From BaseX Documentation
Jump to navigation Jump to search

This article is part of the Advanced User's Guide. It is about the usage of BaseX for processing and storing the live data stream of Twitter. We illustrate some statistics about the Twitter data and the performance of BaseX.

As Twitter attracts more and more users (over 140 million active users in 2012) and is generating large amounts of data (over 340 millions of short messages ('tweets') daily), it became a really exciting data source for all kind of analytics. Twitter provides the developer community with a set of APIs for retrieving the data about its users and their communication, including the Streaming API for data-intensive applications, the Search API for querying and filtering the messaging content, and the REST API for accessing the core primitives of the Twitter platform.

BaseX as Twitter Storage

For retrieving the Twitter stream we connect with the Streaming API to the endpoint of Twitter and receive a never ending tweet stream. As Twitter delivers the tweets as JSON objects the objects has to be converted into XML fragments. For this purpose the parse function of the XQuery JSON Module is used. In the examples section both versions are shown (tweet as JSON and tweet as XML). For storing the tweets including the meta-data, we use the standard insert function of XQuery Update.

Twitter’s Streaming Data

Each tweet object in the data stream contains the tweet message itself and over 60 data fields (for further information see the fields description). The following section shows the amount of data, that is delivered by the Twitter Streaming API to the connected endpoints with the 10% gardenhose access per hour on the 6th of the months February, March, April and May. It is the pure public live stream without any filtering applied.



Day Description Amount
Mon, 6-Feb-2012 Total tweets 30.824.976
Average tweets per hour 1.284.374
Average tweets per minute 21.406
Average tweets per second 356
Tue, 6-Mar-2012 Total tweets 31.823.776
Average tweets per hour 1.325.990
Average tweets per minute 22.099
Average tweets per second 368
Fri, 6-Apr-2012 Total tweets 34.638.976
Average tweets per hour 1.443.290
Average tweets per minute 24.054
Average tweets per second 400
Sun, 6-May-2012 Total tweets 35.982.976
Average tweets per hour 1.499.290
Average tweets per minute 24.988
Average tweets per second 416

Example Tweet (JSON)

  "contributors": null,
  "text": "Using BaseX for storing the Twitter Stream",
  "geo": null,
  "retweeted": false,
  "in_reply_to_screen_name": null,
  "possibly_sensitive": false,
  "truncated": false,
  "entities": {
    "urls": [
    "hashtags": [
    "user_mentions": [
  "in_reply_to_status_id_str": null,
  "id": 1984009055807*****,
  "in_reply_to_user_id_str": null,
  "source": "<a href=\"http:\/\/\" rel=\"nofollow\">twitterfeed<\/a>",
  "favorited": false,
  "in_reply_to_status_id": null,
  "retweet_count": 0,
  "created_at": "Fri May 04 13:17:16 +0000 2012",
  "in_reply_to_user_id": null,
  "possibly_sensitive_editable": true,
  "id_str": "1984009055807*****",
  "place": null,
  "user": {
    "location": "",
    "default_profile": true,
    "statuses_count": 9096,
    "profile_background_tile": false,
    "lang": "en",
    "profile_link_color": "0084B4",
    "id": 5024566**,
    "following": null,
    "protected": false,
    "favourites_count": 0,
    "profile_text_color": "333333",
    "contributors_enabled": false,
    "verified": false,
    "description": "http:\/\/",
    "profile_sidebar_border_color": "C0DEED",
    "name": "BaseX",
    "profile_background_color": "C0DEED",
    "created_at": "Sat Feb 25 04:05:30 +0000 2012",
    "default_profile_image": true,
    "followers_count": 860,
    "geo_enabled": false,
    "profile_image_url_https": "https:\/\/\/sticky\/default_profile_images\/default_profile_0_normal.png",
    "profile_background_image_url": "http:\/\/\/images\/themes\/theme1\/bg.png",
    "profile_background_image_url_https": "https:\/\/\/images\/themes\/theme1\/bg.png",
    "follow_request_sent": null,
    "url": "http:\/\/\/5ktAf",
    "utc_offset": null,
    "time_zone": null,
    "notifications": null,
    "friends_count": 2004,
    "profile_use_background_image": true,
    "profile_sidebar_fill_color": "DDEEF6",
    "screen_name": "BaseX",
    "id_str": "5024566**",
    "show_all_inline_media": false,
    "profile_image_url": "http:\/\/\/sticky\/default_profile_images\/default_profile_0_normal.png",
    "is_translator": false,
    "listed_count": 0
  "coordinates": null

Example Tweet (XML)

<json booleans="retweeted possibly__sensitive truncated favorited possibly__sensitive__editable default__profile profile__background__tile protected contributors__enabled verified default__profile__image geo__enabled profile__use__background__image show__all__inline__media is__translator" 
  numbers="id retweet__count statuses__count favourites__count followers__count friends__count listed__count"
  nulls="contributors geo in__reply__to__screen__name in__reply__to__status__id__str in__reply__to__user__id__str in__reply__to__status__id in__reply__to__user__id place following follow__request__sent utc__offset time__zone notifications coordinates" 
  arrays="urls indices hashtags user__mentions"
  objects="json entities user">
  <text>Using BaseX for storing the Twitter Stream</text>
  <source><a href="" rel="nofollow">twitterfeed</a></source>
  <created__at>Fri May 04 13:17:16 +0000 2012</created__at>
    <created__at>Sat Feb 25 04:05:30 +0000 2012</created__at>

BaseX Performance

The test show the time BaseX needs to insert large amounts of real tweets into a database. We can derive that BaseX scales very well and can keep up with the incoming amount of tweets in the stream. Some lower values can occur, cause the size of the tweets differ according to the meta-data contained in the tweet object.
Note: The AUTOFLUSH option is set to FALSE (default: SET AUTOFLUSH TRUE)

System Setup: Mac OS X 10.6.8, 3.2 GHz Intel Core i3, 8 GB 1333 MHz DDR3 RAM
BaseX Version: BaseX 7.3 beta

Insert with XQuery Update

These tests show the performance of BaseX performing inserts with XQuery Update as single updates per tweet or bulk updates with different amount of tweets. The initial database just contained a root node <tweets/> and all incoming tweets are inserted after converting from JSON to XML into the root node. The time needed for the inserts includes the conversion time.

Single Updates

Amount of tweets Time in seconds Time in minutes Database Size (without indexes)
1.000.000 492.26346 8.2 3396 MB
2.000.000 461.87326 7.6 6997 MB
3.000.000 470.7054 7.8 10452 MB


Bulk Updates

coming soon...