• Registered: 2009-06-09
  • Posts: 5

Re: API 2.1

A couple of bits of feedback, all for the movie search:

1) The use of stop words in the search seem to effect the results and the absence of them seems to help to positive match a movie. For example:

silence+of+the+lambs => 0.45
silence+of+lambs => 0.71
silence+lambs => 1.0

2) When searching for "american+beauty" with v2.0 results in a score of 1 but searching on v2.1 results in a score of 0.79... The rest of the results don't have american and beauty in their title so this score surprises me. Other films that also exhibit this behaviour, that I've come across, are:

  • The Green Mile

  • Dr. Strangelove or: How I Learned to Stop Worrying and Love the Bomb

  • Being John Malkovich

  • There's Something About Mary

3) I'm surprised that the addition of a year within the search doesn't help results and in fact often results in a lower score, I can imagine there will be times where the only way to distinguish a title is via its year of release. And there will be times when neither of these are capable of uniquely identifying a movie, this is where it would be nice to see the director also included in the results?

4) I don't know whether I am being silly here or whether this is a genuine issue. When using the movie search with JSON and when a description has an apostrophe in it, it is encoded to \u2019 which is a right single quote mark where as it should possibly be \u0027, an apostrophe. The use of \u2019 causes php when decoding to ISO-8859-1 to insert the character "?", which is its fall back character when there is no equivalent of that character in the destination encoding.

I hope that is helpful

Last edited by Nuisance (2009-06-14 21:25:35)

  • tooth
  • New member
  • Offline
  • Registered: 2009-03-12
  • Posts: 8

Re: API 2.1

I know you've been busy.  Thanks for your work.  Is there any chance we would be able to test out methods like person.search and person.getinfo soon?

  • Registered: 2009-07-02
  • Posts: 2

Re: API 2.1

Just noticed there that an API search for 'Once' returns The Dark Knight with a score of 1.0. Once appears nowhere on the returned xml. However when using the search engine on the site it comes back as the second result.

  • Registered: 2009-07-03
  • Posts: 25

Re: API 2.1

zag2me wrote:
reven wrote:

is there any chance you can include "Series" and "Part" in the getmovieinfo method?

eg for "Star Wars Episode 1", series would be "Star Wars" and part would be "1".  this would come in handy when sorting films (since some dont have a number in them, "Indiana Jones and the Last Crusade" would appear before "Indiana Jones and the Raiders of the Lost Ark" if sorted alphabetically)

I second that request, it was talked about a bit on the XBMC forums as their software can't group movies at the moment.

http://xbmc.org/forum/showthread.php?t= … mp;page=11

Being able to group movies (Die Hard Series, James Bond, LOTR, etc.) would be a great addition to the api...

thx for this awesome site btw smile

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

I just made some small tweaks to the searching algorithm, they can be easily changed back but I would mind seeing you guys run it through its paces.

Some stuff will definitely show up more reliably, but I am unsure of some outer case examples that might be worse now.

Let me know!

Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

Just updated the first post. The Person.search and Person.getInfo methods are now available. Check out the post for more details.

Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

32 Reply by <><

  • <><
  • Member
  • Offline
  • Registered: 2009-01-08
  • Posts: 18

Re: API 2.1

Hey Travis,

Although the Person.getInfo works, here are things i've noticed:
- Not all actors have the images in the person.getInfo (one example is Carrie-Anne Moss in Matrix), picture is available on website but not through API.
- Not all movies from movie.getInfo have the "id" tag set for person (one example is Harry Potter and the Order of the Phoenix).

Other than that, it's working fine.

One other recommendation is to have a movie.getPersonInfo or something similar which will return all personInfo for each actor for that particular movie, for example:

<people>
--- <person>
------ <filmography>
------ <images>
--- </person>
--- <person>
------ <filmography>
------ <images>
--- </person>
</people>

Instead of having to query TMDb for each persons info for a movie, it can be returned as a full list for a movie.

