Skip to content

Latest commit

 

History

History
220 lines (172 loc) · 8.31 KB

README.md

File metadata and controls

220 lines (172 loc) · 8.31 KB

ABOUT

This is the base implementation of a full crawler that uses a spacetime cache server to receive requests.

CONFIGURATION

Step 1: Install dependencies

If you do not have Python 3.6+:

Windows: https://www.python.org/downloads/windows/

Linux: https://docs.python-guide.org/starting/install3/linux/

MAC: https://docs.python-guide.org/starting/install3/osx/

Check if pip is installed by opening up a terminal/command prompt and typing the commands python3 -m pip. This should show the help menu for all the commands possible with pip. If it does not, then get pip by following the instructions at https://pip.pypa.io/en/stable/installing/

To install the dependencies for this project run the following two commands after ensuring pip is installed for the version of python you are using. Admin privileges might be required to execute the commands. Also make sure that the terminal is at the root folder of this project.

python -m pip install packages/spacetime-2.1.1-py3-none-any.whl
python -m pip install -r packages/requirements.txt

Step 2: Configuring config.ini

Set the options in the config.ini file. The following configurations exist.

USERAGENT: Set the useragent to IR F19 uci-id1,uci-id2,uci-id3. It is important to set the useragent appropriately to get the credit for hitting our cache.

HOST: This is the host name of our caching server. Please set it as per spec.

PORT: This is the port number of our caching server. Please set it as per spec.

SEEDURL: The starting url that a crawler first starts downloading.

POLITENESS: The time delay each thread has to wait for after each download.

SAVE: The file that is used to save crawler progress. If you want to restart the crawler from the seed url, you can simply delete this file.

THREADCOUNT: This can be a configuration used to increase the number of concurrent threads used. Do not change it if you have not implemented multi threading in the crawler. The crawler, as it is, is deliberately not thread safe.

Step 3: Define your scraper rules.

Develop the definition of the function scraper in scraper.py

def scraper (url: str, resp: utils.response.Response): -> list
    pass

The scraper takes in two parameters:

ARGS

url:

The URL that was added to the frontier, and downloaded from the cache. It is of type str and was an url that was previously added to the frontier.

resp:

This is the response given by the caching server for the requested URL. The response is an object of type Response (see utils/response.py)

class Response:
    Attributes:
        url:
            The URL identifying the response.
        status:
            An integer that identifies the status of the response. This
            follows the same status codes of http.
            (REF: https://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html)
            In addition there are status codes provided by the caching
            server (600-606) that define caching specific errors.
        error:
            If the status codes are between 600 and 606, the reason for
            the error is provided in this attribute. Note that for status codes
            (400-599), the error message is not put in this error attribute; instead it
            must picked up from the raw_response (if any, and if useful).
        raw_response:
            If the status is between 200-599 (standard http), the raw
            response object is the one defined by the requests library.
            Useful resources in understanding this raw response object:
                https://realpython.com/python-requests/#the-response
                https://requests.kennethreitz.org/en/master/api/#requests.Response
            HINT: raw_response.content gives you the webpage html content.

Return Value

This function needs to return a list of urls that are scraped from the response. (An empty list for responses that are empty). These urls will be added to the Frontier and retrieved from the cache. These urls have to be filtered so that urls that do not have to be downloaded are not added to the frontier.

The first step of filtering the urls can be by using the is_valid function provided in the same scraper.py file. Additional rules should be added to the is_valid function to filter the urls.

EXECUTION

To execute the crawler run the launch.py command. python3 launch.py

You can restart the crawler from the seed url (all current progress will be deleted) using the command python3 launch.py --restart

You can specify a different config file to use by using the command with the option python3 launch.py --config_file path/to/config

ARCHITECTURE

FLOW

The crawler receives a cache host and port from the spacetime servers and instantiates the config.

It launches a crawler (defined in crawler/__init__.py L5) which creates a Frontier and Worker(s) using the optional parameters frontier_factory, and worker_factory.

When the crawler in started, workers are created that pick up an undownloaded link from the frontier, download it from our cache server, and pass the response to your scraper function. The links that are received by the scraper is added to the list of undownloaded links in the frontier and the url that was downloaded is marked as complete. The cycle continues until there are no more urls to be downloaded in the frontier.

REDEFINING THE FRONTIER:

You can make your own frontier to use with the crawler if they meet this interface definition:

class Frontier:
    def __init__(self, config, restart):
        #Initializer.
        # config -> Config object (defined in utils/config.py L1)
        #           Note that the cache server is already defined at this
        #           point.
        # restart -> A bool that is True if the crawler has to restart
        #           from the seed url and delete any current progress.

    def get_tbd_url(self):
        # Get one url that has to be downloaded.
        # Can return None to signify the end of crawling.

    def add_url(self, url):
        # Adds one url to the frontier to be downloaded later.
        # Checks can be made to prevent downloading duplicates.
    
    def mark_url_complete(self, url):
        # mark a url as completed so that on restart, this url is not
        # downloaded again.

A sample reference is given in utils/frontier.py L10. Note that this reference is not thread safe.

REDEFINING THE WORKER

You can make your own worker to use with the crawler if they meet this interface definition:

from scraper import scraper
from utils.download import download
class Worker(Thread): # Worker must inherit from Thread or Process.
    def __init__(self, worker_id, config, frontier):
        # worker_id -> a unique id for the worker to self identify.
        # config -> Config object (defined in utils/config.py L1)
        #           Note that the cache server is already defined at this
        #           point.
        # frontier -> Frontier object created by the Crawler. Base reference
        #           is shown in utils/frontier.py L10 but can be overloaded
        #           as detailed above.
        self.config = config
        super().__init__(daemon=True)

    def run(self):
        In loop:
            > url = get one undownloaded link from frontier.
            > resp = download(url, self.config)
            > next_links = scraper(url, resp)
            > add next_links to frontier
            > sleep for self.config.time_delay

A sample reference is given in utils/worker.py L9.

THINGS TO KEEP IN MIND

  1. It is important to filter out urls that do not point to a webpage. For example, PDFs, PPTs, css, js, etc. The is_valid filters a large number of such extensions, but there may be more.
  2. It is important to filter out urls that are not with ics.uci.edu domain.
  3. It is important to maintain the politeness to the cache server (on a per domain basis).
  4. It is important to set the user agent in the config.ini correctly to get credit for hitting the cache servers.
  5. Launching multiple instances of the crawler will download the same urls in both. Mechanisms can be used to avoid that, however the politeness limits still apply and will be checked.
  6. Do not attempt to download the links directly from ics servers.