Re: [Qemu-devel] RFC: libyajl for JSON

[Eric Blake]

On 11/02/2015 12:10 PM, Markus Armbruster wrote:

>>> * Single-quoted strings

> 
>> So that is an absolute must for whatever parser we choose.  My first
>> glance at libyajl is that it does NOT currently allow single quotes (not
>> even with any of its existing options), so we'd have to pre-process
>> input before feeding it to yajl.

>>
>> I'll ask on the yajl mailing lists, to get a feel for community
>> receptiveness, before attempting to actually write patches on that front.
> 
> Makes sense.

On IRC, I got pointed to a couple of forks that have already done this,
such as:

https://github.com/ludocode/yajl/commits/master

although the upstream author Lloyd was not on the channel at the time.
Meanwhile, there's 47 open issues on the upstream repo:

https://github.com/lloyd/yajl/issues

which implies slow response by the author, but at least he WAS asking if
anyone would like to help him with maintenance:

https://github.com/lloyd/yajl/issues/161

|  lloyd commented on Sep 24
| ok, give me a minute to hand you a patch on a branch to review.

|  lloyd commented on Sep 24
| ok, I'll merge down and roll a new release within a day.

[still hasn't happened yet...]

|  lamont-granquist commented on Sep 25
| hey @lloyd would you consider adding other maintainers?

|  lloyd commented on Oct 1
| I would absolutely consider additional maintainers. Who's interested?

so I don't know what the future holds for extending things upstream.

To date, I don't have a github account, by conscious personal choice; so
I can't really take him up on that offer directly.  So far, I've been
trying the mailing list to see if he answers that in addition to the
github PR stream, but the archives show it to be pretty full of spam:
http://librelist.com/browser/yajl/

>>
>> Yes; the yajl parser has 4 modes of parse operation based on which of
>> three callbacks you implement: double-only (yajl_double), long long-only
>> (yajl_integer), double-and-int (both yajl_double and yajl_integer, not
>> sure which one has precedence if input satisfies both formats), or
>> custom number (yajl_number, which is given a string, and you turn it
>> into the format of your choice).  Likewise, the yajl formatter has 2
>> functions: yajl_gen_double() (formats doubles) and yajl_gen_number()
>> (you provide literal strings formatted how you like).
> 
> Good.

And one of the open bugs on the tracker was the same one we've
encountered ourselves: yajl is locale-sensitive when using strtod() for
parsing and when using printf() for printing doubles:

https://github.com/lloyd/yajl/issues/79

[I would love for C/POSIX to add strtod_l and printf_l, but that's a
thread for another day]

> 
>>> More extensions or pitfalls might be lurking in our parser.  Therefore,
>>> replacing our parser by a suitable library is not without risk.  I guess
>>> we could do it over a full development cycle.  No way for 2.5.
>>>
>>
>> Absolutely not for 2.5; we've missed softfreeze, and this would count as
>> a new feature.

Another interesting bug report against yajl: one reporter made the point
[1] that yajl is already a superset of the canonical RFC 4627 definition
of JSON [2], because while the RFC states that a JSON document has this
terminal state:
      JSON-text = object / array
yajl will accept _any_ JSON value (not just objects/arrays) as the
overall document.

[1] https://github.com/lloyd/yajl/issues/154
[2] https://www.ietf.org/rfc/rfc4627.txt

So at this point, I want to see if lloyd makes any progress towards an
actual yajl release and/or adding a co-maintainer, before even trying to
get formal upstream support for single quoting.  We could always create
a git submodule with our own choice of fork (since there are already
forks that do single-quote parsing) - but the mantra of 'upstream first'
has a lot of merit (I'm reluctant to fork without good reason).

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature