|
|
|
Juriscraper is a scraper library started several years ago that gathers judicial opinions, oral arguments, and PACER data in the American court system. It is currently able to scrape:
- a variety of pages and reports within the PACER system
- opinions from all major appellate Federal courts
- opinions from all state courts of last resort except for Georgia (typically their "Supreme Court")
- oral arguments from all appellate federal courts that offer them
Juriscraper is part of a two-part system. The second part is your code, which calls Juriscraper. Your code is responsible for calling a scraper, downloading and saving its results. A reference implementation of the caller has been developed and is in use at CourtListener.com. The code for that caller can be found here. There is also a basic sample caller included in Juriscraper that can be used for testing or as a starting point when developing your own.
Some of the design goals for this project are:
- extensibility to support video, oral argument audio, etc.
- extensibility to support geographies (US, Cuba, Mexico, California)
- Mime type identification through magic numbers
- Generalized architecture with minimal code repetition
- XPath-based scraping powered by lxml's html parser
- return all meta data available on court websites (caller can pick what it needs)
- no need for a database
- clear log levels (DEBUG, INFO, WARN, CRITICAL)
- friendly as possible to court websites
First step: Install Python 3.8+.x, then:
On Ubuntu/Debian Linux:
sudo apt-get install libxml2-dev libxslt-dev libyaml-dev
On macOS with Homebrew <https://brew.sh>:
brew install libyaml
pip install juriscraper
You can set an environment variable for where you want to stash your logs (this can be skipped, and /var/log/juriscraper/debug.log will be used as the default if it exists on the filesystem):
export JURISCRAPER_LOG=/path/to/your/log.txt
Some websites are too difficult to crawl without some sort of automated WebDriver. For these, Juriscraper either uses a locally-installed copy of geckodriver or can be configured to connect to a remote webdriver. If you prefer the local installation, you can download Selenium FireFox Geckodriver:
# choose OS compatible package from: # https://github.com/mozilla/geckodriver/releases/tag/v0.26.0 # un-tar/zip your download sudo mv geckodriver /usr/local/bin
If you prefer to use a remote webdriver, like Selenium's docker image, you can configure it with the following variables:
WEBDRIVER_CONN: Use this to set the connection string to your remote
webdriver. By default, this is local, meaning it will look for a local
installation of geckodriver. Instead, you can set this to something like,
'http://YOUR_DOCKER_IP:4444/wd/hub', which will switch it to using a remote
driver and connect it to that location.
SELENIUM_VISIBLE: Set this to any value to disable headless mode in your
selenium driver, if it supports it. Otherwise, it defaults to headless.
For example, if you want to watch a headless browser run, you can do so by starting selenium with:
docker run \
-p 4444:4444 \
-p 5900:5900 \
-v /dev/shm:/dev/shm \
selenium/standalone-firefox-debug
That'll launch it on your local machine with two open ports. 4444 is the
default on the image for accessing the webdriver. 5900 can be used to connect
via a VNC viewer, and can be used to watch progress if the SELENIUM_VISIBLE
variable is set.
Once you have selenium running like that, you can do a test like:
WEBDRIVER_CONN='http://localhost:4444/wd/hub' \
SELENIUM_VISIBLE=yes \
python sample_caller.py -c juriscraper.opinions.united_states.state.kan_p
Kansas's precedential scraper uses a webdriver. If you do this and watch selenium, you should see it in action.
For scrapers to be merged:
- Automated testing should pass. The test suite will be run automatically by Github Actions. If changes are being made to the pacer code, the pacer tests must also pass when run. These tests are skipped by default. To run them, set environment variables for PACER_USERNAME and PACER_PASSWORD.
- A *_example* file must be included in the
tests/examplesdirectory (this is needed for the tests to run your code). - Your code should be PEP8 compliant with no major Pylint problems or Intellij inspection issues.
- We use the black code formatter to make sure all our Python code has the same formatting. This is an automated tool that you must run on any code you run before you push it to Github. When you run it, it will reformat your code. We recommend integrating into your editor.
- This project is configured to use git pre-commit hooks managed by the Python program pre-commit ([website](https://pre-commit.com/)). Pre- commit checks let us easily ensure that the code is properly formatted with black before it can even be commited. If you install the dev dependencies in requirements-dev.txt, you should then be able to run $ pre-commit install which will set up a git pre-commit hook for you. This install step is only necessary once in your repository. When using this hook, any code files that do not comply to black will automatically be unstaged and re- formatted. You will see a message to this effect. It is your job to then re-stage and commit the files.
- Beyond what black will do for you by default, if you somehow find a way to do whitespace or other formatting changes, do so in their own commit and ideally in its own PR. When whitespace is combined with other code changes, the PR's become impossible to read and risky to merge. This is a big reason we use black.
- Your code should efficiently parse a page, returning no exceptions or speed warnings during tests on a modern machine.
When you're ready to develop a scraper, get in touch, and we'll find you a scraper that makes sense and that nobody else is working on. We have a wiki list of courts that you can browse yourself. There are templates for new scrapers here (for opinions) and here (for oral arguments).
When you're done with your scraper, fork this repository, push your
changes into your fork, and then send a pull request for your changes.
