On 2015/12/04 15:16:33, rossberg wrote:
On 2015/12/04 14:35:32, caitp wrote:
> On 2015/12/04 14:34:43, caitp wrote:
> > On 2015/12/04 14:32:33, rossberg wrote:
> > > On 2015/12/04 14:26:09, caitp wrote:
> > > > On 2015/12/04 14:24:21, rossberg wrote:
> > > > > The main thing I'm confused about is why we need the rewriting
queue.
> What
> > > > > prevents you from invoking the rewriting right at the time
where you
> > > currently
> > > > > just queue it up?
> > > >
> > > > The main thing is the ambiguity `({ x } = { x: 1 }) => x` <<
where the
> > meaning
> > > > changes depending on whether `=>` is found or not. Binding
patterns
are
> > > > rewritten differently, so by necessity the rewriting is deferred
until
> it's
> > > > known for sure what is being rewritten
> > >
> > > But is anything ever removed from this queue without being
rewritten? I
> can't
> > > find that happening anywhere, so isn't it always performed
unconditionally
> > > anyway, just delayed?
> >
> > Hmm, that's a good point, right now there might be some redundant
rewriting
> > happening. I'm not sure it should be removed from the queue because it
can't
> be
> > easily, but it can be marked as rewritten to prevent redundant
rewriting
>
> --- but, just to be clear, that redundant rewriting doesn't mean that
it's
able
> to eagerly rewrite, it just means that the redundant rewriting is sort
of
> discarded, because the RewritableAssignmentExpression itself is
discarded
I see. The deeper reason I'm asking is because Niko is now implementing
the
rewriting for array literals with spread, which has a similar problem.
The way
I
thought this would work is that we introduce something dual to the pattern
rewriter, i.e., an expression visitor that is invoked when an ambiguous
cover-expression/pattern is resolved to be an expression. That would then
walk
and rewrite the necessary expressions inside. (In most cases that would
be a
no-op, so it's probably worth introducing a bit in the expression
classifier
to
indicate the necessity for rewriting.)
You already have that kind of visitor, it seems, so I wonder why it isn't
possible to invoke it directly at the resolution points.
It may be possible, I've been looking at this CL for too long now, almost 4
months. Should it wait to see if I can find a better way to deal with the
cover
grammar?
https://codereview.chromium.org/1309813007/
--
--
v8-dev mailing list
v8-dev@googlegroups.com
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to v8-dev+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
