Click here to Skip to main content
15,860,859 members
Articles / Desktop Programming / Win32

Making a Search Engine

Rate me:
Please Sign up or sign in to vote.
4.94/5 (51 votes)
3 May 2013CPOL6 min read 238K   27.6K   124   59
This article discusses the making of a search engine.

Search Suggest

search suggest

search page for crawler

Crawler status

Introduction

This project is still not complete. You must use it to crawl thousands of URLs because you may find that it crawls the same URL for the last 100 times. (This is because of some unidentified problem in conversion of relative to absolute URL.)

 

Like most search engines, this one also has a crawler whose basic aim is to retrieve the source code of a given URL and then break the content into words with which we can create an array of tag words which will represent the content of the site. It is not a fool proof method, but can work for sites with lots of words in it, like a blog or article or a discussion forum, etc.

For example:

TAG CLOUD of http://www.google.co.in

Tag cloud of http://www.google.co.in

TAG CLOUD of http://www.wikipedia.org

tag cloud of wikipedia.org

TAG CLOUD of http://www.w3schools.com

tag cloud of w3schools

It is clear how tag cloud can highlight the key words which could describe a given URL.

Now these keywords are stored in a database and then used to find the relevant URL for given keywords.

Background

It all started when my friend showed me his search engine with 4 URLs in an XML file. It seemed like an auto complete rather than a search engine, but later that night it was 2:00 am, and I couldn't sleep at all because of that auto complete feature with which I was too impressed. I wanted my own... there was a thunder storm of ideas in my mind. After 3 sleepless nights, on the 3rd day at 6:00 a.m., I was ready with my search engine working with 100 URLs in database... that was the time when I finally slept comfortably and full of satisfaction. It took me 3 days because every day I started from the beginning because I was not satisfied with the performance of the crawler or there was some problem.

Using the Code

I have basically divided every task into small parts so that the work could become easy. So you would find lots of classes in the project.

Some important classes are given below:

  • Panda-> It retrieves the source code of a given URL using a get_sourcecode(url) from Module 'func' and then passes it to juicer(class) for extracting the useful information and after the work is completed, reports to the panda_manger (boss of all the pandas) and again assigns a new job to panda if any.
  • panda_manger-> It manages all the pandas and from this class we assign any web URL for crawling and it will automatically assign this work to any free panda and if no panda is free, then it creates a new panda and assigns the work to it. When a panda finishes, it works and reports back to panda_manger then panda_manger checks whether there is any URL left to be crawled, if any then it gives the command to crawl that URL to the panda.
  • juicer->This could be said to be the main class for the whole crawler which extracts keywords from the source code of any website and then saves it into the database.
  • juicer->extract_juice-> This class gets the source code and then converts it into HtmlAgilityPack.HtmlDocument using the method LoadHtml(source). It is used because we need an HTML parser to scrap the text out of HTML (building a custom one will require a lot of time) and since it supports xpath for retrieving any element(s) from source, it simplifies our work a lot.
    Now we pass the HtmlDocument.DocumentNode(.SelectNodes) to the various methods of this class for extraction of some type of information.
    VB.NET
    Dim doc As New HtmlAgilityPack.HtmlDocument()
    doc.LoadHtml(source)
    process_texttag(doc.DocumentNode.SelectNodes("//meta"))
    prcess_anchor(doc.DocumentNode.SelectNodes("//a"))
    process_image(doc.DocumentNode.SelectNodes("//img"))
  • juicer->process_metatag-> Currently, it just processes the meta tag containing keywords and then assigns every word a 45% to total word count in the document and 30% of total word count to the words which are mentioned in the keywords but can't be found in the document.
  • juicer->process_anchor->This is the most important part if you want the search engine to automatically move to the next link without manually entering every link to the crawler and the hardest part was to convert the relative URL to absolute URL. But Microsoft saved me with the URI class which can easily be used to convert the URI.

Database Structure

At the starting point, we have a 4 tables stored in a database named as "Crawler" by default.

