Parker 0.9.0

A web spider for collecting specific data across a set of configured sites

Parker is a Python-based web spider for collecting specific data across a set of configured sites.

Non-Python requirements:

• Redis - for task queuing and visit tracking

• libxml - for HTML parsing of pages

Installation

Install using pip:

$ pip install parker

Configuration

To configure Parker, you will need to install the configuration files in a suitable location for the user running Parker. To do this, use the parker-config script. For example:

$ parker-config ~/.parker

This will install the configuration in your homedir and will output the related environment variable for you to set in your .bashrc.

Changes

0.9.0

• Completely refactored storage and added the ability to store consumed data to Amazon’s S3 service. This requires your AWS Access Key ID to be set up as an environment variable.

• Added an ‘unclassified’ path prefix for sites configured without a classification.

• Added a ‘crawl_uri_filters’ site configuration list which should contain regular expressions to match any URIs you would like filtered out of the crawl.

• Added a ‘seconds_until_expire’ site configuration value which sets the expiry of the sets stored in Redis. Would recommend you ensure this is set to an interval similar to how long it takes Parker to crawl your site so that when it finishes, the Sets will expire shortly after to save memory.

0.8.0

• Updated key-value functionality to allow a sub-selector to return a list of values.

• Added the ability to add an expiry time in seconds to the RedisSet objects, with a default of 5 days.

• Added an extra set to track URIs that are already on the crawl queue. This should hopefully cut down on duplication but may eat memory if there are multiple possiple URIs for the same page.

0.7.3

• Patch to fix an issue where the consumer was overlooking media URIs that start with / and are therefore relative to the base_uri configuration.

• Added boto to the requirements for future use.

0.7.2

• Patch to fix an issue where the crawler was overlooking URIs that start with / and are therefore relative to the base_uri configuration.

0.7.1

• Patch to fix an issue where, if class is not present in the site config, the path includes “None”.

0.7.0

• Rework the client to allow for improved proxy failover should we need it. Improve testing a little to back this up.

• Add tagging to the configuration. These are simply passed through to the resulting JSON objects output by the model so that you can tag them with whatever you want.

• Add classification to the configuration. Again this is passed through, but is also used in the output file path from the consumer worker.

0.6.0

• Add tracking of visited URIs as well as page hashes to the crawl worker. Use that to reduce the number of URIs added to the crawl queue.

0.5.1

• Fix an issue with the order of key-value reference resolution that prevented the effective use of unique_field if using a field that was a kv_ref.

• Add some Parker specific configuration so we can specify where to download, in case the PROJECT env variable doesn’t exist.

0.5.0

• Update ConsumeModel to post process the data. This enables us to populate specific data from a reference to a key-value field.

• Reorder changes so newest first, and rename to “Changes” in the long description.

0.4.2

• Bug fix to fix RST headers which may be the problem.

• Remove the decode/encode which is not the issue.

0.4.1

• Bug fix to see if RST in ASCII fixes issues on PyPI.

0.4.0

• Added handling for a PARKER_CONFIG environment variable, allowing users to specify where configuration files are loaded from.

• Added the parker-config script to install default configuration files to a passed location. Also prints out an example PARKER_CONFIG environment variable to add to your profile files.

• Updated documentation to use proper reStructuredText files.

• Add a CHANGES file to track updates.