About the search, it seems to be a lot better now...it doesn't return random results as much at all...

Also, has anything changed in the backend with download posters/backdrops etc? Everything seems to be downloading a lot slower...

Take Care

Last edited by <>< (2009-07-09 01:05:28)

meta<browser/>
http://themetabrowser.com

<><'s Website

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

<>< wrote:

- Not all actors have the images in the person.getInfo (one example is Carrie-Anne Moss in Matrix), picture is available on website but not through API.

This is the same situation when we started out with posters. Some really low res versions of the images exist in the db but we're not using those, at all. Only images that have been uploaded are pulled for the API. It's easy to tell if this is the case by looking at the URL on the website. If it's an image from images.themoviedb.org it will be a user uploaded image, and in the api.

<>< wrote:

Not all movies from movie.getInfo have the "id" tag set for person (one example is Harry Potter and the Order of the Phoenix).

They will, your API key had probably already cached the results for that movie. Just wait for the first version to clear and they'll come down with the IDs (approx. every 8 hours).

<>< wrote:

One other recommendation is to have a movie.getPersonInfo or something similar which will return all personInfo for each actor for that particular movie, for example:

<people>
--- <person>
------ <filmography>
------ <images>
--- </person>
--- <person>
------ <filmography>
------ <images>
--- </person>
</people>

Instead of having to query TMDb for each persons info for a movie, it can be returned as a full list for a movie.

That would be a doozy of a query. I'd be searching the db for hundreds of items in a single pass, vs. right now, many small ones. Hard to say which is more efficient, I'll give it some thought.

<>< wrote:

About the search, it seems to be a lot better now...it doesn't return random results as much at all...

Great.

<>< wrote:

Also, has anything changed in the backend with download posters/backdrops etc? Everything seems to be downloading a lot slower...

The amount of traffic is just starting to pin the hard drives in the box. Nothing has changed except we're busier than ever, last month I think we pushed 7TB of bandwidth! There's some stuff in the works but obviously with me footin' the bill every month, it has to be done very economically.

Cheers man.

Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

  • Registered: 2009-01-12
  • Posts: 15

Re: API 2.1

Trying out the new API and things seem to be running smoothly so far.

I did a search for "Star Wars: The Phantom Menace" but got back no results ("nothing found"), I had to to do a full name search "Star Wars: Episode I - The Phantom Menace".

The same search critia on the website returned all the Star Wars movies and a few extras.

Is there a trick to performing "wild" searches like this...??

Shane

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

Hey Shane,

2 things you can do to try tweaking the results (for your search query):

 Star+Wars+*

AND

"Star+Wars:+The+Phantom+Menace"

Both of those will show it in the set of results but they are in essence wild card searches so the movies listed are ordered more by a combination of popularity + name match, than anything else.

Cheers,

Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

36 Reply by <><

  • <><
  • Member
  • Offline
  • Registered: 2009-01-08
  • Posts: 18

Re: API 2.1

Travis,

Can you please run the resizer on backdrops for the thumbs that are still 75x42 and make them 300x169?

The thumbs load a lot faster than the poster size for a preview, but some of the thumbs are wayyy to small to make out what the original might look like.


And, another request, is to have the image size (of the original) listed along with the attributes of the image, it will make it easier for users to only download the size they want rather than download and then see the size, will save some bandwidth also.


Thanks

Last edited by <>< (2009-07-17 18:14:04)

meta<browser/>
http://themetabrowser.com

<><'s Website

  • Registered: 2009-01-12
  • Posts: 15

Re: API 2.1

travisbell wrote:

Hey Shane,

2 things you can do to try tweaking the results (for your search query):

 Star+Wars+*

AND

"Star+Wars:+The+Phantom+Menace"

Both of those will show it in the set of results but they are in essence wild card searches so the movies listed are ordered more by a combination of popularity + name match, than anything else.

