think there is a misunderstanding of the difference between "Only One" and "One*". Only One will literally only sync one and it will honor the RealWatchedEndTime setting; if none are really watched, none will be synced. One* will TRY to find a really watched episode but if it doesn't find one it will look for one that was manually set watched.

I really appreciate your testing , it looks to be very thorough. I will look at all of the results tomorrow.

Some of the results do not look as I expect. One thing that jumps off the page is that rows 16 and 17 are identical and they probably should not be. Ditto for several of the other identical rows.

[tmiranda]

Originally Posted by tmiranda

Wow, awesome testing. I need to study these results in more detail but here is some initial feedback.

Rows 16, 20, and 24 probably will NOT have the same values in columns F, H, and J. Let's just look at column H to keep it simple. Row 16 should have the biggest number because it includes literally everything. (i.e. episodes manually set watched and episodes that were really watched) Row 20 should have the smallest because it will ONLY sync episodes that have a RealWatchedEndTime. Row 24 should be somewhere in between because if it does not find an episode with a RealWatchedEndTime it will look for one that was manually set watched.

From your comment on the side note I think there is a misunderstanding of the difference between "Only One" and "One*". Only One will literally only sync one and it will honor the RealWatchedEndTime setting; if none are really watched, none will be synced. One* will TRY to find a really watched episode but if it doesn't find one it will look for one that was manually set watched.

I really appreciate your testing , it looks to be very thorough. I will look at all of the results tomorrow.

Some of the results do not look as I expect. One thing that jumps off the page is that rows 16 and 17 are identical and they probably should not be. Ditto for several of the other identical rows.

I found the bug that causes rows the duplicate rows (16 & 17, 18 & 19, etc) and it will be fixed in version 0.8.

[Skirge01]

Originally Posted by tmiranda

I found the bug that causes rows the duplicate rows (16 & 17, 18 & 19, etc) and it will be fixed in version 0.8.

Glad I could help. Testing is the easy part! Is there a way to get a text file of what it wants to send up to Trakt (really, all I'd need is TV Show, Episode, and Movie names)? I'd like to run a comparison program on the results to see what's different across each run. That would even speed up some of the testing, since I wouldn't have to actually send it up to Trakt and then clear everything. If this isn't something it currently does or is too much wasted effort, I definitely understand.

[tmiranda]

Originally Posted by Skirge01

Glad I could help. Testing is the easy part! Is there a way to get a text file of what it wants to send up to Trakt (really, all I'd need is TV Show, Episode, and Movie names)? I'd like to run a comparison program on the results to see what's different across each run. That would even speed up some of the testing, since I wouldn't have to actually send it up to Trakt and then clear everything. If this isn't something it currently does or is too much wasted effort, I definitely understand.

Sorry, I don't fully understand what you are looking for. Can you explain a little more?

I hope to have the next version available this weekend and I'll try to incorporate whatever it is you need to help with the testing.

[Skirge01]

No worries, Tom.

For all Sage DB entries which your plugin matches to something in Trakt, I would need a text file containing the Trakt TV Show name, season #, episode #, and episode name; plus, the Trakt movie name.

This doesn't need any special formatting, so long as it's uniform across all runs, regardless of settings:

TVShowName,Season#,Episode#,EpisodeName

*OR*

MovieName

I'll then use something like WinMerge to compare the different runs to see exactly what is or is not being sent to Trakt during each test. My intent is to see if there are any commonalities across the runs and what is being skipped and whether or not it's an actual issue that it was skipped.

That would be enough.

Now, in a perfect world, I'd love to also have a list of any Sage DB entries which failed to match in Trakt. I don't know what would be easier for you, but if that were either a separate file or collectively isolated in the same file (e.g. after all the matched entries), I might be able to help identify issues with the fuzzy matching.

[tmiranda]

Originally Posted by Skirge01

No worries, Tom.

For all Sage DB entries which your plugin matches to something in Trakt, I would need a text file containing the Trakt TV Show name, season #, episode #, and episode name; plus, the Trakt movie name.

This doesn't need any special formatting, so long as it's uniform across all runs, regardless of settings:

TVShowName,Season#,Episode#,EpisodeName

*OR*

MovieName

I'll then use something like WinMerge to compare the different runs to see exactly what is or is not being sent to Trakt during each test. My intent is to see if there are any commonalities across the runs and what is being skipped and whether or not it's an actual issue that it was skipped.

That would be enough.

