Twitter Streamer is a Python command-line utility to dump Twitter streaming API statuses/filter method data to stdout.
It began life as a testing tool for Tweepy, and to satisfy my curiosity. It's currently in an early beta test state, and needs testing and improvement. (see Known issues, below)
You will need:
- Python 2.7, Tweepy and its prerequisites.
- Your Twitter API keys.
Once you have your API keys, export the following environment variables:
export CONSUMER_KEY="your-consumer-key"
export CONSUMER_SECRET="your-consumer-secret"
export ACCESS_KEY="your-access-token"
export ACCESS_SECRET="your-access-token-secret"
These environment variables must be set correctly for twitter-streamer to function properly. If any of these variables are not available, you'll see an error message and twitter-streamer will terminate.
If you try to use invalid API keys, you'll see something like this output:
StreamListener.on_error: 401
StreamListener.on_error: 401
StreamListener.on_error: 401
...
Basic usage:
python streamer.py [options] "track terms" ...
You can print a usage summary by invoking streamer.py with the -h or --help option.
The positional track parameter provides one or more track search terms for the Twitter statuses/filter API. Commas denote an or relationship, while spaces denote an and relationship.
You can provide multiple track parameters, which will expand the search terms.
Please refer to the track documentation for specific limitations and usage examples. See this question for help filtering on hashtags using URL encoding.
Stream (filter) statuses containing both car and dog:
$ python streamer.py "car dog"
Stream statuses containing either boat or bike:
$ python streamer.py "boat,bike"
Stream statuses containing (water and drink) or (eat and lunch):
$ python streamer.py "water drink" "eat lunch"
The default output mode dumps the unprocessed stream's JSON string for each received element, followed by a newline.
Each raw stream element is expected to be a well-formed JSON object, but the stream elements are not separated by commas, nor is the stream wrapped in a JSON list. Therefore, if you expect to parse the entire output of this program's JSON data, you will need to transform it into well-formed JSON, or take each newline-separated element as an independent JSON object rather than treat the stream as a JSON array.
The program will produce CSV output when using --fields (-f) field specifiers.
See Field Output Selectors, below.
As of v0.0.6-dev, you can follow user ids by using the --follow (-F) option.
The --follow option accepts a list of comma-separated integer user id values.
I plan to add user screen_name lookup at some point, in the meantime,
you can use the helper script scripts/lookup-users.py:
$ python scripts/lookup-users.py twitter,tweepy
tweepy: 14452478
twitter: 783214
(Please keep in mind that per the Streaming API docs, following users does not filter results to only those users, rather, it adds the selected users' streams to the incoming stream results.)
As of v0.0.4, you can add location-based search criteria by specifying the --locations
option. The value is a comma-separated list of longitude, latitude pairs that
define one or more bounding boxes to include in the stream. (This implies that
the number of comma-separated --location= values must be a multiple of 4, and
in fact Tweepy enforces this for us.)
Example:
$ python streamer.py -f=place.full_name,coordinates.coordinates,text --locations="-122.75,36.8,-121.75,37.8"
This produces a stream of status updates as CSV, with the place.full_name,
coordinates.coordinates, and text fields (if available). Here is
an example with longitude and latitude obscured in order to protect privacy:
"San Jose, CA","[longitude, latitude]",@user is a boy
"Oakland, CA","[longitude, latitude]",@user1 @user2 @user3 @user4 @user5 We are all 1 big happy family #BELIEBERS that what we are 24/7 were #STRONG
"California, US",,i'll be here awhile #311
where longitude and latitude are floating-point numbers representing Twitter's
notion of the Tweet's location.
There are several fields that are used to determine location: coordinates,
place, and geo (the latter is deprecated.) Note that
Twitter prioritizes the location data by preferring coordinates over place when
determining if a tweet should be included based on geo location.
Note that including --locations
parameter will not further filter other search terms (such as track keywords)
-- per the location reference, it acts as an OR when
combined with track keywords.
See the Twitter's Tweets structure reference for more information
about location-based information, and the location for more
about the location parameter.
Recent development versions (0.0.5-dev and higher) support a new option:
--location-query. It allows you to reference a Twitter Place name, and
automatically use the resulting coordinates as the value of the --location
parameter. (Currently the resulting --location-query bounding box overrides
any values passed in the --location command line option)
The --location-query value is passed to the Tweepy API.geo_search method
(which uses the Twitter [geo/search][twitter-go-search]) as the query parameter.
The value is case-insensitive, and the value must match an existing Twitter Place
full_name field. In general, you can use this pattern:
--location-query="{city-name}, {state-abbrev}"
--location-query="{state-name}, US"
where {city-name} is a well-known city name, {state-name} is a full state name, and {state-abbrev} is a standard two-letter state abbreviation. You can search for other types of names, but you'll have to do you own research or experiment to find valid values.
Example:
--location-query="San Jose, CA"
--location-query="California, US"
Matching is done without regard to spaces, but
the Twitter API might fail to return expected matches if you deviate too far from the
pattern shown above. If in doubt, enable full debug logging, by passing
-l DEBUG on the command line.
- Note that there are open issues regarding
the accuracy of Twitter's
locationsfiltering. You may need to do your own post-filtering of the results to ensure they are within the desired bounding box. - Using
--location-query="United States"results in an error because Twitter doesn't return a bounding box for the United States. It does for Canada. Go figure.
The -f (or --fields) parameter allows a comma-separated list of output fields.
The field values will be emitted in the order listed in the given -f
parameter value. Output will be formatted as CSV records.
You can access nested elements by using dotted notation: user.name accesses
the name element of the user object. See Twitter's tweets
structure reference for a list of valid elements.
If you reference a non-existent element, the output column will be empty.
If you prefer to have an error message displayed and terminate processing
specify the -t or --terminate-on-error option.
Example 1: list created_at and text fields for 'elections'
python streamer.py -f "created_at,text" elections
Example results:
2012-11-09 20:26:47 Volatility the Likely Outcome of Elections http://t.co/trmmSpXp #Barron's
2012-11-09 20:26:50 @WHLive then why the president ordered Boeing to release the layoff news AFTER the elections?
Example 2: list user.name and text fields for tweets containing dogs and cats
python streamer.py -f "user.name,text" "dogs cats"
Example Results:
User name 1,Cats and dogs in Mexico. http://t.co/gYJvhdvv
User name 2,I actually like both cats and dogs but I've been an introvert for about 27 years now.
See TODO.md
- Needs a bit of cleanup -- obsolete code remains from prior project and will be refactored or removed.
- Error handling needs work.
The current default mode retries the connection with a delay
in the event of most failures; this keeps Streamer running despite network
problems or other errors.
If you specify the
--terminate-on-errors(-t) option, Streamer will terminate with an error message on errors rather than retrying certain operations. This is a work in progress. - Log messages go to stderr.
- If you receive 401 errors during authentication, ensure your system's date and
time settings are correct. Auth can fail if your clock is out of sync with
Twitter's servers. The
scripts/twitter-time-compare.shscript shows Twitter's server and the local server times for comparison.
##License## (MIT License) - Copyright (c) 2012-2013 Exodus Development, Inc. except where otherwise noted. Please refer to LICENSE.md for the gory details.