Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: How to create a Web crawler?

news.ycombinator.com

Post reply on HN
Print view View original

1–10 of 10 posts

Ask HN: How to create a Web crawler?

#1

Post by gurpreet42 »

I am creating a website for price comparison. I want to fetch the price of a single product from multiple shopping carts. I have some questions regarding this. 1) What is the best way to fetch price form different websites? 2) Sometime a single product have different name on different shopping carts. How to handle this problem? 3) When I will send multiple requests on a website (shopping cart) using my created web crawler (or web spider) will they block me or can they take a legal action against me? 4) How can I make the process automated to omit manual error.

I will start with 3 shopping carts initially. Most of the shopping carts are not providing any type of API or other type of access to their products. Currently my approach is to tear down the HTML and find required information from HTML.

I want to go with C# & .Net technology because I am good in it (this is what I thought).

Please suggest what is the best way of doing this.

Re: Ask HN: How to create a Web crawler?

#2

Post by Piskvorrr »

Okay, step #1: go search the web for "web spidering" and "web crawling". Go read the relevant articles (and Wikipedia's, too; a good starting point IMHO). Step #2: update this question when you have a question that's answerable. See, entire books can be (and have been) written on the subjectS that you ask about (and that's even before getting started with the legal aspects).

As for question number 4: Impossible unless you have strong AI (see also: How Apple automatically collated their map data from three different sources and what they got as a result). You see, fully automated data collection from multiple source will introduce more errors than manual methods (which are slow, OTOH); automated collection and manual verification is necessary.

Re: Ask HN: How to create a Web crawler?

#3

Post by gurpreet42 »

Piskvorrr wrote: post #2

Okay, step #1: go search the web for "web spidering" and "web crawling". Go read the relevant articles (and Wikipedia's, too; a good starting point IMHO). Step #2: update this question when you have a question that's answerable. See, entire books can be (and have been) written on the subjectS that you ask about (and that's even before getting started with the legal aspects). As for question number 4: Impossible unles…

I have searched on web and have created one solution, this is semi automated means I have to put some URLs and some parameters to it to get desired result.

Its worked for me up to some extent.

Re: Ask HN: How to create a Web crawler?

#5

Post by gurpreet42 »

mmariani wrote: post #4

If you're willing to veer off from C# there's a great framework in Python called Scrapy [0], it's fast and easy to pick up. You could crawl and save the results in your database with Scrapy, and get them back with your C# backend. [0] http://scrapy.org/

I can switch to any platform, provided that would be best one. I will check this framework and try to create a demo application so that i can get better idea. There are many companies that use web crawling. Are they using Pyton or something other ?

Re: Ask HN: How to create a Web crawler?

#6

Post by dgunn »

Some sites will say you can't scrape. They'll say so in their TOS if that's the case. If you absolutely need that info, there are ways to not look like a spider. One way is to use a browser driver (pretty heavy option so I would just not use those sites for now). I can't remember the name of the browser drivers available, but that's one possibility. I doubt they'll have libraries for your languages though.

I would definitely recommend python for web scraping. There are a ton of tutorials out there. I believe there are even a few from Google which are pretty good.

Re: Ask HN: How to create a Web crawler?

#7

Post by gurpreet42 »

dgunn wrote: post #6

Some sites will say you can't scrape. They'll say so in their TOS if that's the case. If you absolutely need that info, there are ways to not look like a spider. One way is to use a browser driver (pretty heavy option so I would just not use those sites for now). I can't remember the name of the browser drivers available, but that's one possibility. I doubt they'll have libraries for your languages though. I would de…

Thanks for your suggestions. I will try python for my further efforts.

Re: Ask HN: How to create a Web crawler?

#9

Post by philipDS »

An alternative to Python or C# for web crawling could be Node.js. It's pretty good at it and you have a few libraries that can help you: * Node.io (https://github.com/chriso/node.io) * Phantomjs-node for dynamic content (https://github.com/sgentle/phantomjs-node) * Cheerio for a jQuery server-side implementation (https://github.com/MatthewMueller/cheerio) * Node-jquery as an alternative to Cheerio (https://github.com/coolaj86/node-jquery)

A single product might have a different name, but you might try to scrape the product ID's if they exist. Product IDs should be unique. If both websites provide those ID's you could compare those in your database. If that's not possible: as a small hack you could also use Amazon Mechanical Turk to issue manual tasks to compare product names. This way, real people will check if two products are the same in case there is doubt. This will cost you a little, but you could give those people 5 cents per product comparison or something like that.

For question 3, some websites don't allow you to crawl their content. Read their ToS :-)

For the rest I agree with Piskvorrr, you could do some trial and error and learn on the fly or read some books (and still do trial and error and learn on the fly ;-)). Good luck!

Re: Ask HN: How to create a Web crawler?

#10

Post by gurpreet42 »

philipDS wrote: post #9

An alternative to Python or C# for web crawling could be Node.js. It's pretty good at it and you have a few libraries that can help you: * Node.io ( https://github.com/chriso/node.io ) * Phantomjs-node for dynamic content ( https://github.com/sgentle/phantomjs-node ) * Cheerio for a jQuery server-side implementation ( https://github.com/MatthewMueller/cheerio ) * Node-jquery as an alternative to Cheerio ( https://git…

thanks ..
Post reply on HN