Cheers,

Okay, I'll give this a go and see what it does it for me.

Cheers
Shane

  • Registered: 2008-11-09
  • Posts: 12

Re: API 2.1

Does anyone else have this problem?  I tried searching for ghost in the shell and the web request basically hung then eventually timed out. No matter what I searched on yesterday, I got the same problem. Anyone else seeing this?
The Matrix, Constantine, also don't work

EDIT: it seems to be work now.
PS. Is there a full list of the different job types? I am using strongly typed enums for the xml serializer in .NET and I'd like to not have it barf on a never-before-seen job type smile thanks!

Last edited by jackluminous (2009-07-19 18:17:48)

  • Registered: 2009-07-06
  • Posts: 3

Re: API 2.1

DieBagger wrote:
zag2me wrote:
reven wrote:

is there any chance you can include "Series" and "Part" in the getmovieinfo method?

eg for "Star Wars Episode 1", series would be "Star Wars" and part would be "1".  this would come in handy when sorting films (since some dont have a number in them, "Indiana Jones and the Last Crusade" would appear before "Indiana Jones and the Raiders of the Lost Ark" if sorted alphabetically)

I second that request, it was talked about a bit on the XBMC forums as their software can't group movies at the moment.

http://xbmc.org/forum/showthread.php?t= … mp;page=11

Being able to group movies (Die Hard Series, James Bond, LOTR, etc.) would be a great addition to the api...

thx for this awesome site btw smile

I would also love to see this feature implemented in the new API.
Also could we get the MPAA ratings returned in the search? Something like what cmarrin suggested

http://forums.themoviedb.org/post/1813/#p1813

cmarrin wrote:

<ratings>
  <rating>
    <rating_org>BBFC</rating_type>
    <rating_name>18</rating_name >
    <rating_advisory>Language</rating_advisory >
  </rating>
  <rating>
    <rating_org>MPAA</rating_type>
    <rating_name>R</rating_name >
    <rating_advisory>Language</rating_advisory >
  </rating>
</ratings>

  • phaza
  • New member
  • Offline
  • Registered: 2009-07-28
  • Posts: 1

Re: API 2.1

Should the *.getInfo-actions return arrays in json? Seems a bit odd to me.

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

jackluminous wrote:

Is there a full list of the different job types?

Hey, there isn't off hand but I can see about getting one.

loserpenguin15 wrote:

Also could we get the MPAA ratings returned in the search?

Adding ratings is on the list but they're not even in the database yet, and with little time to work on building out features like this it might be a while.

phaza wrote:

Should the *.getInfo-actions return arrays in json? Seems a bit odd to me.

Do you mean the single record is being returned as a single item array?

Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

  • Registered: 2009-08-03
  • Posts: 9

Re: API 2.1

I'm working on updating XBMC's scraper to the 2.1 api, and it seems like a good time to bring up a current issue with images. As it stands now, it's very hard to associate 1 image with another via regexp. Currently the XBMC scraper uses string hacks to append _mid or _thumb to the "original" image. A bit of grouping would sanify things. For example:

<images>
  <poster id="6929">
    <image size="original" url="http://images.themoviedb.org/posters/6929/Eternal_Sunshine_of_a_Spotless_Mind_.jpg"/>
    <image size="thumb" url="http://images.themoviedb.org/posters/6929/Eternal_Sunshine_of_a_Spotless_Mind__thumb.jpg"/>
    <image size="cover" url="http://images.themoviedb.org/posters/6929/Eternal_Sunshine_of_a_Spotless_Mind__cover.jpg"/>
    <image size="mid" url="http://images.themoviedb.org/posters/6929/Eternal_Sunshine_of_a_Spotless_Mind__mid.jpg"/>
  </poster>
  <backdrop id="42047">
    <image size="original" url="http://images.themoviedb.org/backdrops/42047/fanart.jpg"/>
    <image size="poster" url="http://images.themoviedb.org/backdrops/42047/fanart_poster.jpg"/>
    <image size="thumb" url="http://images.themoviedb.org/backdrops/42047/fanart_thumb.jpg"/>
  </backdrop>
</images>

It would be much easier to guarantee that we're working with the same image that way, and no string hacks needed. I realize it's a bit late for such a change, but I feel it would be a big improvement.