Be sure to remember to update the __init__.py file as well, since it
contains a list of completed scrapers.
Before we can accept any changes from any contributor, we need a signed and completed Contributor License Agreement. You can find this agreement in the root of the repository. While an annoying bit of paperwork, this license is for your protection as a Contributor as well as the protection of Free Law Project and our users; it does not change your rights to use your own Contributions for any other purpose.
To get set up as a developer of Juriscraper, you'll want to install the code from git. To do that, install the dependencies and geckodriver as described above. Instead of installing Juriscraper via pip, do the following:
git clone https://github.com/freelawproject/juriscraper.git . pip install -r requirements.txt python setup.py test # run tests against multiple python versions via tox tox # run network tests (on demand, not run via default command above) python setup.py testnetwork
You may need to also install Juriscraper locally with:
pip install .
If you've not installed juriscraper, you can run sample_caller.py as:
PYTHONPATH=`pwd` python juriscraper/sample_caller.py
The scrapers are written in Python, and can can scrape a court as follows:
from juriscraper.opinions.united_states.federal_appellate import ca1
# Create a site object
site = ca1.Site()
# Populate it with data, downloading the page if necessary
site.parse()
# Print out the object
print str(site)
# Print it out as JSON
print site.to_json()
# Iterate over the item
for opinion in site:
print opinion
That will print out all the current meta data for a site, including
links to the objects you wish to download (typically opinions or oral
arguments). If you download those opinions, we also recommend running the
_cleanup_content() method against the items that you download (PDFs,
HTML, etc.). See the sample_caller.py for an example and see
_cleanup_content() for an explanation of what it does.
It's also possible to iterate over all courts in a Python package, even if they're not known before starting the scraper. For example:
# Start with an import path. This will do all federal courts.
court_id = 'juriscraper.opinions.united_states.federal'
# Import all the scrapers
scrapers = __import__(
court_id,
globals(),
locals(),
['*']
).__all__
for scraper in scrapers:
mod = __import__(
'%s.%s' % (court_id, scraper),
globals(),
locals(),
[scraper]
)
# Create a Site instance, then get the contents
site = mod.Site()
site.parse()
print str(site)
This can be useful if you wish to create a command line scraper that
iterates over all courts of a certain jurisdiction that is provided by a
script. See lib/importer.py for an example that's used in
the sample caller.
A sample driver to run the PACER District Court parser on an html file is included. It takes HTML file(s) as arguments and outputs JSON to stdout.
Example usage:
PYTHONPATH=`pwd` juriscraper/pacerdocket.py tests/examples/pacer/dockets/district/nysd.html
We got that! You can (and should) run the tests with
tox. This will run python setup.py test for all supported Python runtimes,
iterating over all of the *_example* files and run the scrapers against them.
Each scraper has one or more *_example* files. When creating a new scraper,
or covering a new use case for an existing scraper, you will have to create an
example file yourself. Please see the files under tests/examples/ to see
for yourself how the naming structure works. What you want to put in your new
example file is the HTML/json/xml that the scraper in question needs to test
parsing. Sometimes creating these files can be tricky, but more often than not,
it is as simple as getting the data to display in your browser, viewing then copying
the page source, then pasting that text into your new example file.
Each *_example* file has a corresponding *_example*.compare.json file. This
file contains a json data object that represents the data extracted when parsing
the corresponding *_example* file. These are used to ensure that each scraper
parses the exact data we expect from each of its *_example* files. You do not
need to create these *_example*.compare.json files yourself. Simply create
your *_example* file, then run the test suite. It will fail the first time,
indicating that a new *_example*.compare.json file was generated. You should
review that file, make sure the data is correct, then re-run the test suite. This
time, the tests should pass (or at least they shouldn't fail because of the newly
generated *_example*.compare.json file). Once the tests are passing,
feel free to commit, but please remember to include the new *_example*
and *_example*.compare.json files in your commit.
Individual tests can be run with:
python -m unittest -v tests.local.test_DateTest.DateTest.test_various_date_extractions
Or, to run and drop to the Python debugger if it fails, but you must install nost to have nosetests:
nosetests -v --pdb tests/local/test_DateTest.py:DateTest.test_various_date_extractions
- Support for additional PACER pages and utilities
- Support opinions from for all intermediate appellate state courts
- Support opinions from for all courts of U.S. territories (Guam, American Samoa, etc.)
- Support opinions from for all federal district courts with non-PACER opinion listings
- For every court above where a backscraper is possible, it is implemented.
- Support video, additional oral argument audio, and transcripts everywhere available
Deployment to PyPi should happen automatically when a tagged version is pushed to master in the format v*.*.*. If you do not have push permission on master, this will also work for merged, tagged pull requests. Simply update setup.py, tag your commit with the correct tag (v.*.*.*), and do a PR with that.
If you wish to create a new version manually, the process is:
- Update CHANGES.md
- Update version info in
setup.py
- Install the requirements in requirements_dev.txt
- Set up a config file at ~/.pypirc
Generate a distribution
python setup.py bdist_wheel
Upload the distribution
twine upload dist/* -r pypi (or pypitest)
Juriscraper is licensed under the permissive BSD license.