Table Name : "Keyword_index"

This table is used to provide the suggested result in the search box. It contains all the words which the crawler has encountered till now and no word is repeated.

mysql result for keyword_index

search suggest for a

For multiple keywords, we first break the keyword from " ", then find the urlhash for the word and then find what the other words are that urlhashes contain and show the words. {Not yet implemented because I was getting an unidentified error and if I used SQL join, then it took more than 1 minute to search the database for 1 keyword and time increased exponentially.}

Table Name :"Keyword_list"

This table contains all the words in a tag cloud of the given URL. And to identify this word belongs to which URL, we store the urlhash of that website with the word.

describe keyword_list

Image 11

Table Name :"url_webpage"

This table is used to store all the links which it finds in the pages that it crawled till now. In this table, we also use MD5 hash of URL to refer to that URL instead of the original URL because no one knows how long the URL which we may find can be. So we change it to MD5 because it does not mater how long the URL is, the MD5 will always be 32 chars long.

describe url_webpage

Image 13

Table Name : "url_image"

This table is used to store all the links of the image which are found in the webpage it crawls. It will be used for image search, but I have not yet completed its processing and front end(PHP code).

So I will not discuss it right now. But I would like to tell that it will take a lot of processing power.

Image 14

How the Crawler Works

crawler layout

Points of Interest

It was very annoying to get rid of the relative URL. I tried several ways of resolving it, but every one ends up with some bug. Then I got a magic class, named as URI which solved all my problems, but while writing this article I should crawl codeproject.com and show the result for CodeProject search, but then a problem hit my crawler after identifying that I have deleted my database of URL, otherwise I would have posted a screenshot of that. It was something like /search.aspx (some text)(same text repeated again)(and again, increasing with each crawl). It may be a problem with my code. I will try later to identify this problem and post the solution.

 

<hr /><h3>Currently Crawler is not following robots.txt so be careful while crawling.</h3><h3>It is you responsibility if crawling a site which is not to be crawled.   </h3><p>Sorry, but currently i am working on a new model of crawler. </p>

License

This article, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)


Written By
Student
India India
I just love coding. But due to my studies it became very tough for me to manage both.

Comments and Discussions

 
GeneralMy vote of 5 Pin
phabtar29-Apr-13 19:12
phabtar29-Apr-13 19:12 
GeneralMy vote of 5 Pin
_Vitor Garcia_22-Apr-13 6:28
_Vitor Garcia_22-Apr-13 6:28 
GeneralRe: My vote of 5 Pin
kburman622-Apr-13 6:59
professionalkburman622-Apr-13 6:59 
GeneralMy vote of 5 Pin
Ștefan-Mihai MOGA12-Apr-13 19:04
professionalȘtefan-Mihai MOGA12-Apr-13 19:04 
Questiongood one Pin
aankur818-Apr-13 19:44
professionalaankur818-Apr-13 19:44 
GeneralMy vote of 5 Pin
JMK893-Apr-13 23:48
professionalJMK893-Apr-13 23:48 
GeneralGood Job Pin
Rajneesh Rai3-Apr-13 0:32
Rajneesh Rai3-Apr-13 0:32 
GeneralMy vote of 5 Pin
ring_02-Apr-13 19:02
ring_02-Apr-13 19:02 
Superb!
GeneralMy vote of 5 Pin
Prasad Khandekar2-Apr-13 18:40
professionalPrasad Khandekar2-Apr-13 18:40 
GeneralMy vote of 5 Pin
Oso Oluwafemi Ebenezer29-Mar-13 2:00
Oso Oluwafemi Ebenezer29-Mar-13 2:00 
NewsRe: My vote of 5 Pin
kburman630-Mar-13 1:27
professionalkburman630-Mar-13 1:27 
GeneralRe: My vote of 5 Pin
Oso Oluwafemi Ebenezer30-Mar-13 6:49
Oso Oluwafemi Ebenezer30-Mar-13 6:49 
GeneralMy vote of 5 Pin
Prasad Khandekar20-Mar-13 4:53
professionalPrasad Khandekar20-Mar-13 4:53 
GeneralMy vote of 5 Pin
npdev1319-Mar-13 18:34
npdev1319-Mar-13 18:34 
GeneralMy vote of 5 Pin
rspercy6519-Mar-13 13:29
rspercy6519-Mar-13 13:29 
AnswerRe: My vote of 5 Pin
kburman620-Mar-13 3:14
professionalkburman620-Mar-13 3:14 
GeneralMy vote of 5 Pin
Abinash Bishoyi19-Mar-13 6:35
Abinash Bishoyi19-Mar-13 6:35 
GeneralRe: My vote of 5 Pin
kburman619-Mar-13 7:22
professionalkburman619-Mar-13 7:22 

General General    News News    Suggestion Suggestion    Question Question    Bug Bug    Answer Answer    Joke Joke    Praise Praise    Rant Rant    Admin Admin   

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.