Re: I created a faster JSON parser

[Herman , Géza]

Mattias Engdegård <mattias.engdegard@gmail.com> writes:

> 15 mars 2024 kl. 14.35 skrev Herman, Géza 
> <geza.herman@gmail.com>:
>
> > I implemented this idea, here is the latest version (now, 
> > object_workspace is a Lisp_Object):
>
> That's considerably slower than before, especially for small 
> inputs, so it's probably a no-go.
Have you benchmarked it?  I really doubt that there is a 
significant performance difference.  Maybe it's possible to find 
some case, where this modification matters a bit, but for most 
cases, the difference should be negligible. The amount of 
allocations done is the same in both cases, which is for large 
files, only 20-30 allocations.  Compared to the thousands of 
allocations for storing the actual Lisp objects.

> Here are some remaining tasks:
>
> 2. Don't allocate any temporary storage before you know that 
> it's actually necessary.
Except for the very-very trivial cases, memory allocation is 
always necessary.  I can lower the allocation sizes, but I don't 
think it's worth it. We're only talking about sizes of the KB 
range.

> 3. Stop using the object_workspace when not required, which is 
> everywhere except possibly when reading arrays into Lisp 
> vectors.
object_workspace is necessary whenever a JSON contains an object 
or array.  So basically always.