Re: I created a faster JSON parser

[Herman , Géza]

Philip Kaludercic <philipk@posteo.net> writes:

> "Herman, Géza" <geza.herman@gmail.com> writes:
>
> > It replaces json-parse-string and json-parse-buffer 
> > functions. The
> > behavior should be the same as before, with the only exception 
> > that
> > objects with duplicated keys are not detected if :object-type 
> > is not
> > 'hash-table.
>
> Is that a problem?
Not sure.  I just mentioned it because it's a behavior change. 
But I intentionally designed it this way, because it is faster. 
To me, it makes some sense that if the user specifies 'alist or 
'plist, then they want to have all the object members, even if the 
keys are duplicated.  I didn't find a clear direction from JSON 
descriptions of how duplicated keys should be handled.

> > This parser runs 8-9x faster than the jansson based parser on 
> > my
> > machine (tested on clangd language server messages).  An 
> > additional
> > tiny benefit is that large integers are parsed, instead of 
> > having an
> > "out of range" error.
>
> That sounds interesting, but I am reminded of this article:
> https://seriot.ch/projects/parsing_json.html.  There seem to be 
> plenty
> of difficult edge-cases when dealing with JSON input, that 
> should
> probably be tested if Emacs has it's own custom parser built-in.
I've now run my parser on the tests in this repo, it passes all of 
them.