Discussions on Developer Support

awesome! that works!!!!!!!!!!

Thanks a lot!

We strip out html that you include in a description. You can use the WYSIWYG controls of the editor to add some formatting to your description, but video is not one of the things that are allowed.

If you want to add a video for a topic, you can add it in the weblinks section, located on the right of the edit page, below the image: http://www.freebase.com/edit/topic/guid/9202a8c04000641f800000000fd8bd57

Thanks, but how could I to have a link in the topic view? for example in the view: http://www.freebase.com/view/en/mariana_gurkova

I would like to have a link to youtube video. This is possible? could I have a topic with a link in a topic view?

Regards

but it also looks like we don't actually store the entire Wikipedia entries since the 'raw' service is also returned truncated entries: http://www.freebase.com/api/trans/raw/guid/9202a8c04000641f800000000001daf5

If this isn't enough, I suggest posting your problem (with example urls) to the developers mailing list. The developers responsible for the wikipedia loads hang out there.

Hi Will,

I forgot to mention that I did explicitly set the maxlength to something like 10000 bytes but that didn't help either.

I'm guessing the full content is not available on Freebase ?

With the example above on hamsters, the article content ends with the sentence "Pet Syrian hamsters are descended..." followed by a link to read the full article on Wikipedia.

Thanks once again for your assistance.

why yes, use acre.freebase.extend_query

acre.freebase.extend_query(q, {"query.uri":uri}); where query.uri is the mql variable to be replaced and uri the variable in the code.

There appear to be less than 80,000 films (/film/film) in the database. Where are you getting 300,000 from?

The answer to your question really depends on what library you are using and how you're processing the results, but one of the easier ways to skip ahead is to use the "page" envelope parameter. As far as I can tell this parameter still isn't documented, but if you look at the Query Editor, you can figure out how it works. If you're currently using the "cursor" parameter, switch to "page" instead. If you're using something like the Python library's mqlreaditer, you'll probably need to switch to managing your own paging (it does it internally). Alternatively, you could modify your query to exclude the piece you've already processed (e.g. sort them by release date and exclude ones before/after a certain date), but you always risk introducing a different unrelated bug with that type of modification.

Just pass the envelope parameter to mqlread with the page parameter set correctly:

myEnvelope = {'page':page} result = session.mqlread([myQuery],myEnvelope)

The first thing you need to understand, which I don't think is at all obvious, is that there are two different Javascript environments running in two different contexts -- the Acre server and the user's browser. An Acre script is Javascript, but it's running on the server, not the user's web browser.

You can include scripts in-line by using the

An acre server-side variable can be used by a client JS script just by referring to it using the $ syntax.

Here's a simple example: http://acre.freebase.com/#app=/user/masouras/play&file=client_js_example

The recipe 'Pass a Javascript Object on The Server Side to The Client Side' is also useful if you want to pass more complex data structures around:

http://www.freebase.com/docs/acre_recipes

You can't escape slashes?

Or use something like this example? http://pipes.yahoo.com/pipes/docs?doc=operators#Regex

or:

http://www.kiesler.at/index.php?module=wiki&page_op=raw&page=YahooPipesRegex&lay_quiet=1

http://brooksbayne.wordpress.com/2009/01/11/using-regular-expressions-with-yahoo-pipes/

Forgive me for not knowing the correct terminology here ... it is the name of the items and not the content. Pipes debugger displays the items ok but refuses to work with them in any way. As far as I know the only modules in Pipes that manipulate the names of the items are the Item Builder/RSS Item Builder and the Rename Operator. Neither one of those modules would recognize the items inputted. Here is a link to the pipe if that will help:

http://pipes.yahoo.com/paulspipes/9980d338e99794b9fb3d31cbba289670

MQL does not offer any functionality for manipulating the results in such a way. However, you can fairly easily add this functionality by using a layer of code on top of your query to traverse the resulting JSON data structure and replace the "/" characters in all the keys and values it find.

