Re: [Cbor] The three horses (Re: Implementing float->int numeric reduction)

[Anders Rundgren <[e-mail]>]

On 2023-08-17 15:50, Joe Hildebrand wrote:
> 
>> On Aug 17, 2023, at 2:37 AM, Anders Rundgren <[e-mail]> wrote:
>>
>> Imagine MSFT, Google, Amazon, and the JDK folks begin integrating CBOR support in their respective platforms and/or cloud services.  In their worlds, merging (1) and (2) should be a no-brainier.
> 
> Not if there is a performance impact from 2, which there is.

I guess you are talking about map sorting?  Just for fun (well...), I did a test with a map having 100000 keys.
With the worst case scenario (with respect to the sorting algorithm), I ended up with 15 seconds for running all set (key, fixedObject) operations.
With presorting I got 10 ms.  Presorting isn't pretty, but it makes insertion time per item independent of map size.

For the more normal case with a dozen keys, I can't imagine that map sorting would even be measurable.
You won't get around duplicate testing though, regardless if you are in (1), (2), or (3).

Anders

> 
>> With some kind of configuration facility, (1) and (3) also seems plausible.  In my case it may be down to two tests controlling 1 + 20 lines of code (after some minor restructuring).
> 
> It was MUCH more code for me, particularly to fire all of the required errors.  Perhaps that was because I had to refit it into a system that didn't anticipate those kinds of errors or configuration, I suppose.
> 
> —
> Joe Hildebrand
>