They should take single matcher during construction, and their match method should take a generic range, using the new generic matcher support.

Expected usage:

REQUIRE_THAT(get_keys(), AllMatch(KeyPattern({1, 2, 3})));
REQUIRE_THAT(get_numbers(), NoneMatch(IsOdd{}));
REQUIRE_THAT(get_strings(), AnyMatch(Contains("webscale") && !Contains("MongoDB")));

the matching semantics

When running build.ps1 -t Test, the following output describes the failing tests:

1. Failed : NUnit.Framework.Api.FrameworkControllerTests.ExploreTestsAction_FileNotFound_ReturnsNonRunnableSuite
   Expected: String containing "The system cannot find the file specified."
   But was: "Le fichier sp├®cifi├® est introuvable. (Exception from HRESULT: 0x80070002)"
   at NUnit.Framework.Api.Framew

In a pact-node consumer test, I have -

describe('edit a policy', () => {
    const REQUEST_BODY = {
      name: 'account-holder',
      version: '2020-07-10T01:22:09.488Z',
    };
    const RESPONSE_BODY = {
      name: Matchers.like('account-holder'),
      version: Matchers.ISO8601_DATETIME_WITH_MILLIS_FORMAT,
      additionalAttributes: Matchers.like({}),
    };
    beforeEach(

I would like to drop Spring Boot support for 1.5.x. It is EOL https://spring.io/blog/2018/07/30/spring-boot-1-x-eol-aug-1st-2019
I would also add some notes in the docs, that we adhere to the spring boot lifecycle in order to be transparent about dropping support for versions.

As part of the process of making the Pact docs more user friendly, we'd like to start writing feature documentation once, and providing code examples for each of the languages, rather than everyone having to write the same docs over and over again in each language.

The documentation for the consumer version selectors can be our first opportunity to give this approach a try. I have written the beh

Feature request

Is your feature request related to a problem? Please describe

The analysis.py script provides just one confidence interval: 95%. While that is related to the 2-sigma interval of normal distribution, so in theory allows for easy estimation of bigger confidence intervals (like 3, 4, 5-sigma), it assumes a normal distribution, which the differences don't follow, and may be