Re: [Qemu-devel] [PATCH v5 17/28] qapi: Allow true, false and null in schema json

[Kevin Wolf <kwolf@redhat.com>]

Am 01.04.2015 um 13:03 hat Markus Armbruster geschrieben:
> Kevin Wolf <kwolf@redhat.com> writes:
> 
> > Am 01.04.2015 um 11:33 hat Markus Armbruster geschrieben:
> >> Kevin Wolf <kwolf@redhat.com> writes:
> >> 
> >> > Am 31.03.2015 um 22:09 hat Markus Armbruster geschrieben:
> >> >> Kevin Wolf <kwolf@redhat.com> writes:
> >> >> 
> >> >> > Am 24.03.2015 um 21:03 hat Eric Blake geschrieben:
> >> >> >> From: Fam Zheng <famz@redhat.com>
> >> >> >> 
> >> >> >> In the near term, we will use it for a sensible-looking
> >> >> >> 'gen':false inside command declarations, instead of the
> >> >> >> current ugly 'gen':'no'.
> >> >> >> 
> >> >> >> In the long term, it will allow conversion from shorthand
> >> >> >> with defaults mentioned only in side-band documentation:
> >> >> >>  'data':{'*flag':'bool', '*string':'str'}
> >> >> >> into an explicit default value documentation, as in:
> >> >> >>  'data':{'flag':{'type':'bool', 'optional':true, 'default':true},
> >> >> >>          'string':{'type':'str', 'optional':true, 'default':null}}
> >> >> >
> >> >> > FWIW, I don't think that's a very friendly syntax for humans, it's a bit
> >> >> > verbose. But that's no reason not to allow true/false/null, of course.
> >> >> 
> >> >> Here's my current thinking.
> >> >> 
> >> >> Longhand:
> >> >> 
> >> >>     # mandatory
> >> >>     'name': { 'type': 'str' }
> >> >>     # optional, with a default
> >> >>     'flag': { 'type': 'bool', 'default': true }
> >> >>     # optional, no default
> >> >>     'string': { 'type': 'str', 'default': null }
> >> >> 
> >> >> Presence of 'default' implies optional.
> >> >> 
> >> >> Equivalent shorthand, if any:
> >> >> 
> >> >>     'name': 'str'
> >> >>     '*string': 'str'
> >> >
> >> > A nice shorthand for defaults would be:
> >> >
> >> >     '*name': 'str' = 'default'
> >> >
> >> > Though that would be neither valid JSON nor Python any more. Do we
> >> > actually rely on this property anywhere or is it only parsed by the QAPI
> >> > generator anyway and we can extend the language in such ways?
> >> 
> >> I guess JSON / Python was chosen as QAPI schema language to save us the
> >> bother of defining a syntax and building the tools to work with it, like
> >> an Emacs mode.  JSON's not exactly my favourite choice, but at least
> >> it's not XML.
> >> 
> >> What we have now isn't JSON, but it's still a subset of Python, and the
> >> Python tools work.  If we go beyond Python, they'll break.
> >> 
> >> If we decide to sacrifice these tools for readability, then we can just
> >> as well replace the syntax entirely.  Preferably by something where I
> >> don't have to put every identifier in quotes.
> >> 
> >> In short, you're welcome to hack up qapi.py some more for schema
> >> readability, but either keep Emacs Python mode working, or provide a
> >> replacement :)
> >
> > What features does this Python mode provide that you use?
> 
> Syntax highlighting, automatic indentation, possibly more I rely on
> without noticing.  My fingers know, but they don't talk.
> 
> > For the JSON schema, the only thing I really use from the editor is
> > syntax highlighting, and that shouldn't be bothered by an addition like
> > this (in vim it doesn't hurt anyway). I also don't use any other Python
> > tools, though I'm not sure that none exist that would make sense with
> > the schema.
> >
> > So if there are practically relevant advantages in being real Python
> > code, then we need to consider that, of course. That's why I was asking.
> > But if there aren't, it's just an arbitrary restriction that could be
> > removed.
> 
> If we decide to revise the decision to borrow existing syntax and roll
> our own instead, let's 'make' 'our' 'own' 'syntax' 'not' 'suck'.
> 
> Anyway, we got bigger fish to fry right now.

Okay, I got it. Asking me to create a completely new syntax and the
generator for it is the longhand version for "no".

Kevin