A parser for RIPE Atlas measurement results
A parsing library for RIPE Atlas measurement results
Why this exists
RIPE Atlas generates a lot of data, and the format of that data changes over time. Often you want to do something simple like fetch the median RTT for each measurement result between date X and date Y. Unfortunately, there are dozens of edge cases to account for while parsing the JSON, like the format of errors and firmware upgrades that changed the format entirely.
To make this easier for our users (and for ourselves), we wrote an easy to use parser that’s smart enough to figure out the best course of action for each result, and return to you a useful, native Python object.
How to install
The stable version should always be in PyPi, so you can install it with pip:
$ pip install ripe.atlas.sagan
Better yet, make sure you get ujson and sphinx installed with it:
$ pip install ripe.atlas.sagan[fast,doc]
Some setups (like MacOS) have trouble with building the dependencies required for reading SSL certificates. If you don’t care about SSL stuff and only want to use sagan to say, parse traceroute or DNS results, then you can do the following:
$ SAGAN_WITHOUT_SSL=1 pip install ripe.atlas.sagan
Quickstart: How To Use It
You can parse a result in a few ways. You can just pass the JSON-encoded string:
from ripe.atlas.sagan import PingResult my_result = PingResult("<result string from RIPE Atlas ping measurement>") print(my_result.rtt_median) 123.456 print(my_result.af) 6
You can do the JSON-decoding yourself:
from ripe.atlas.sagan import PingResult my_result = PingResult( json.loads("<result string from RIPE Atlas ping measurement>") ) print(my_result.rtt_median) 123.456 print(my_result.af) 6
You can let the parser guess the right type for you, though this incurs a small performance penalty:
from ripe.atlas.sagan import Result my_result = Result.get("<result string from RIPE Atlas ping measurement>") print(my_result.rtt_median) 123.456 print(my_result.af) 6
What it supports
Essentially, we tried to support everything. If you pass in a DNS result string, the parser will return a DNSResult object, which contains a list of Response’s, each with an abuf property, as well as all of the information in that abuf: header, question, answer, etc.
from ripe.atlas.sagan import DnsResult my_dns_result = DnsResult("<result string from a RIPE Atlas DNS measurement>") my_dns_result.responses.abuf # The entire string my_dns_result.responses.abuf.header.arcount # Decoded from the abuf
We do the same sort of thing for SSL measurements, traceroutes, everything. We try to save you the effort of sorting through whatever is in the result.
Which attributes are supported?
Every result type has its own properties, with a few common between all types.
Specifically, these attributes exist on all *Result objects:
- created An datetime object of the timestamp field
- firmware An integer representing the firmware version
- origin The from attribute in the result
- is_error Set to True if an error was found
Additionally, each of the result types have their own properties, like packet_size, responses, certificates, etc. You can take a look at the classes themselves, or just look at the tests if you’re curious. But to get you started, here are some examples:
# Ping ping_result.packets_sent # Int ping_result.rtt_median # Float, rounded to 3 decimal places ping_result.rtt_average # Float, rounded to 3 decimal places # Traceroute traceroute_result.af # 4 or 6 traceroute_result.total_hops # Int traceroute_result.destination_address # An IP address string # DNS dns_result.responses # A list of Response objects dns_result.responses.response_time # Float, rounded to 3 decimal places dns_result.responses.headers # A list of Header objects dns_result.responses.headers.nscount # The NSCOUNT value for the first header dns_result.responses.questions # A list of Question objects dns_result.responses.questions.type # The TYPE value for the first question dns_result.responses.abuf # The raw, unparsed abuf string # SSL Certificates ssl_result.af # 4 or 6 ssl_result.certificates # A list of Certificate objects ssl_result.certificates.checksum # The checksum for the first certificate # HTTP http_result.af # 4 or 6 http_result.uri # A URL string http_result.responses # A list of Response objects http_result.responses.body_size # The size of the body of the first response # NTP ntp_result.af # 4 or 6 ntp_result.stratum # Statum id ntp_result.version # Version number ntp_result.packets.final_timestamp # A float representing a high-precision NTP timestamp ntp_result.rtt_median # Median value for packets sent & received
What it requires
As you might have guessed, with all of this magic going on under the hood, there are a few dependencies:
There’s a full battery of tests for all measurement types, so if you’ve made changes and would like to submit a pull request, please run them (and update them!) before sending your request:
$ python setup.py test
You can also install tox to test everything in all of the supported Python versions:
$ pip install tox $ tox
Complete documentation can always be found on Read the Docs, and if you’re not online, the project itself contains a docs directory – everything you should need is in there.
Who’s Responsible for This?
Sagan is actively maintained by the RIPE NCC and primarily developed by Daniel Quinn, while the abuf parser is mostly the responsibility of Philip Homburg with an assist from Bert Wijnen and Rene Wilhelm who contributed to the original script. Andreas Stirkos did the bulk of the work on NTP measurements and fixed a few bugs, and big thanks go to Chris Amin, John Bond, and Pier Carlo Chiodi for finding and fixing stuff where they’ve run into problems.
But why “Sagan”? The RIPE Atlas team decided to name all of its modules after explorers, and what better name for a parser than that of the man who spent decades reaching out to the public about the wonders of the cosmos?
- abuf.py: error handling for NS records, extended rcode, cookies and client subnets
- Catch problems parsing SSL certificates
- Add support for non-DNS names in subjectAltName extensions
- Replaced pyOpenSSL with cryptography
- Added parsing of subjectAltName X509 extension
- Added first version of WiFi results
- Added a parse_all_hops kwarg to the Traceroute class to tell Sagan to stop parsing Hops and Packets once we have all of the last hop statistics (default=True)
- Remove dependency on IPy: we were using it for IPv6 canonicalization, but all IPv6 addresses in results should be in canonical form to start with.
