On Jan 6, 2014, at 13:49 , Geoffrey Garen 
<gga...@apple.com<mailto:gga...@apple.com>> wrote:

Let me try to clarify with two more motivating examples:

(1) Nodes.cpp:

FunctionParameters::FunctionParameters(ParameterNode* firstParameter, unsigned 
size)
    : m_size(size)
{
    unsigned i = 0;
    for (ParameterNode* parameter = firstParameter; parameter; parameter = 
parameter->nextParam()) {
        auto pattern = parameter->pattern();
        pattern->ref();
        patterns()[i++] = pattern;
    }
}

If I had to describe this algorithm in English, I’d say, “Collect and retain 
all the [auto] from the list of parsed parameters.” I think that explanation 
would be stronger if “[auto]” were a concrete noun.

Does anybody prefer auto in this context? If so, why?

While I wouldn't say that I prefer auto here, it doesn't really bother me in 
this example. Personally, I would read that as "Collect and retain all of the 
patterns from the list of parsed parameters", and I'd say it the same way if 
auto had been an explicit type.

(2) ApplyStyleCommand.cpp:

    auto children = elementChildren(*dummySpanAncestor);
    for (auto child = children.begin(), end = children.end(); child != end; 
++child) {
        if (isSpanWithoutAttributesOrUnstyledStyleSpan(&*child))
            toRemove.append(&*child);
    }

I don’t understand why we’re &*’ing here. That’s a surprising idiom I haven’t 
seen before, which I would expect to be a no-op. My first question when reading 
this is, “What is the type of ‘child’, such that I would need to &* it?”.

Is this "&*child” obvious to everyone else?

It's actually pretty obvious to me, but I've been spending a lot of time with 
similar iterators lately. I don't think it's that much clearer without the 
auro. I've definitely found this to be confusing in the codebase much before 
auto entered the picture. (In this case, I'd argue strongly for using an 
accessor method on the iterator or a temporary variable)

- Bem

_______________________________________________
webkit-dev mailing list
webkit-dev@lists.webkit.org
https://lists.webkit.org/mailman/listinfo/webkit-dev

Reply via email to