pythonwhois-alt

Module for retrieving and parsing the WHOIS data for a domain. Supports most domains. No dependencies.


Keywords
whois, nic, domain, domains, python, python3, python37, pythonwhois
License
WTFPL
Install
pip install pythonwhois-alt==2.4.5

Documentation

pythonwhois-alt

This is a fork of powerful Sven Slootweg's pythonwhois library which unfortunately seems abandoned. My changelog is here.

Install it with:

pip install pythonwhois-alt

Original description goes under.


pythonwhois

A WHOIS retrieval and parsing library for Python.

Dependencies

None! All you need is the Python standard library.

Instructions

The manual (including install instructions) can be found in the doc/ directory. A HTML version is also viewable here.

Goals

  • 100% coverage of WHOIS formats.
  • Accurate and complete data.
  • Consistently functional parsing; constant tests to ensure the parser isn't accidentally broken.

Features

  • WHOIS data retrieval
    • Able to follow WHOIS server redirects
    • Won't get stuck on multiple-result responses from verisign-grs
  • WHOIS data parsing
    • Base information (registrar, etc.)
    • Dates/times (registration, expiry, ...)
    • Full registrant information (!)
    • Nameservers
  • Optional WHOIS data normalization
    • Attempts to intelligently reformat WHOIS data for better (human) readability
    • Converts various abbreviation types to full locality names
      • Airport codes
      • Country names (2- and 3-letter ISO codes)
      • US states and territories
      • Canadian states and territories
      • Australian states
  • pwhois, a simple WHOIS tool using pythonwhois
    • Easily readable output format
    • Can also output raw WHOIS data
    • ... and JSON.
  • Automated testing suite
    • Will detect and warn about any changes in parsed data compared to previous runs
    • Guarantees that previously working WHOIS parsing doesn't unintentionally break when changing code

IP range WHOIS

pythonwhois does not yet support WHOIS lookups on IP ranges (including single IPs), although this will be added at some point in the future. In the meantime, consider using ipwhois - it offers functionality and an API similar to pythonwhois, but for IPs. It also supports delegated RWhois.

Do note that ipwhois does not offer a normalization feature, and does not (yet) come with a command-line tool. Additionally, ipwhois is maintained by Philip Hane and not by me; please make sure to file bugs relating to it in the ipwhois repository, not in that of pythonwhois.

Important update notes

2.4.0 and up: A lot of changes were made to the normalization, and the performance under Python 2.x was significantly improved. The average parsing time under Python 2.7 has dropped by 94% (!), and on my system averages out at 18ms. Performance under Python 3.x is unchanged. pythonwhois will now expand a lot of abbreviations in normalized mode, such as airport codes, ISO country codes, and US/CA/AU state abbreviations. The consequence of this is that the library is now bigger (as it ships a list of these abbreviations). Also note that there may be licensing consequences, in particular regarding the airport code database. More information about that can be found below.

2.3.0 and up: Python 3 support was fixed. Creation date parsing for contacts was fixed; correct timestamps will now be returned, rather than unformatted ones - if your application relies on the broken variant, you'll need to change your code. Some additional parameters were added to the net and parse methods to facilitate NIC handle lookups; the defaults are backwards-compatible, and these changes should not have any consequences for your code. Thai WHOIS parsing was implemented, but is a little spotty - data may occasionally be incorrectly split up. Please submit a bug report if you run across any issues.

2.2.0 and up: The internal workings of get_whois_raw have been changed, to better facilitate parsing of WHOIS data from registries that may return multiple partial matches for a query, such as whois.verisign-grs.com. This change means that, by default, get_whois_raw will now strip out the part of such a response that does not pertain directly to the requested domain. If your application requires an unmodified raw WHOIS response and is calling get_whois_raw directly, you should use the new never_cut parameter to keep pythonwhois from doing this post-processing. As this is a potentially breaking behaviour change, the minor version has been bumped.

It doesn't work!

  • It doesn't work at all?
  • It doesn't parse the data for a particular domain?
  • There's an inaccuracy in parsing the data for a domain, even just a small one?

If any of those apply, don't hesitate to file an issue! The goal is 100% coverage, and we need your feedback to reach that goal.

License

This library may be used under the WTFPL - or, if you take issue with that, consider it to be under the CC0.

Data sources

This library uses a number of third-party datasets for normalization:

Be aware that the OpenFlights database in particular has potential licensing consequences; if you do not wish to be bound by these potential consequences, you may simply delete the airports.dat file from your distribution. pythonwhois will assume there is no database available, and will not perform airport code conversion (but still function correctly otherwise). This also applies to other included datasets.

Contributing

Feel free to fork and submit pull requests (to the develop branch)! If you change any parsing or normalization logic, ensure to run the full test suite before opening a pull request. Instructions for that are below.

Please note that this project uses tabs for indentation.

All commands are relative to the root directory of the repository.

Pull requests that do not include output from test.py will be rejected!

Adding new WHOIS data to the testing set

pwhois --raw thedomain.com > test/data/thedomain.com

Checking the currently parsed data (while editing the parser)

./pwhois -f test/data/thedomain.com .

(don't forget the dot at the end!)

Marking the current parsed data as correct for a domain

Make sure to verify (using pwhois or otherwise) that the WHOIS data for the domain is being parsed correctly, before marking it as correct!

./test.py update thedomain.com

Running all tests

./test.py run all

Testing a specific domain

./test.py run thedomain.com

Running the full test suite including support for multiple python versions

tox

Generating documentation

You need ZippyDoc (which can be installed through pip install zippydoc).

zpy2html doc/*.zpy