You can use Acre, the freebase development platform, to do this easily: http://www.freebase.com/apps

(check out the video on the right)

I think you need a Pipes expert to help you out here. Have you asked the Yahoos?

If you're going to go the Acre route, have a look at Art van Hoff's RSS app. It's so old that it's been broken by Acre changes, but I repaired the most obvious problems and put a copy at http://rss-copy.tfmorris.user.dev.freebaseapps.com/ The Google Maps won't work since the API key is bound to the web site address which is different, but there should be enough there for you to quickly figure out how to generate a GeoRSS feed.

There's no specific restriction on query rate as far as I know, but it benefits everyone to be as efficient with queries as possible.

MQL is pretty powerful, so there's probably a way to collapse things and get all the information that you need in a single query, unless you're doing something quite exotic. One thing you can do, which you might not be aware of, is to request linked properties in a single query,

For example, you could do

"type":"city",
"name":"Boston",
"mayor" : {"birthplace" : {"geolocation":{"longitude":null"}}}

to get the longitude of the mayor's birthplace for all cities named Boston. (Those are made up property names, so this query is just for illustration purposes and won't actually run)

Sorry about that, the example is wrong. It should use the fully qualified property name: /music/album/artist $("#ent_exp").suggest({type: '/music/album', mql_filter: [{"/music/album/artist": [{"id": "/en/carly_simon"}] }] })

Example 5 in the docs is correct. I've removed the incorrect example in the next version of the docs. Thanks!

Thanks!

Here is another question that I could not find in the documentation for Suggest: Is it possible to do a compound mql_filter.

For example, could I ask for the albums from carly_simon OR janet_jackson?

Perfect! Thank you so much for your help.

When using this method for Suggest, I noticed (and saw it noted in your docs) that using mql_filter impacts performance considerably. My goal is to find actors whether they were in film or television, but it looks like this would be prohibited because the types are exclusive (/film/actor & /tv/tv_actor) so I would not be able to use the filter in the same way.

Should I create a new base with type "performance" (for example) and combine the two types of data programatically in my own base?

Thank you both for your reply. I have done both types of queries, but my question is Suggest-specific. I hope I can clarify.

I understand that I can use mqlfilter_ to provide a union of music artists. For example:

// Both Jewel & Leona Lewis have an album named "Spirit"
   $("#ent_exp")
       .suggest({
          type: '/music/album',  
          mql_filter: [{"/music/album/artist": [{"id|=": ["/en/leona_lewis", "/en/jewel"]}] }]
   });

This works because both Jewel and Leona are of type /music/album/artist. However, I'd like to list all the films OR TV shows that Brad Pitt was in. I know that I can list any films or TV shows like this:

 // Any film or TV show
   $("#ent_exp")
       .suggest({
          type: ['/film/actor', '/tv/tv_actor'],
          type_strict: 'any'
   });

But when I try to constrain it to Brad Pitt, I'm not sure of the syntax, or if it's possible because Brad Pitt is of a different type depending on film or TV:

 // Brad Pitt films or TV shows
   $("#ent_exp")
       .suggest({
          type: ['/film/film', '/tv/tv_program'],
          mql_filter: [{"/film/actor": [{"id": "/en/brad_pitt"}] OR {"/tv/tv_actor": [{"id": "/en/brad_pitt"}]}]    // I know this syntax is wrong
   });

There have been a number of issues with the pipeline that reconciles Wikipedia articles with Freebase topics in the past, particularly to do with articles which have been merged and split multiple times. I believe the Stockholm topic has been affected by this issue.

A new version of this pipeline has recently been rolled out which should have fixed most of these issues, so the live data you'll see on freebase.com will be from this pipeline. However, the WEX dumps are older than the new pipeline, so you'll see the incorrect data there.

Hope this helps.

I don't think any problems with IDs should affect the stability of Freebase GUIDs. Having said that, I believe the original poster was refering to the dump which is described as "a processed dump of the English language Wikipedia." (emphasis added).

