Sample Nested Queries

the whole fb feed - all page posts, its likes, comments, likes of the comments, subcomments, likes of the subcomments

{
  "id": 1,
  "name": "scrape",
  "type": "nested-query",
  "query": {
    "path": "feed",
    "fields": "caption,message,created_time,type,description,likes{link,name,username},comments{message,created_time,from,likes{link,name,username},comments{message,created_time,from,likes{link,name,username}}}",
    "ids": "<page_id>"
}

extract page metrics `page_impressions`, `page_fans` and `page_engaged_users` for last 5 days

{
  "id": 2,
  "name": "page_insights",
  "type": "nested-query",
  "query": {
    "path": "",
    "fields": "insights.since(5 days ago).metric(page_impressions,page_fans,page_engaged_users)",
    "ids": "<page_id>"
  }
}

According to the last facebook graph api all insights metrics that are needed to extract must be explicitely listed in the query, i.e., there is no general get-all-metrics data query type of call.

extract posts metrics `page_posts_impressions` and `post_impressions` for all posts:

Note day we are using since(now) specification of insights time to get the most recent values, otherwise it may paginate over small periods of time and consume lots of request to facebook api.

{
  "id": 3,
  "name": "posts_insights",
  "type": "nested-query",
  "query": {
    "path": "feed",
    "fields": "insights.since(now).metric(page_posts_impressions,post_impressions)",
    "ids": "<page_id>"
  }
}

You can try the examples above by calling Facebook Graph api directly in a http client(e.g. Postman) as follows: GET https://graph.facebook.com/<api_version>/<path>?fields=<fields>&ids=<ids>&access_token=<access_token> You can get access token in Graph Api Explorer.

Configuration

Facebook Graph API

This extractor extracts data from facebook graph api: https://developers.facebook.com/docs/graph-api/ . You can try it live here: https://developers.facebook.com/tools/explorer

Nested Query

Lets say that in Facebook Graph Api every endpoint represents a node in a graph. Example of a node could be /me - ie user info, me/posts - ie posts of the current user. To get data from a particullar endpoint one can make typical REST api call GET me/posts or make a nested api call that basically allows to extract the whole subtree of a node. for example GET me?fields=posts,comments,likes extracts all posts comments and likes of an id(in this case it is me - current user). For more info see Making Nested Queries in https://developers.facebook.com/docs/graph-api/using-graph-api#reading

Configuring nested query

In configuration under parameters there is an array of queries(see sample configuration). Each query besides obvious properties such as id, name, type(currently only nested-query type), disabled also contains object query with the following properties:

path : enpoint url so the absolute url will be like graph.facebook.com/version/path. Typically it is endpoint feed. Can be an empty string if we want to start extracting from the "root" node that is the page itself.
fields: fields parameter of the graph api nested-query
ids: comma separated list of ids(typically page-ids) that will be prepended with path. It is also a parameter of graph api. If empty string than all ids from accounts object will be used. Can also be completely removed from the query.
limit - size of one page(response). Default is 25, maximum 100. Useful when fb api returns error that the request is "too big" - in such case use smaller limit. This parameter also affects the total number of request made to fb api.
since - relates to the created_time of path parameter i.e., if path is "posts" then it takes all posts with created_time since the specified date in since parameter. If path is empty then it does not have any effect. Can be specified relatively, e.g. 10 days ago.
until - same as since above but specifies date until data with created_time date.

The most important parameter is fields - tells what is going to extract. so here are few hints:

you can specify additional params of a node with dot e.g to specify since and limit params of posts: posts.limit(100).since(2016-12-24){message,likes,comments{comments}}
you can also specify date range using since or/and until that accepts unix timestamp values(in seconds) or date in format yyy-mm-dd or relative values: e.g. all posts posted in last 10 days posts.since(10 days ago){message,likes,comments}
relative date specification is parsed by this cool PHP function strtotime
if an object(e.g comments) does not have nesting specified it will extract some of its columns but once the nesting is specified e.g commens{likes} then one has to explicitely specify all its column in the nesting e.g. comments{from,message,created_time,likes}
for each row id is extracted autmatically and no need to be specified in the query
to extract all posts with its comments, subcomments,likes and sublikes the query would look like this:

{
  "fields": "posts{message,story,created_time,likes,comments{from,message,created_time,comments,likes}}",
  "path": "",
  "ids": "<some_page_id>"

}

Sample configuration:

Note that you can specify facebook api version via api-version parameter. Default is v2.8.

{
  "storage": {},
  "parameters": {
    "accounts": {
      "<pageId1>": {
        "id": "<pagId1>",
        "name": "my fancy page",
        "category": "entertainment"
      },
      "<pageId2>": {
        "id": "<pageId2>",
        "name": "keboola",
        "category": "software"
      }
    },
    "api-version": "v2.8",
    "queries": [
      {
        "id": 1,
        "name": "qname",
        "type": "nested-query",
        "query": {
          "path": "feed",
          "fields": "message,story,likes,comments{from}",
          "ids": "<pagId1>,<pageId2>"
        }
      }
    ]
  },
  "authorization": {
    "oauth_api": {
      "id": "{OAUTH_API_ID}"
      }
    }
  }

Result tables description

For each query extractor generates a number of tables prefixed with query name. Each table represents one type of node so typically tables would be queryname_post, queryname_likes, queryname_comments queryname_insights. Same nested structure type will be in the same table. So for example comments and subcomments will be in the same table comments. Every table has different columns but the following columns will always be the same:

id: unique id of the row, some tables may not have it, e.g. ads insights
parent_id: the parent node id, e.g. comments parent-id refers to a post row id
ex_account_id: top root source under the extraction is beign done, it could be a page id or ads account id and it is specified in config under ids property.
fb_graph_node: describes the row "vertical position" of the resulting tree. e.g for comments it will be page_feed_comments, for subcomments(i.e. comments of comments) it will be page_feed_comments_comments etc
insights data objects will be flatten into columns key1, key2 and value along with columns metric name, title, description etc

Authorization

Register component api to oauth-bundle-v2 by calling POST to https://syrup.keboola.com/oauth-v2/manage with manage-token, storage-api token in the header and body:

{
  "component_id": "keboola.ex-facebook-insights-v2",
  "friendly_name": "Facebook Insights",
  "app_key": "xxx",
  "app_secret": "xxx",
  "oauth_version": "facebook",
  "permissions": "manage_pages,public_profile,read_insights,pages_show_list",
  "graph_api_version": "v2.8"

}

more info about authorization registration here: https://github.com/keboola/oauth-v2-bundle

Development

The app is written in Clojure(1.8), evaluated and build via Boot-clj which requires Java Development Kit. To try the app locally check target commands in the Makefile For example to build and run the app locally type from the repo root:

make build-jar run-jar

youonf/ex-facebook-graph-api