- Removed the parse_abuf script because no one was using it and its Python3 support was suspect anyway.
- Handle case where a traceroute result might not have dst_addr field.
- Change condition of traceroute’s last_hop_responded flag.
- Add couple of more traceroute’s properties. is_success and last_hop_errors.
- Add tests to the package itself.
- Fix for Issue #56 a case where the qbuf value wasn’t being properly captured.
- Fixed small bug that didn’t accurately capture the DO property from the qbuf.
- We now ignore so-called “late” packets in traceroute results. This will likely be amended later as future probe firmwares are expected to make better use of this value, but until then, Sagan will treat these packets as invalid.
- Added a type attribute to all Result subclasses
- Added support for a lot of new DNS answer types, including NSEC, PTR, SRV, and more. These answers do not yet have a complete string representation however.
- Changed the name of TracerouteResult.rtt_median to TracerouteResult.last_rtt_median.
- Modified the DnsResult class to allow the “bubbling up” of error statuses.
- We skipped this number for some reason :-/
- Breaking Change: the Authority and Additional classes were removed, replaced with the appropriate answer types. For the most part, this change should be invisible, as the common properties are the same, but if you were testing code against these class types, you should consider this a breaking change.
- Breaking Change: The __str__ format for DNS RrsigAnswer to conform the output of a typical dig binary.
- Added __str__ definitions to DNS answer classes for use with the toolkit.
- In an effort to make Sagan (along with Cousteau and the toolkit) more portable, we dropped the requirement for the arrow package.
- 1.0! w00t!
- Breaking Change: the data property of the TxtAnswer class was changed from a string to a list of strings. This is a correction from our own past deviation from the RFC, so we thought it best to conform as part of the move to 1.0.0
- Fixed a bug where non-ascii characters in DNS TXT answers resulted in an exception.
- Fixed a bug related to non-ascii characters in SSL certificate data.
- Added a wrapper for json loaders to handle differences between ujson and the default json module.
- Minor fix to make all Result objects properly JSON serialisable.
- Added iortiz’s patch for flags and flags and sections properties on DNS Answer objects.
- Changed README.md to README.rst to play nice with pypi.
- Added pierky’s new RRSigAnswer class to the dns parser.
- Fixed a bug in how Sagan deals with inappropriate firmware versions
- Added pierky’s fix to fix AD and CD flags parsing in DNS Header
- Added rtt_min, rtt_max, offset_min, and offset_max to NTPResult
- Support for NTP measurements
- Fixes for how we calculate median values
- Smarter setup.py
- Complete Python3 support!
- Added better Python3 support. Tests all pass now for ping, traceroute, ssl, and http measurements.
- Modified traceroute results to make use of destination_ip_responded and last_hop_responded, deprecating target_responded. See the docs for details.
- Added support for making use of some of the pre-calculated values in DNS measurements so you don’t have to parse the abuf if you don’t need it.
- Fixed a bug in the abuf parser where a variable was being referenced by never defined.
- Cleaned up some of the abuf parser to better conform to pep8.
- Fixed a bug where DNS TXT results with class IN were missing a .data value.
- Fixed a problem in the SSL unit tests where \n was being misinterpreted.
- Made abuf more robust in dealing with truncation.
- Replaced SslResult.get_checksum_chain() with the SslResult.checksum_chain property.
- Added support for catching results with an err property as an actual error.
- Fixed a bug in how the on_error and on_malformation preferences weren’t being passed down into the subcomponents of the results.
- Support for seconds_since_sync across all measurement types
- “Treat a missing Type value in a DNS result as a malformation” (Issue #36)
- Minor bugfixes
- Added a median_rtt value to traceroute Hop objects.
- Smarter and more consistent error handling in traceroute and HTTP results.
- Added an error_message property to all objects that is set to None by default.
- Totally reworked error and malformation handling. We now differentiate between a result (or portion thereof) being malformed (and therefore unparsable) and simply containing an error such as a timeout. Look for an is_error property or an is_malformed property on every object to check for it, or simply pass on_malformation=Result.ACTION_FAIL if you’d prefer things to explode with an exception. See the documentation for more details
- Added lazy-loading features for parsing abuf and qbuf values out of DNS results.
- Removed the deprecated properties from dns.Response. You must now access values like edns0 from dns.Response.abuf.edns0.
- More edge cases have been found and accommodated.
- Fixed the deprecation warnings in DnsResult to point to the right place.
- Better handling of DNSResult errors
- Rearranged the way abufs were handled in the DnsResult class to make way for qbuf values as well. The old method of accessing header, answers, questions, etc is still available via Response, but this will go away when we move to 0.2. Deprecation warnings are in place.
- Smarter code for checking whether the target was reached in TracerouteResults.
- We now handle the destination_option_size and hop_by_hop_option_size values in TracerouteResult.
- Extended support for ICMP header info in traceroute Hop class by introducing a new IcmpHeader class.
- Broader support for SSL checksums. We now make use of md5 and sha1, as well as the original sha256.
Release history Release notifications
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size ripe.atlas.sagan-1.3.0-py2.py3-none-any.whl (36.7 kB)||File type Wheel||Python version py2.py3||Upload date||Hashes View hashes|
|Filename, size ripe.atlas.sagan-1.3.0.tar.gz (107.4 kB)||File type Source||Python version None||Upload date||Hashes View hashes|
Hashes for ripe.atlas.sagan-1.3.0-py2.py3-none-any.whl