WEX (or something very similar to it) is an input to Freebase, so any GUIDs that it uses are disjoint from Freebase's. Of course, the IDs aren't really as "globally unique" as they might be if there's overlap between the ID spaces of Freebase and Wex.

If you were to work with the actual Freebase data dumps (The Freebase Wikipedia Extraction (WEX) is a processed dump of the English language Wikipedia), the IDs should match up (and it's bug if they don't).

Here's the formatted code, I didn't see the Markdown business before posting…

    [{
        "!pd:/location/location/contains": [{
            "id": "/en/san_francisco_bay_area"
        }],
        "name": null,
        "type": "/location/us_county",
        "/location/location/contains": [{
            "type": "/location/citytown",
            "name": null,
            "!sv0:/location/mailing_address/citytown": [{
                "!sv0:/business/company/headquarters": [{
                    "optional": true,
                    "name":     null
                }]
            }],
            "!sv1:/location/mailing_address/citytown": [{
                "!sv0:/organization/organization/headquarters": [{
                    "optional": true,
                    "name":     null
                }]
            }],
            "!sv2:/location/mailing_address/citytown": [{
                "!sv0:/education/educational_institution/address": [{
                    "optional": true,
                    "name":     null
                }]
            }]
        }]
    }]

In fact, the entire post looks like a mud fence.

What I'm trying to do is identify keywords that are strongly associated with any arbitrary location (for example: San Francisco, Silicon Valley or Greater Los Angeles).

Now, there's the San Francisco view page, which contains quite a bit of the information that I want, but not all. This is the sort of info I want to gather: names of : * cities/towns containedy * neighborhoods containedby * tourist destinations associated witth * cities/towns containedy * neighborhoods containedby * tourist destinations associated with * transit info (like from the SF view page) associated with * names of schools/colleges/universities located there * companies & organizations headquartered there * names of persons in the news (politicians, celebrities, captains of industry) but not * statistics like population * names of every: * taco stand * copy shop * bong vendor * dogpatch * trailer park * campground * hobo camp

My final use for this is just a long list of words that will be used to filter other data to identify regional items.

Incidentally I've been doing quite a bit of editing on companies (trying to clear up whether a company is still active for the purposes of the headquarters bit), updating various entities "containedby" info and other incidental tweaks. Hopefully I'm doing it right.

I'd prefer not to have to have a different query for every wad of data I want but I also realize that this is quite a bit of data. Here's another example that finds Stanford but not UC Berkeley: http://bit.ly/b8ild (it's a full permanent link to a Query page). It's also really messy

Suggestions? I have some other related questions. Separate post for those though.

There seems to be two parts in you problem:

• identify the entity you want data about (e.g. San Francisco)

• getting the relevant data you want for that entity

For the first problem, a simple solution would be to use the name of the location you want when you do a query. For example:

 [{ 'name' : 'San Francisco', 'type' : '/location/location'}]

will match the san francisco topic.

The properties /location/contains and /location/contained_by can help you navigate the georgraphic hierarchy. Note that the 'contains' property of a given location does not include all the locations that are contained but the immediate descendants. You have to walk that tree on your own using MQL.

For the second problem, once you have the location you want, you can use MQL to get the specific additional properties that you wish. For example, this query: http://tinyurl.com/lblnx3 will give you all the locations contained within san francisco. You can add as many properties you want in the same query, so you don't have to do a different query for every property.

An alternative approach to what you are doing is to use our geo service to get your proximity data: http://www.freebase.com/docs/geosearch

This service also takes mql as parameters so that you can ask it 'Give me all the museums close to San Francisco for example'.

Thanks, but I can DO that already...

My embedded example was my attempt to do the first part, that is determine all the locations I was interested in (it does find all the counties in the SFBA and each of the cities each county contains. Then using the address city/town, I was trying to find corporations, organizations and schools, but it doesn't do a very good job.

That's what I need help with.

