Get various information from Facebook Graph API

Finding information from dataset has always fascinated me. And facebook is such a vast reservoir of data, that it's practically a crime for the enthusiasts to leave it untapped.

I've been looking for some way to get different questions answered by facebook.

Like, who has liked my post most? Whose post I liked most? Comments, photos everything can be explored to find these sort of information. There are also a lot of apps for these purposes, but as a developer, you should be able to do it yourself, at least I think so.

I'm going to explain this one: "Who liked your last N number of feeds most?"

In this post, I've explained the way to connect to graph api. Keep in mind that facebook has made the fql obsolete from versions 2.0+. So, being an expert in fql is not going to help much in the future.

Now, facebook has opened up different root nodes in the graph api. They could be found in this link . We are going to use the 'user/feed' node for our purpose. Now, of course facebook is not going to give all data in one request, nor you should expect it to. It returns data in a paged format. There are three types of pagination that facebook uses. Our node uses time based pagination.

Let's look at this code chunk:

import requests
from facepy import GraphAPI
import pprint
from pymongo import MongoClient
from bson.son import SON

client = MongoClient('localhost', 27017)
db = client.likes
like_table = db.like_table

We're going to use mongo a little bit. Due the scope of this post, I'm not going to describe it's installation or usage. This link has very good resources on it.

feeds = graph.get('friends_or_self_id/feed?limit=1')

To find this id, we can simply issue a request in the browser in the address 'http://graph.facebook.com/username'. We could also append parameters like until and since to specify the time period we are concerned about. However, I faced trouble getting them working. So, I've opted for counter instead.

while True:
    try:
        likes = feeds['data'][0]['likes']
        while True:
            try:
                for m in likes['data']:
                    like_table.insert(m)
                likes = requests.get(likes['paging']['next']).json()
            except KeyError:
                print("keyerror in likes pagination")
                break
        if(counter == 30):
            break
        counter = counter + 1
        feeds = requests.get(feeds['paging']['next']).json()
    except KeyError:
        print("keyerror happened")
        break

This shall retrieve the last 30 feeds on the timeline of that user. Then, let's find the most appearance of a particular id. This is where mongo is coming handy.

list = (like_table.aggregate([
    {'$group': {
        '_id': '$id',
        'count': {'$sum': 1}}
    },
    {"$sort": SON([("count", -1), ("_id", -1)])}
]))

for k in list['result']:
    print(k['_id'])
    print(k['count'])

This is just one simple example of the endless possibilities of facebook. I'm a big fan of it just because of it's database. People can practically be 'profiled'.
The code is also hosted in github.

Useful FQL examples

I've been exploring the facebook graph API and fql. FQL seems to be a very powerful tool. Here are some of them which I've been using recently:

1. Finding friends user id by looking up his name:

select name, uid from user where strpos(name, "part_of_your_friends_name") >= 0 and uid in (SELECT uid2 FROM friend WHERE uid1=me())

2. Find the id of a group by its name:

select id, name from profile where name="group_name"

3. Find posts of a user in a group:

SELECT post_id, message, created_time  FROM stream WHERE strpos(field_name,'possible_string') >=0 and  source_id=group_Id and actor_id=user_id  and created_time >=unix_timestamp and created_time < unix_timestamp limit A_NUMBER


//will be continued...

note:
To get the unix timestamp you can use this code in Javascript:

var d = new Date();
d.setMonth(d.getMonth() - 3);
three = Math.round(d.getTime() / 1000); //3 months ago

console.log(three);

