[mythtv-users] Metadata is generic in v31

[James Linder]

> On 12 Jun 2021, at 8:04 pm, Paul Gardiner <lists at glidos.net> wrote:
> 
> On 12/06/2021 03:27, James Linder wrote:
>>> On 26 Sep 2020, at 11:44 pm, Paul Gardiner <lists at glidos.net> wrote:
>>> 
>>> On 23/09/2020 09:10, Paul Gardiner wrote:
>>>> On 20/09/2020 18:53, Paul Gardiner wrote:
>>>>> On 20/09/2020 17:11, Peter Bennett wrote:
>>>>>> On 9/20/20 11:54 AM, Paul Gardiner wrote:
>>>>>>> On 20/09/2020 10:05, Paul Gardiner wrote:
>>>>>>>> On 20/09/2020 00:28, jam at tigger.ws wrote:>> On 20 Sep 2020, at 1:57 am, Paul Gardiner <lists at glidos.net> wrote:
>>>>>>>>>> 
>>>>>>>>>> I've recently updated to 31 from 29.1 and so far run into just one problem. Metadata looked up from within the video player is now generic to the series rather than specific to the episode, so I don't see the episode title or the description of what happens in the episode. There also seems to be a problem with downloading artwork.
>>>>>>>>>> 
>>>>>>>>>> Anyone else seen similar?
>>>>>>>>> 
>>>>>>>>> Paul great confusion here. It seems the movie grabber gets different data at different times.
>>>>>>>>> Also I find (and this must be myth)
>>>>>>>>> 
>>>>>>>>> I->Change Video->Retrieve
>>>>>>>>> gets generic metadata
>>>>>>>>> 
>>>>>>>>> Whereas menu->scan
>>>>>>>>> retrivies specific data
>>>>>>>>> 
>>>>>>>>> but the scan way leaves out some movies and the I way is needed. Also the I way does not log what it is doing. I'll play with the logging to see if I can make sense here.
>>>>>>>>> 
>>>>>>>>> What we need is the developer[s] who crafted this feature to spend a little time polishing the debugging so we can unravel what is happening. The code is complicated and python is alien as it contavenes my personal ethos.
>>>>>>>> 
>>>>>>>> This may be wrong, but the package dependency changes suggest that there may have been a rework of the xml parsing, using a different API. When I get a chance, I'll see if I can find a corresponding commit and try to find anything that looks dodgy.
>>>>>>> 
>>>>>>> Looks like the API change was 3 years ago, so I doubt that's what caused the problem: I can't see it staying broken for 3 years.
>>>>>>> 
>>>>>> Using V31, I have "Preform metadata update after video scan" enabled. After putting video files into the videos storage group on the backend, I run mythutil --scanvideos.  It reliably gets the descriptions of each episode for the series I have loaded into videos. Do you have the file names set up correctly? I use a directory name that is the series name and file names that start SnnEnn for example S01E01.  Sometimes it gets the same artwork for the whole series and sometimes it gets different artwork for each season. It rarely finds preview images.
>>>>> 
>>>>> Actually, it works fine for me if I perform a scan, just as Jam claimed earlier. I get the generic data only if I try to retrieve via
>>>>> Info->Change Video->Retrieve: A choice of program matches is displayed,
>>>>> and I can choose the correct one (there is only one exact title match). Then, I'd expect to be given the chance to choose Artwork etc., but that
>>>>> bit of UI never appears. Could I have part of the installation missing,
>>>>> or might the theme have something missing? I'm using Stepps.
>>>> I suspect this issue as the cause: https://code.mythtv.org/trac/ticket/13518. I see the problem mentioned in this issue in my logs, and I think the failure to fetch thumbnails may be shortcircuiting some of the code used in selecting best match.
>>>> As a cause of the issue, I'm suspicious of: https://github.com/MythTV/mythtv/commit/49e987798ace7d125496cbf874b6bcf43791c8fa. In theory, such a change shouldn't cause change in behaviour, but is const foolproof?
>>> 
>>> I was wrong on both accounts, but I have a partial fix now. This pull request restores the thumbnail and artwork download from ttvdb when manually searching from the Info menu: https://github.com/MythTV/mythtv/pull/235
>> Being a bear of little brain I do not understand the pull request, and certainly the video metadata is very messed-up, and worse than it was when I wrote the above.
>> Of the last dozen videos I tried to update, 1 worked om scan, the rest do not retrieve any data (I->change->retrieve)
> 
> I can explain what was happening and why I made the change. I was never sure the change was altogether correct but, at the time, it certainly had positive effects. This change alone was not the final fix for the problems. It was merged with a whole set of other related changes. Those, at least for a while, made metadata work well. Possibly, ttvdb has changed since then.
> 
> Before the change, we were wrongly generating image urls of the form
> 
> 	http://www.thetvdb.com/banners/banners/xxxxx
> 
> and thumb urls of the form
> 
> 	http://www.thetvdb.com/banners/_cache/banners/xxxxx
> 
> To work, they needed to be:
> 
> 	http://www.thetvdb.com/banners/xxxxx
> 
> and
> 
> 	http://www/thetvdb.com/banners/_cache/xxxxx
> 
> 
> I could have corrected that by just removing "banners/" from second argument of the concat in both cases, but then the image case would have involved removing that string and adding it back, which is sort of odd, so I took the view that the image case should be a simple concat and the thumb should be achieved by altering the second argument to concat.
> 
> I've noticed some problems with metadata just recently, so I wonder if the site has changed again lately.

Paul thanks!
I’ve just been reading what Leslie wrote, and while I’m happy to (and do) build myth from src, I emphasize with the ‘just works’ idea particually with respect to meta data. I’ll pull the latest and see if that affects, but then I’m at a loss of what to do. Hacking the DB springs to mind, maybe even a GUI with the specific aim of dealing with metadata, but that puts me firmly back in the hobbey category.
Cheers
James