You'll need to use some of the more advanced capabilities of MQL to do this kind of thing. A query along the lines of this will give you the most recently updated links from any /film/film topic (I've just realised you'll have to extend the query a bit to get deletions of links as well). If you then repeat that query with the "type":"/film/film" clause in the target clause rather than the source clause, you'll get all links made to /film/film topics.

That leaves values, which may be harder. For a specific topic, you can use /type/reflect/* pseudoproperties (see section 3.7.3 of the MQL Reference Guide), but I can't see a way of doing that for all /film/film topics.

I think this is a domain-specific question, the answer to which will vary with what you're looking at. In the case of Ocean's Eleven, the remake is flagged as being adapted from the original. The same is (now) true of the Italian Job as well.

In other cases, say this book and this book, there is no connection between the two. I'm not sure what sort of answer you're looking for here.

I think adaption and adapted work are your best bet.

The namesake and name source types might also be useful for things named after something else.

There are certain properties in each type that are marked as disambiguators - which is exactly what you want. The freebase suggest (this is the drop-down that tries to auto-complete your input in the search box) flyout is supposed to show you those properties, so if you are hovering over an item in the returned list, you should be able to see the difference.

 Of course this will only work if these propertiesare filled out. In the case they are not, your last resort is to look at any wikipedia or other external links that are attached to them. 

If you fully qualify your property id it will be optional:

http://tinyurl.com/l9jvha

[{

  "id" : "/en/the_godfather",
 "/music/album/artist" : [{ "name" : null }]
}]

You can also specify 'optional' : true explicitely in your clause.

$("#searchtype").suggest({
    type:"/type/type",
    mql_filter:[{
      "/type/type/domain": {
        "key": [{"namespace": "/"}]
      }
    }]
  })

 should get you what you want.

That worked.  Thanks Will!

Jim Weaver

There are a few different porn related genres (e.g. Softcore porn, gay porn, etc), so you'll want to make sure to exclude them all.  There's a huge amount of gay porn which is in the gay porn "base" so you could exclude anything which is part of that base.  You could use the film rating, although I doubt that's very well populated in the database.

You'll probably end up having to use a set of hand tuned heuristics which get you "close enough" to the effect that you're trying to achieve.  Of course, the first thing you'll have to do is decide what you consider "porn."  For example, a movie like Showgirls has a NC-17 rating, but also is listed in IMDB and had a general theatrical release.

Thanks. Yes I noticed that there are various generes which I could consider as part of a "porn" group and exlude. But, roughly, half of the movies in the base do not have any genere attached. Rating is much more sparingly used and doesen't help.

And many of those non "tagged" entries do have explicit and quite "graphic" tiles ...

I can of course consider only the films tagged by genre, with genres outside the porn "domain", but this will reduce the base to 1/3, rougly, or even less ...

If you're worried about "naughty words" in the titles, then you could create a blacklist of words which cause films to be excluded.  Of course that won't catch the ever popular double entendres.

Another approach to the problem would be to work on enhancing the database so that all the movies have a genre attached.  Create a work flow and set of tools, then recruit some people to help you do all the tagging (or just pay folks to do it).

I think that this might be solvable, at least for films that are clearly pornography.

First, I'm told by Jon Reitsma that everything in the Gay Porn base has the genre "/en/gay_pornography".

Next, I know that most pornographic actors in Freebase have the profession, "/en/pornographic_actor", thanks to a mapping from the equivalent category in Wikipedia.

This means we could mod Typewriter to create a data game to crowdsource the marking of pornographic films as such so they can be excluded from queries.

I created this query which displays films not marked as pornography that has actors with the profession pornographic actor. 

http://tinyurl.com/mprjhb

It would need some refinement to also exclude films marked with subgenres of Pornographic movie as well.

That's very useful

I'll try and work a bit to adapt it to all subgenres and post a few results

Thanks

That query returns things like Scary Movie 4 and Toxic Avenger due to cross over actors.  They're not great movies, but they're not porn by any stretch of the imagination either.