TheUni

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

Hey TheUni,

With 2.1, I added an id attribute so it should be real easy to group images together. This fixes the problem you mention, no?

<image type="poster" size="original" url="http://images.themoviedb.org/..." id="19078"/> 
<image type="poster" size="cover" url="http://images.themoviedb.org/..." id="19078"/> 
<image type="poster" size="thumb" url="http://images.themoviedb.org/..." id="19078"/> 
<image type="poster" size="mid" url="http://images.themoviedb.org/..." id="19078"/>
Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

  • Registered: 2009-08-03
  • Posts: 9

Re: API 2.1

travisbell wrote:

Hey TheUni,

With 2.1, I added an id attribute so it should be real easy to group images together. This fixes the problem you mention, no?

Unfortunately that does not help. The problem is, the images can be in any order, and there's no sane way to tell if we're still on the same image when going from one line to the next. The xbmc scrapers are pretty basic in this regard. We'd have to Scrape once for the id, cache it, then scrape again for thumb, and again for original. If things were grouped as I proposed, we'd know that as long as we were in the poster element, we're dealing with the same image.

I hope that hopes to explain things. My other argument is that our string hacks are currently broken (and there's not an easy way for us to fix) so we're grabbing the same (original) twice. Making it easier to get the thumb would significantly lessen XBMC's bandwidth load on tmdb smile

Thanks for listening
TheUni

  • From: Paramaribo, Suriname, SA
  • Registered: 2009-07-14
  • Posts: 5

Re: API 2.1

Is there something wrong with the API 2.0 and 2.1 ?? I am not able to get any information within the browser or a linux server which is trying to download all available posters and backgrounds for my 50.000 DVD catalogue ...

Any idea what is going wrong ??

Have a nice weekend and thanks so far.

mywebserver's Website

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

No issues that I'm aware of and everything looks to be working for me.

Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

  • Registered: 2009-08-18
  • Posts: 3

Re: API 2.1

Thing missing from Personal Info:

Date of Death:Living or Date
Gender: Male or Female
Actor Bio
Under <filmography> you need to include the imdb# for each film...

Everything works great so far

Michael T.

Last edited by michael1t (2009-08-18 01:22:47)

  • Flowy
  • New member
  • Offline
  • Registered: 2009-09-05
  • Posts: 1

Re: API 2.1

Nice improvements, i like that

But one question about the langueges, when are you going to start the implemation?
This feature would be great, many thanks

Flowy

  • From: Calgary, Canada
  • Registered: 2008-09-01
  • Posts: 770

Re: API 2.1

michael1t wrote:

Date of Death:Living or Date
Gender: Male or Female
Actor Bio
Under <filmography> you need to include the imdb# for each film...

Some of those would be fantastic additions, agreed. I'll see what I can do.


Flowy wrote:

But one question about the langueges, when are you going to start the implemation?

Languages are a huge under taking and before I were to start that I feel as though I need to nail down some other feature requests.

It's on the list but I can't really speculate as to when.

Travis Bell
themoviedb.org | The open movie database
Follow TMDb on Twitter
Found a bug? Got a feature request? Check our Lighthouse first.

travisbell's Website

  • Registered: 2009-01-12
  • Posts: 15

Re: API 2.1

While searching for "Seven+Samurai*" through the API returned no results, the web interface did (return the right result), while using "Shichinin+no+samurai*" through the API returned the correct result?

Also tried "Crouching+tiger+hidden+dragon*" to see what I'd get back and got the correct result??

Shane