Add more data types in the tests

[vincentparrett]

The current speed tests really only test quoted strings, it should test all data types.

Also some validity testing of numbers (ie does it treat 1.2.3 as a number or a string?),

hex numbers

dates (ISO 8601)

booleans

Test handling of escape characters, unicode escapes

Use specific test files that are never written to so that each library is parsing the same thing, rather than a file it generated itself.

Add more methods to ILib to enable generating more complex documents with differnent data structures and data types. That might be more difficult to achieve due to the differences in the libs. Perhaps rather than calling Lib.Add it should just be Lib.BuildDocA Lib.BuildDocB and have each lib build the same document - although how you would validate that I don't know.

Handle errors in the tests - if a library raises and exception then the report should not report timing and just say ERROR - it took me a while to realise my parser was failing the load test because I didn't read the earlier output.

The timing isn't very accurate, but these tests do give a reasonable indication.

There is probably more I could think of but it is all a lot of work so I won't be upset if you ignore my suggestions ;)

[hydrobyte]

The Seed test was intended as an introductory test only. If you combine these results with the recently added Open File test, you'll see that the differences remain the same.

The Validation test isn't a JSON standard compliance test. I simply found a test suite from another project and applied it. What's noteworthy is that a subset of JSON standard validation was sufficient to fail all tested solutions. The McJSON is an exception because I included them in the unit tests.

That might be more difficult to achieve due to the differences in the libs.

Yes, indeed. Lets keep things simple for a while.

There is probably more I could think of but it is all a lot of work so I won't be upset if you ignore my suggestions ;)

Absolutely not. I'm just trying not to escalate the complexity for now, because in my point of view, simple tests were enough to demonstrate significant differences in solutions that were previously considered completely solid. But remember, my goal has always been to put all the solutions in one place so that others can have some kind of quick comparison.