[olympia]

Not sure what you mean on 'hard-coding' the difficult titles. If you really think I am hard-coding the specific titles in the scraper then the answer is that I certainly not. That would be stupid.

I am playing with all the possible title types (there are far too many) and trying to sort them into an order so they are racing with each other.

You're correct, some of the now-working titles are against my bet, but at the end I filtered out some title types which makes them work. Now my new bet is on some potential issue with other titles in an other set of movies, but we will see.

I am dog-tired today after work so probably the above explanation is not crystal clear, and I don't guarrantee that the new version will be any better and not screwing up everyting instead, but here it is:

http://pastebin.com/raw.php?i=DBbdnmUc

[carmatana]

Not sure what you mean on 'hard-coding' the difficult titles. If you really think I am hard-coding the specific titles in the scraper then the answer is that I certainly not. That would be stupid.

I was kidding, I am surprised about the great job you are doing given the quality of the input (akas.imdb).

I will try the latest version and post the results in a couple of hours.

[carmatana]

My deepest admiration,

1 mismatch: Mesrine: Public Enemy #1, the same output as reported 2 or 3 posts above: Mesrine: Public Enemy No. 1

But, technically, the output is correct, it just happens that imdb decided to replace the american title with a new one.

So, I think it is done. I tried with another folder where I store watched movies that I really liked, most of them from non-English countries, and no problem.

One question:

As mentioned I have a movie.nfo just with the imdb url for each movie, so I do not have problems of movie identification or false positives. However, in one of these many tests I removed all these nfo files and ran the scraper using the "Use folder names..." option. I got some mis-identifications (18 over 220 movies)

What is your recommended folder naming to minimize this?

In most cases, it seems that there is another movie or tv program/show with the same title, and I think that having the movie's year as part of the folder name resolves these conflicts, but seems that no.

In other cases it is difficult to understand why the scraper returned what it returned, ie.

The Science of Sleep (2006), was returned as "The Science of Superman"

My folder naming is:

Movietitle (year) -something

If movietitle has a non-standard character (ie ":") I usually replace it with " - "

- Something could be anything, ie:

i) "My wife's name" to identify that it belongs to her collection.
ii) "DTS" because I have a media player that does not play some formats, etc.
iii) "nosubs" to know that I do not have subtitles for that movie.
iv) or anything.

Thank you very much for all your attention and support.

[olympia]

Let's go for the zero mismatch, then we will see if including the '(new title)' type will cause any issues. Here we go:
http://pastebin.com/raw.php?i=DyqH3yjc

Regarding your question about movie identification:
By default (if no nfo file) the scraper only matches the folder name (if that is what checked) with the original title on imdb. I.e. the identification is original title based. That's why you experience that many mismatches without nfo.

I cannot resolve this. Or maybe I could, but then the scraper would list ALL the AKA titles on single search in the search result dialog which would be confusing.

There are 2 workarounds:
- use the original title as folder name
- use nfo (as you are doing it today).

But this should only cause issues when the original title |= USA title. For an USA movie/title, the successful identification rate should be close to 100% I would say.

[carmatana]

Hi Olympia, this one produced a very strange output,

I would say that the number of mismatches is 2 1/2 and 2 of them do not exist in AKAS !!!

Output:

Mesrine: Killer Instinct (2008)
tt1259014 France / Canada / Italy

Output: "Mesrine - Killer Instinct"

This output is similar but in akas does not exist that title, did you convert the ":" for " - " with a regex command?

Mesrine: Public Enemy #1 (2008)
tt0411272 France / Canada

Output: "Mesrine - Public Enemy No1"

Idem

The Boat That Rocked (2009)
tt1131729 UK / Germany / USA / France

Output: Pirate Radio

Actually not a mismatch, but I have the wrong title. The "The Boat that Rocked" is the old American title and the "Pirate Radio" is the New one. I have just discovered this.

I count it as 1/2 mismatch here or +1 in all the previous versions.

[olympia]

This output is similar but in akas does not exist that title, did you convert the ":" for " - " with a regex command?

No, I am not doing anything like that and actually both title appear correctly with ":" on my side. That sounds weird.

Actually not a mismatch, but I have the wrong title. The "The Boat that Rocked" is the old American title and the "Pirate Radio" is the New one. I have just discovered this.

Well, yes. That's what I was referring when I was saying that I am almost sure that sometimes we need the "New title" and sometimes the old to get the wanted title. Honestly I don't know which would be better in overall, but probably we should remove this 'new title' catching, no?

[carmatana]

Perhaps I did something wrong, let me try again to confirm.

[carmatana]

Confirmed, I did something wrong.

I ran previous tests over my video collection backup drive, and there I deleted from the movie.nfo the lines that force the title for foreign, non-english movies. But I backed up my collection during the weekend and this movie.nfo files were overwritten with the originals, that have the title forcing lines, and for the Mesrines movies I have " - " instead of ":"

So, forget my last report.

It should have been:


100% of accuracy !!!!!!!!!!!!!!!

This counting "The Pirate Radio" as the right title for the "The Boat that Rocked". In relation to this and:

Well, yes. That's what I was referring when I was saying that I am almost sure that sometimes we need the "New title" and sometimes the old to get the wanted title. Honestly I don't know which would be better in overall, but probably we should remove this 'new title' catching, no?

I would say that if someone is looking for the "USA / International" title and is scrapping her collection with the "IMDB" scrapper, it is very likely that her reference is http://www.imdb.com and she will be expecting to get the title displayed there (the old if there have not been changes, the new if there have been changes)

If, after having scrapped her collection, the movie "changes" its title she won't notice that. So no problem. If she re-builds her library, she may notice the change but probably she will verify against IMDB and then she will realize that the scrapper did its job well and it will be up to her to re-acommodate whatever she wants to re-acommodote, i.e. renaming its foldername or forcing the old title via the movie.nfo.

I am in this last situation, so it is up to me what I want to do but I cannot blame the scraper for doing the its job well.

Ready to upload this new version to XBMC repositories?

One more time: my admiration for all your diligent work and your great support.

ps I hope that my movie collection is a good sample of others people's collections and that it does not generate side effects on them

[olympia]

I commited the above changes to the offical repo today.

Thank you for testing carmatana! That level of commitment always helps us to make things better!

Cheers!

[User 102910]

.

[mkortstiege]

Yes. <originaltitle> is what you want in your .nfo files. AFAIK not all skins support displaying em. The built-in search is not yet considering the original title when searching. Will double check and put up a PR if needed.