Answers to the exercises of the book Natural Language Processing with Python: Chapter 1

from __future__ import division
from nltk.book import *

#author: Anik

Checkout Chapter 2 Excercise Answers

ex 2
print 26**100

ex 3
print ['Monty', 'Python'] * 20

ex 4
print len(text2)
print len(set(text2))

ex 5
romance

ex 6
text2.dispersion_plot(["Elinor", "Marianne", "Edward", "Willoughby"])
Ellenor appears everywhere, and probably Willoughby and Marianne had a relationship

ex 7
print text5.collocations()

ex 8
easy

ex 9
easy

ex 10
my_sent = ["All", "quiet", "on", "the", "western", "front"]
joined =  ' '.join(my_sent)
print joined
print joined.split()

ex 11
phrase1 = ["All", "quiet", "on", "the", "western", "front"]
phrase2 = ["We", "were", "Soldiers"]
print len(phrase1 + phrase2)
print len(phrase1) + len(phrase2)

ex 12
print "Monty Python"[6:12]
print ["Monty", "Python"][1]
second one, tokenized

ex 13
sent = "Flags of our fathers"
print sent.split()
print sent.split()[2][2]

ex 14
sent3 = ['In', 'the', 'beginning', 'God', 'created', 'the', 'heaven', 'and', 'the', 'earth', '.']
indices = [i for i, x in enumerate(sent3) if x == "the"]
print indices

ex 15
bs = [w for w in text5 if w.startswith('b')]
print sorted(bs)

ex 16
print range(10)
print range(10,20)
print range(10, 20, 2)
print range(10, 20, -2)

ex 17
print text9.index('sunset')
print text9[621:644]

ex 18
print sorted(set(sent1+sent2+sent3+sent4+sent5+sent6+sent7+sent8))

ex 19
print len(set([w.lower() for w in text1]))
print len([w.lower() for w in set(text1)])

ex 20
print "ANIK".isupper()
print not "ANIK".islower()
no difference

ex 21
print text2[len(text2)-2:len(text2)]

ex 22
four_letter_words = sorted([w.lower() for w in text5 if len(w) is 4])
print FreqDist(four_letter_words)

ex 23
upper_case_words = [w for w in text6 if w.isupper()]
for word in upper_case_words:
    print word

ex 24
word_list = [w for w in text6 if w.endswith('ize') and 'pt' in w and w.istitle()]
print word_list

ex 25
sent = ['she', 'sells', 'sea', 'shells', 'by', 'the', 'sea', 'shore']
print [w for w in sent if w.startswith('sh')]
print [w for w in sent if len(w) > 4]

ex 26
print (sum([len(w) for w in text1])) / len(text1)

ex 27
def vocab_size(text):
    return len(set([word.lower() for word in text if word.isalpha()]))
print vocab_size(text1)

ex 28
def percent(word, text):
    return 100 * text.count(word) / len(text4)
print str(percent('a', text4)) + '%'

ex 29
too descriptive

Connect to facebook Graph API with python

Yes, this type of posts are all over the world wide web. However, there are thousands of methods and not all of them are going to work for you. I've tried to find the easiest way to interact with Graph API and that's what I'm going to describe, along with the problems I've faced throughout the process. I've been using PyDev in Eclipse.

1. Get python-facebook sdk. this seems to be the official one. Install it with this command

sudo pip install git+git://github.com/pythonforfacebook/facebook-sdk.git

2. If you have opened up eclipse before this process, you are going to get an import error. PyDev doesn't refresh the PYTHONPATH automatically. You have to got to

File > Properties > Edit the launch configuration > Refresh tab > Check refresh all resources

3. Now, you would need an access token to interact with the Graph API. Usually this is done with a browser requesting with app id and secret token and then facebook returns an access token. That's a bit tricky, I was looking for an easier method. However, for testing purpose, facebook also gives access tokens directly.
Go to this link : https://developers.facebook.com/tools/explorer and click on generate access token. It shall ask for the permissions you want to have in your application and then generate an access token for you. However, It's temporary and shouldn't be hard coded in your application forever.