Now, in a perfect world, I'd love to also have a list of any Sage DB entries which failed to match in Trakt. I don't know what would be easier for you, but if that were either a separate file or collectively isolated in the same file (e.g. after all the matched entries), I might be able to help identify issues with the fuzzy matching.

I see. What you are looking for is mostly in the logfile now.

For shows that matched, look for "createDeltaFile" at the end of the run. That will contain the JSON (which is a formatted string) of all the items that were matched on trakt. (The plugin compares the trakt database before and after the sync to keep track of what was actually sent.)

For shows that didn't match anything on trakt look for "did not timestamp". This will contain the show name (movie or episode) in the Sage DB that was not matched in the trakt DB.

Let me know if you can use that or if you need something else.

[Skirge01]

Quite a bit of editing/converting, through three different programs, but I think I can make that work for most scenarios. I feel bad making it sound like that's difficult, considering what you would have to do to log it separately for me.

However, the log files are huge when I do the "All Watched" and Sage isn't able to contain it in "only" 4 log files. Is there a way to increase the number of log files Sage writes?

[tmiranda]

Originally Posted by Skirge01

Quite a bit of editing/converting, through three different programs, but I think I can make that work for most scenarios. I feel bad making it sound like that's difficult, considering what you would have to do to log it separately for me.

However, the log files are huge when I do the "All Watched" and Sage isn't able to contain it in "only" 4 log files. Is there a way to increase the number of log files Sage writes?

I think the better solution is to reduce the amount of info my plugin puts into the logfile I'm actually working on that now......

[jusjoken]

Originally Posted by Skirge01

Quite a bit of editing/converting, through three different programs, but I think I can make that work for most scenarios. I feel bad making it sound like that's difficult, considering what you would have to do to log it separately for me.

However, the log files are huge when I do the "All Watched" and Sage isn't able to contain it in "only" 4 log files. Is there a way to increase the number of log files Sage writes?

change the following in the sage properties and restart.

num_logfiles_to_keep=4

[tmiranda]

The next version is ready. I emailed it to Skirge01 and dstanley, if anybody else wants it let me know.

Changelog:

// - Added scrobbling.
// - Improved episode matching by searching for all episodes of a show before searching by episode title.
// - Improved episode, movie, and show matching by looking for user-defined keyword substitutions.
// - Do a more graceful exit if user manually stops the database sync process.
// - Added the ability to sync a specific movie, show, or episode to improve testability.
// - Fixed a bug that caused shows to be set as synced if the sync failed and it had duplicates.
// - Fixed a bug that prevented honoring the RealWatched setting if sync mode was Everything.
// - Fixed a bug that prevented sync_movies and sync_tv settings to be ignored.
// - Fixed a bug that prevented Trakt from finding matches with strings containing "?" even though they were URLEncoded.

[tmiranda]

The next version is almost ready. I'm doing some final testing and plan on releasing it tomorrow.

Change log:

• Now uses phoenix core services to determine media type, parse titles, and search for IMDB IDs. This is slower, but more accurate.
• Added support for imported videos.
• Status now automatically updates as the sync progresses.
• Improved sync progress messages.
• If no ShowTitle is found, use the physical file name as the title.
• Clarified configuration info messages to indicate that sync TV and sync Movies determines what gets synced.
• Fixed a bug in TraktShow that would cause an error if the show JSON did not contain any episode information.
• Fixed a bug that caused shows to be set/cleared watched even if they were set not to.

Note that using phoenix core services has increased the accuracy of the initial sync but it is slower. You can plan on 3-4 seconds per item. If you have a lot of items, well .... you can do the math.

The good news is that it now shows you the progress so you can see what is happening.

[tmiranda]

Version 0.10 is ready, change log above.

To install, unzip the attached to your sage install folder. Then go to Plugins -> All Available. Look under "UI Mod" for Trakt and install it.

Note that this procedure is a little different. I added a UI mod plugin to enable the status updating while the sync is in progress. When you install the UI Mod plugin everything else you need will be installed.

[Skirge01]

Originally Posted by tmiranda

Note that using phoenix core services has increased the accuracy of the initial sync but it is slower. You can plan on 3-4 seconds per item. If you have a lot of items, well .... you can do the math.

I'll get back to you around Christmas on my testing.

In all seriousness... the feedback, especially, is nice. Starting my tests in reverse, so the really time consuming ones will be the final ones.

I don't know if it was necessary/preferred, but I uninstalled 0.09 completely and then installed 0.10 fresh.