Local Government Scraper Framework

This is an experimental project that should be considered alpha and unstable.

The Local Government Scraper Framework is a set of tools for scraping UK local government websites. It was made for fun to see what a toolkit for people who wanted to turn government websites in to open data would look like.

Ideas:

1. Each council has a folder/package that contains contains scrapers.

2. There are scraper classes for types of thing that might want scraping, like councillors

3. There are classes for the types of thing. For example, there is a Councillor class. A CouncillorScraper is expected to produce a set of Councillor objects. The Councillor objects know how to be saved, cleaned, etc.

4. Raw data is scraped and normalised in to a simple structure with little processing of the values. The data is processes later, for example to match party names to identifiers.

5. Scrapers for common CMSs exist, making sub-classing on of them easy. All that should be needed is the base URL, if a CMS is known and a scraper class exists for it.

6. It’s possible to use a Django like command line interface for running scrapers. Scrapers can be run by tags, council identifiers, failing, etc

Using

This is very new and likely to change a lot. If you want to actually use this project, it’s recommended you also join the Democracy Club Slack to talk about it with us. Everything is likely to change, and this code is in no way supported.

Requirements

• Python 3.10
• pipenv

Installation

Check out the code and run pipenv install in the directory

Running

At the moment, the only type of scraper supported is councillors.

To scrape all councillors, run:

python manage.py councillors --all-councils

This will take some time. Add -v for verbose output.

To run a single council run e.g:

python manage.py councillors --council KIR

Where KIR is the council ID from the MHCLG register.

Running the scrapers will create a data directory with raw and JSON folders and a file in each per councillor.

Contributing

Install pre-commit hooks: $ pre-commit install

If you want to add a scraper for a new council:

1. Find the register code for it using the link above

2. Create a python package in scrapers

3. Create a file called councillors.py (more types of scraper are planned, but not supported yet).

4. Create a class called Scraper and sub-class either the base scraper or a CMS specific scraper (see below)

5. Test the scraper with python manage.py councillors --council [package name] -v

Councillor Scraper classes

There are 4 types of councillor scraper class supported.

All scrapers require a base_url property on the class. You can optionally set a list of tags and a disabled flag on the class.

BaseCouncillorScraper

This is the most basic scraper class. It requires two methods: get_councillors and get_single_councillor.

get_councillors must return a iterator that contains the raw content representing a councillor.

Each item returned will be passed to get_single_councillor.

get_single_councillor must return a Councillor object.

If this pattern doesn’t work for a council, then the required run method can be overridden. Run needs to call self.save_councillor with the raw scraped data for each councillor, and a councillor object. It then needs to call self.report().

HTMLCouncillorScraper

Expects a dict containing CSS selectors for example:

list_page = {
	"container_css_selector": ".container .col-md-8",
	"councillor_css_selector": ".col-sm-3",
}

Where container_css_selector is the CSS selector for the elements that contains a list of councillors, and councillor_css_selector is the selector for each element that contains a single councillor.

The get_single_councillor method is required, and needs to return Councillor object.

PagedHTMLCouncillorScraper

A subclass of HTMLCouncillorScraper that supports pagination in on the container page (the list of coucillors is split over different pages).

Looks for a new key in self.list_page called "next_page_css_selector" and uses that to iterate over the pages calling get_single_councillor for each as it goes.

CMISCouncillorScraper

This is a scraper sub-class for councils using CMIS. You can tell these because they normally have CMIS in the URL somewhere.

All that should be required is a base_url. This normally ends in Councillors.aspx or Members.aspx and is the list of councillors per ward.

ModGovCouncillorScraper

Similar to the CMIS class, this scrapes ModernGov URLs. ModernGov sites have URLs that contain something like mgMemberIndex.aspx . If mg is in the URL, it’s likely it’s a ModGov site. You can test this by looking for the WDSL page.

If the URL with councillors on is http://democracy.kirklees.gov.uk/mgMemberIndex.aspx then try requesting http://democracy.kirklees.gov.uk/ mgWebService.asmx?WSDL. If you see XML, then it's using ModGov with the API turned on.

The base_url should be the URL above the mgWebService.asmx script, e.g. http://democracy.kirklees.gov.uk/ or http://democracy.kirklees.gov.uk/councillors/ if it’s installed at a sub-path.

Skipping councillors

The contract of get_single_councillor is that it must return a Councillor object.

However in some cases the source requires that this can't happen. Two examples:

1. With HTMLCouncillorScraper when iteration over all rows in a table, we sometimes see inline header rows. They can get passed to get_single_councillor, but don't contain a councillor
2. Adur and Worthing have one page for all their councillors and we need to remove Adur councillors from Worthing's scraper and vise versa.

To deal with this we have SkipCouncillorException. If get_single_councillor raises this exception then the loop continues on to the next councillor.

Councillor objects

All scrapers in some way need to make a set councillor objects. CMISCouncillorScraper and ModGovCouncillorScraper handle this automatically, but the HTMLCouncillorScraper and BaseCouncillorScraper don’t.

Councillor objects require a url, identifier, name, party and division:

from lgsf.councillors import CouncillorBase

councillor = CouncillorBase(
    url,
    identifier=identifier,
    name=name,
    party=party,
    division=division,
)

All councillor scrapers need to have self.councillors = set(), and each scraper needs to populate this with Councillor objects.