4. Paste the following code in your editor. Among tons of others, I've found this one little and easy.

import facebook
token = 'your secret token'
graph = facebook.GraphAPI(token)
profile = graph.get_object("me")
friends = graph.get_connections("me", "friends")
friend_list = [friend['name'] for friend in friends['data']]
print friend_list

And see the output in console, a nice list of your friends. Remember two points:

1. If it fails to find 'facebook' module, you might need to restart eclipse.
2. If it fails to find Graph API, re-install facebook module. You can remove it by issuing

   pip freeze

   to get the actual python-facebook module name and then

   pip uninstall that-module

    

Oh My Zsh! for Linux Users

If you use linux and do not use zsh then you are missing something you don't want to miss. Actually I'm grateful to my senior for introducing it to me.

Oh My Zshell is a project for extending and beautifying zshell. Follow that link and don't ever look back to any other shells. :)

For installation and usage:

http://railscasts.com/episodes/308-oh-my-zsh
http://www.stevendobbelaere.be/installing-and-configuring-the-oh-my-zsh-shell/

Screenshot of my current shell, using guake and oh-my-zshell :

Restore missing "Create New Document" and extend it's functionality in ubuntu 13.04 Nautilus context menu

Hi,

One of the reported bugs of 13.04 is that, the creating blank document menu item is missing in nautilus context menu. It's restoration instruction was found in this link .

You just have to create an empty document inside your Templates folder in home directory and the missing menu item would be restored.

Now you can leverage the templates folder's functionality by adding different types of file schemas. For example you can create files with .txt, .html, .sh, .sql extensions and add appropriate template codes inside. For example, I have these files in this folder with those extensions and with the following contents:

untitled.txt

//nothing

untitled.html

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
    "http://www.w3.org/TR/html4/loose.dtd">
<html lang="en">
<head>
    <meta http-equiv="content-type" content="text/html; charset=utf-8">

    <title>Title Goes Here</title>
</head>
<body>
<p>This is my web page</p>
</body>
</html>

bash.sh

#!/bin/bash
//nothing

obsql.sql

CREATE TABLE prefix_table_name
(
  prefix_table_name_id character varying(32) NOT NULL,
  ad_org_id character varying(32) NOT NULL,
  ad_client_id character varying(32) NOT NULL,
  isactive character(1) NOT NULL DEFAULT 'Y'::bpchar,
  created timestamp without time zone NOT NULL DEFAULT now(),
  createdby character varying(32) NOT NULL,
  updated timestamp without time zone NOT NULL DEFAULT now(),
  updatedby character varying(32) NOT NULL,
  
  
  CONSTRAINT prefix_table_name_key PRIMARY KEY (prefix_table_name_id),
  CONSTRAINT prefix_table_name_ad_client FOREIGN KEY (ad_client_id)
      REFERENCES ad_client (ad_client_id) MATCH SIMPLE
      ON UPDATE NO ACTION ON DELETE NO ACTION,
  CONSTRAINT prefix_table_name_ad_org FOREIGN KEY (ad_org_id)
      REFERENCES ad_org (ad_org_id) MATCH SIMPLE
      ON UPDATE NO ACTION ON DELETE NO ACTION
)

You can add any number of file types as you want.

Regular Commands in Linux

Hi,

Most of the time you may notice that you are having to give some regular commands in terminal when you start working in the morning. For example, in my case-opening up chrome, pgadmin, springsource toolsuite, skype, firefox, guake. Initially I had to click a few shortcuts or enter commands in the terminal. Then I've searched for a quicker and easier way and came up with this:

1. Opened up a text file in gedit/sublime-text.
2. Wrote this line:

#!/bin/bash
guake & /home/anik/ides/springsource/sts-3.1.0.RELEASE/STS & pgadmin3 & firefox & skype

3. Saved it as regularCommands.sh
4. Pasted the file inside /bin folder of the file system.
Then I was able to open up a terminal and just put "regularCommands" without the quotes or even omitting the ".sh" and all those programs were starting up one after another. Saves me time and energy.
Hope that helps. :)