• Install the project dependencies with yarn
• Generate the binary for your platform with yarn build.binary. This will also compile the project from TS -> JS

Run yarn test.harness from your terminal

You can run one or selected tests using TESTS env variable. If you want multiple test files to be run separate them with commas. Use paths relative to the ./spec directory.

E.g. run TESTS=parameters-ac1.oas2.txt,validate-body-params/form-byte-format-fail.oas2.txt yarn test.harness

All test files are matched using a glob **/*.txt. This means that you can:

• nest test files in subdirectories
• skip files by suffixing name with .skip or some other suffix.

• Create a new file in the ./spec directory. It can have any name and any extension, it does not really matter.
• Use the following template and put your stuff:

====test====
Test text, can be multi line.
====spec====
openapi 2/3 document
====server====
command line arguments to run Prism with. You have to use the ${document} template variable.
====command====
curl command to get the response
====expect====
expected output from curl

====test====
Hello
====spec====
openapi: 3.0.2
paths:
  /todos:
    get:
      responses:
        200:
          description: Get Todo Items
          content:
            'application/json':
              example: hello
====server====
mock -p 4010 ${document}
====command====
curl -i -X GET http://localhost:4010/todos -H "accept: application/json"
====expect====
HTTP/1.1 200 OK
content-type: application/json; charset=utf-8
content-length: 7
Date: Fri, 21 Jun 2019 09:25:34 GMT
Connection: keep-alive

"hello"

• 1 scenario per file.
• Be precise with the separators. They should be 4 before AND after the word. ====
• The 6 keywords are test,spec,server,command,expect,expect-loose, nothing else at the moment
• You can run all the tests on the same port 4010, but you can also choose another one
• The curl command does not support piping stuff into other tools; so if you're trying to be cool and do curl | grep, well maybe next time.
• All the curl commands must have the -i flag, otherwise the trace parser won't understand the output
• Some of the harness test are relying on an instance of httpbin working locally on your computer and responding on the http://httpbin address. You can easily do that using the kennethreitz/httpbin docker image

• A RegExp is used to split the content
• A temporary file with the specification file is stored on your disk
• Prism gets spawn with the specified arguments and waited to be running
• The curl command is performed
• The outputs get converted using http-string-parser, a veeery old package transforming curl output to a consumable format
• Gavel is used to validate the request — it will automagically ignore headers that can change and consider only the "fundamental" one such as content negotiation ones and stuff around.

There are 3 different keywords you can use for the "expect" portion of the test: expect, expect-loose, expect-keysOnly. With each of these, gavel will be used to validate the actual output with the expected output. expect-loose will not add any additional checks on top of the gavel validation. expect and expect-keysOnly will add additional checks on the output as described below:

• expect: additionally uses jest's .toStrictEqual() which ensures the output matches exactly with the expected body supplied. This includes all keys and values match and are in the exact order.
• expect-keysOnly: additionally validates that only the keys are exactly the same and in the exact order. The values of the output are not validated. This is useful when you want to test dynamic mode, where the values will be different every time, but also want to validate that the order of the keys of the response are correct. This is meant to be used only for tests that return a a JSON object. This is not meant to work with any other response body/type.

• Async callback was not invoked within the 5000ms timeout specified by jest.setTimeout.Timeout --> most likely the binary is crashing, failing to start Prism. The best way to troubleshoot is to copy the openapi content from your harness test into a separate OpenAPI file, and try mocking that file with the CLI directly. That will give you more information as to why Prism failed to start.