On the other hand, a movie like Aunt Peg *is* porn, but you can't tell from the title (unless you're a huge porn fan) or from any information which is in Freebase, so you'd have to do external research to exclude it.

Clearly, the description for Scary Movie 4, Toxic Avenger (and pretty much any Troma production) is going to be caught by this query.  This is why it would be the input to a typewriter-like tool -- for human screening.

Regarding Aunt Peg -- there will always be edge cases.  I think we should at least get started fixing the most egregious cases and then work out how to handle the outliers later.  

Here's an updated version of the query: http://tinyurl.com/klzy5o

I did some checks and will be doing more on the topic

As far I was able, this is the actual full list of porn-related genres:

http://tinyurl.com/nnzrwu

There are some other generes that actually are borderline, though

Currently, it returns 6396 movies. 

Considering the nearly 73K film/film entries in the base, it's just a tiny amount ...

BTW, I used names instead if ids. Does this in your opinion have any impact on performances?

Also, the query on people's profession actually returns a few thousand movies, so id does not seem a definitive solution.

For the sake of precision I did this query:

[{
  "type":   "/film/film",
  "starring": [{
    "actor": {
      "/people/person/profession": {
        "id": "/en/pornographic_actor"
      }
    }
  }],
  "return": "count"
}]

 and it returns 4861 entries

A few more tweaks:

- There are 131 Adult Entertainers who don't have Porn as their profession, so you probably want to exclude them by type instead

    "a:actor": {
      "type": {
        "id": "/base/adultentertainment/adult_entertainer"
      },
      "optional": "forbidden"
    }

-  There are some pornos that only have a title and director.  Most them seem to be actor/directors, so you could exclude them using

  "directed_by": {
    "/people/person/profession": {
      "id": "/en/pornographic_actor"
    },
    "optional": "forbidden"
  },

- You could add various adult awards and film festivals as exclusion criteria if you're still getting undesirable results.

@naram: ids are always better than names, but in this case it probably wouldn't make a huge difference.  Also, 6396 movies sounds about right in terms of proportion.  I imagine that there are a small number of thousand other films that are not yet marked with a pornographic genre.

@tom: I agree that there are additional heuristics we should bake in to the query.

My hope is to create a typewriter-like app that takes these as inputs to generate a work queue.  Users would mark films as the "pornographic movie" genre.  

When people want to exclude these from applications, they could use a query similar to the one that naram gives above.  Or, alternatively, we could create a denormalization that flattens the genre hierarchy so that the query is simpler.  It might take the form of a node called "Porn sentry" that points to all films that have any of the genres listed in naram's query.  It would be updated nightly and only writeable by the bot.

Ok, I've tried a few tweaks and the results are not that much positive: too many false negatives or false positives, alas.

One of the original options was to develop some tools and manually catalogue the movies. If I can make a "tool" good enough, then this is what I'll try.

Here's my idea.

1) create an environment (domain?) I can write to without overwriting or otherwise disturbing the main film/film base. A sort of test base ...

2) create an applicaton in acre that shows the movies (may be in reverse date order so to start with the most recent) in the base with a few info and links (freebase itself, optionally wikipedia and IMDB) so that we can asses the content of the file. Then is needed something like two buttons like "Porn" - "No Porn". Pressing the button just writes in the test base associating the genre info to a movie ID

3) create a way to link this test infoand the base so that it is possible to filter out the movies already categorized as "porn" or "no porn"

It will take time and a few people but in this way we could try to solve the issue.

As an alternative, we could add to genre=porn to the movies directly in the film base. But I'm not sure if this is feasible. Also I'd prefer not to mess things in the general base, even if it were possible.

Now, I'm very new at this kind of development.  First: is either one of the previous solution possible? Second: will it be complicated to do? Can somebody give me some starting tips?

Thanks

I think you and Robert are basically describing the same kind of kind of application.  It's certainly feasible to do.  It's just a matter of a) doing the programming and b) recruiting/hiring folks to use the app to generate the data.

