Re: Help in JSON encode/decode project

[Andrew Janke]

On 3/9/20 2:01 AM, Kai Torben Ohlhus wrote:
> On 3/9/20 2:56 PM, Andrew Janke wrote:
>>
>> I took a look in to RapidJSON and agree that it looks like a better
>> library for this than jsoncpp, both for performance, and code quality &
>> documentation.
> 
> 
> Now I am curious how did you get convinced of this change?  Did you run
> some large scale tests yourself?
> 
> Best,
> Kai
> 

No, I don't have time for that. :)

I took a look at this benchmark [1] (which happens to be by the author
of RapidJSON) and thought it was pretty convincing; the code is put
together well, and I see it referenced in a lot of places. It shows a
pretty stark performance difference between jsoncpp and RapidJSON. And
the fact that the library author actually made and ran benchmarks made
me think well of RapidJSON. I saw a couple other benchmarks floating
around in which RapidJSON also beat out jsoncpp.

Then I took a look at the RapidJSON library itself, reading through its
documentation and part of its codebase, and reviewing its GitHub
activity. This struck me as a high quality library. I like its API a bit
better than jsoncpp's, and it's much better documented. The author
clearly also *gets* Unicode encoding and the subtleties of the various
definitions and "standards" for JSON. So I decided I was more
comfortable using RapidJSON than jsoncpp, even if performance weren't a
factor.

Since RapidJSON is a header-only library, it was easy to vendor it into
the JsonStuff project and make the transition without introducing any
external dependencies, so I went ahead and did it.

Cheers,
Andrew

[1] https://github.com/miloyip/nativejson-benchmark