You should be able to separate the films into three categories: 1) porn, 2) not porn (films that have G ratings, etc), and 3) unknown.  It's only those in category 3 that you need to worry about handling manually.

Names of topics derived from Wikipedia usually have equivalents in other languages, but I'm not sure how much other information is available in non-English languages.

The web client is English only, but if you use the MQL query language, you can access the underlying data in whatever languages are available.  The following query will return the name of the Germany topic in a) the default (English), b) German, and c) all available languages, tagged with their language.

http://www.freebase.com/app/queryeditor?q=[{%20"id":%20"/en/germany",%20"name":%20null,%20"n1:name":%20[{%20"lang":%20"/lang/de",%20"value":%20null%20}],%20"n2:name":%20[{%20"lang":%20null,%20"value":%20null%20}]%20}] 

[{
  "id":   "/en/germany",
  "name": null,
  "n1:name": [{
    "lang":  "/lang/de",
    "value": null
  }],
  "n2:name": [{
    "lang":  null,
    "value": null
  }]
}]​

 See http://mql.freebaseapps.com/ch03.html#history  Something along the lines of the following two queries which get the incoming and outgoing links, both valid and invalid (ie deleted or overwritten), sorted in reverse order by time:

[{
  "type":          "/type/link",
  "valid":         null,
  "master_property": null,
  "source": {},
  "target_value":  null,
  "target":{"id":"/en/penguins"},
  "timestamp":     null,
  "sort":          "-timestamp"
}]

[{
  "type":          "/type/link",
  "valid":         null,
  "master_property": null,
  "source": {"id":"/en/penguins"},
  "target_value":  null,
  "target":{},
  "timestamp":     null,
  "sort":          "-timestamp"
}]

When you say 'base' are you talking about a Freebase base?  You're aware that all Freebase bases are basically public, right?

One technique that they use for the Wall Street Journal film reviews is to use the as_of timestamps to basically snapshot the database in a state that's acceptable to you.  This means you need to fix up the "real" base to the what you want it to look like, but you can then "freeze" it until time you come back and do the next review.

If you're talking about having your staging database outside of Freebase, you can obviously manage things in whatever way suits your purposes.

Hiya.

The "trans" service is what you're after here; please see the relevant section of the MQL documentation.

Hope this helps.

All text strings are encoded in UTF-8.

In the example you gave, that is being displayed as expected; the topic really does have the name "Old House at S(square root)(registered trademark)vres". This is obviously not the correct name for the topic, but that's "just" a data problem which can easily be corrected.

Hope this helps.

Yes, this particular page look fine now, thanks. But I do not think that you fixed the problem in general - looking at the data I saw that specifically  french names share this problem. I guess it might have been some encoding problem as you initially imported the data.

e.g.

http://www.freebase.com/view/en/charles_n_gre 

http://www.freebase.com/view/en/t_te_trois_quarts_et_profil_buste_de_diego 

 It would be great if you could fix the data in the database and update the dumps.

Thanks,

Elen

It looks like the person doing the entry was probably using a Mac Roman character encoding, but claiming it was UTF-8.  There are actually more like 300+ entries messed up, but it's still a vanishingly small percentage of the database.  Nonetheless, I've reported it as bug https://bugs.freebase.com/browse/FREEBASE-888  Whether or not it'll get fixed before the next quarterly data dump or not will depend on how it gets prioritized relative to other work, but anyone can fix it themselves using the transformation described in the bug report.

BTW, all the web links back to the SF MOMA web site are broken as well.

as_of_time also works in the URL:

http://www.freebase.com/view/en/on_the_road?as_of_time=2008-05-01

http://www.freebase.com/view/en/on_the_road

The TSV data dumps don't currently include any keys. If you'd like to download keys in bulk, you can use the Link Export (1.6 GB, http://download.freebase.com/datadumps/). In that file, IMDB and Netflix keys will look like this:

/guid/9202a8c04000641f8000000000009e89 /type/object/key /authority/imdb/title tt0083658
/guid/9202a8c04000641f8000000000009e89 /type/object/key /authority/netflix/movie 70053131

Will that work for you? Let me know if there's anything else I can help with.

I think it should work, depending on the formatting within the dump files, and how long it takes me to successfully pull down a 1.6 gig file...

I'll post back when done. Thanks! 

OK. It took about an hour to download, which was faster than I expected. It also took about an hour to unzip, which was quite a bit more.

I am afraid that I now have a 25 gig file here that I am not sure how to approach. I think i still have much reading to do on your site (or somewhere) to figure out how I can start parsing some data out of that file.

Do I need to run a graph database over here locally for this? 

Or, let me back up. What I am really trying to get is something that would normally be returned by a SQL query that looks something like this:

SELECT imdb_id, netflix_id WHERE imdb_id IS NOT NULL and netflix_id IS NOT NULL; 

Is there a simpler way to get this subset of data easily? 

You don't need any kind of database to start using this data. If you're using a Unix-like environment, the simplest way to get at the keys you want is to directly filter it as a text-file for the key-namespaces you want using grep, like this:

grep "/authority/imdb/title" freebase-datadump-quadruples.tsv > imdb_keys.tsv

In this file, keys are represented as "id property key-namespace key-value". You could also filter by "/type/object/key" to get all the keys in Freebase.

That worked like a charm with the imdb data giving 19,558 records.

The netflix version returned 0 records, so I shortened the "search for" string down to "/netflix/" and got a lot of records with:

/user/hal/netflix/movie  (23,929 records)

 and   /user/hal/netflix/role  (8,397 records)

Are these the netflix records you referred to above as:

guid/9202a8c04000641f8000000000009e89 /type/object/key /authority/netflix/movie 70053131

?

Correlating the /user/hal/netflix/movie records with the 

/authority/imdb/title records, i know have 11,189 matches -- that is 11,189 movies in the freebase dump file that have both an imdb id and a netflix id.

Does that sound about right? 

As I write this, there are 11,182 topics on the live graph which have both IMDB and Netflix title keys, so 11,189 from the slightly out of date data in the dumps is a perfectly reasonable number.

As an aside, 19 of those topics are not currently typed as a film. The first one of those that I looked at, The Human Factor, is a book and needs to be split into two topics; similar things may well apply for the other 18. Just a small health warning :-)

Is this your list of topics?

You need to edit the view (blue-colored text 'Edit view' link below the 'Share This' button).

Click on the 'advanced filters' in blue text under the 'Filters' and change the first timestamp value from 'Jun 15 2009 12:00 am' to that of 'Jun 16...' etc.

See this image as an example on restricting the topics to ones you just added.

Now see the gear icon (next to the Add more button with the green "+") click it and you have an option to do many great things, one of which is to add the topics of this particular view to a type. Click on the 'Add to another type' and enter in the type name (Second Life Region?) and that should do it..

 Example picture here.

@librarianavenger, I'm sorry to hear about your list importer implosion :(  Unfortunately this is a known bug, that the list importer degrades in performance and starts flaking out when you put lots of topics in at once.  In my experience, it's best to import in batches of around 50 at a time, to be on the safe side.  I'd love to get the tool rewritten in Acre or somthing, and fix it up; I was talking to jdouglas about this just yesterday in fact (Cc'ing him so he sees this thread).

 The workaround gmackenz suggests should work, however.

And can I just say how pleased I am to see you're working on the SL base again?  YAY.  Also, did you know we have another Hack Day coming up on July 11th?

gmackenz - thanks for the excellent workaround - sadly I seem to get an error when either filtering things I created on the 16th, or navigating there via activity history. I'll keep trying and hope something shakes loose.

 Hi Skud!!!! Freebase is totally feeding the librarian compulsion. We should get brunch this weekend. I miss you guys!

That's because Government Position Held is a Compound Value Type (CVT), which doesn't have a name itself. You'll need to expand out the properties you're interested in from the CVT, for example like this.

Hope this helps.

Thanks that worked.  Could you help me now with this... I need a disjunctive query, for example to get the governing officials from San Francisco OR Edmonton... I tried this, but does not work

[{
  "type": "/government/governmental_jurisdiction",
  "id":   null,
  "name": { "San Francisco", "Edmonton" },
  "governing_officials": [{
    "office_holder": null,
    "office_position_or_title": null,
    "from":          null,
    "to":            null
  }]
}]

Thanks in advance

You need the "one of" operator:

 [{
  "type": "/government/governmental_jurisdiction",
  "id":   null,
  "name|=": [ "San Francisco", "Edmonton" ],
  "governing_officials": [{
    "office_holder": null,
    "office_position_or_title": null,
    "from":          null,
    "to":            null
  }]
}]

Still not working - this time it correctly identifies the above mql as a write, but returns the error /api/status/error/mql/type, "Type /type/object does not have property query".

I've created CLI-8201 in JIRA. 

It doesn't understand the envelope, so if you remove the outer { "query" : ... } and just keep the array, like so 

http://tinyurl.com/prta7m 

it should work. It just did for me. 

• Sorry, the format of the above writing is kind of messy. To see the function code (e.g. expand(....)) please visit here:

• http://acre.freebase.com/#app=/user/namwkim/gendata&file=data

There's already a genealogy graphing application here http://genealogy.alexander.user.dev.freebaseapps.com/index?person=%2Fen%2Felizabeth_ii_of_the_united_kingdom

which seems to work fine with Elizabeth II.  Why not just use it?

A common programming problem with recursion is a logic flaw which causes the recursion to be infinite.  Are you sure your recursion terminates?  I'd start by adding debug logging to your app so that you can tell what progress it's making.  That should narrow things down to the specific section of the code that's got the bug in it.

My version is the extended version of the application. The application does not get the whole genealogy data, but just a subset. Mine is trying to get the whole data. you can try charles darwin on both application and see difference.

 I am sure my app terminates. You may not look at the code carefully, but I am maintaining a Set containing all the visited nodes. So, I never visit already visited nodes. 

So instead of what genealogists call an "hour glass chart" showing direct descendants and direct ancesters, you want an "everyone chart" that includes all siblings/cousins as well?  You understand that this will quickly make the graph layout exceeding difficult, right?

Why do you think the problem is too many queries?  Since you're sure that the recursion terminates, presumably you know how long it takes.  How much time is being spent collecting data?  How much time is being spent by GraphViz trying to lay things out?  Are the timeouts dependent on purely the number of nodes in the graph or also the complexity of their interconnections (edges)?  Do you limit the number of nodes that you attempt to lay out?  What happens if you change the type of graph that you're asking GraphViz to plot?  What happens if you skip the GraphViz stage altogether and just output a list of names and dates?

If you could get the attention of a Metaweb staffer who knows Acre, you could ask them:

• what the time limit is for constructing a page using Acre
• whether the limit can be increased
• whether the timer can be reset by some simple technique such as using a periodic "Working..." message

Thanks so much for such a great advice.

I am collecting genealogy data for my research. that is why I was trying to get the whole genealogy (e.g. ancestors of descendant etc)

I asked a Metaweb staffer about increasing time limit, but did not get a reply. : )

I talked to professor and we decided to narrow focus to parent-child relationship and spose-relationship. So I don't think I need to get the whole genealogy anymore. 

 Anyway, thank you so much again.

Nam

This is really a Python programming question.  Check any Python tutorial for how to use dictionaries (dict(), {}, etc).

Having said that, the example that you gave in your other note over in General Support sounds dangerous to me in that it's going to potentially give the user the ability to generate lots of new topics without the verification that's implicit in manually selecting things from an autocomplete box.

Remember that even if things are in a user-specific domain, they are visible globally, so it's going to be confusing to have lots of duplication.

Can you describe a little bit more about the type of data that you want users to be able to upload in this way?