Re: Generalizing find-definition

[Yuri Khan]

On Fri, Nov 7, 2014 at 1:40 AM, Stephen Leake
<address@hidden> wrote:

>> I think it should return "baz" here, if we want to go this way at all.
>> And then that value could be passed to a "search-definition-function",
>> which will perform a full non-precise scan.
>
> When would that be useful?

When function names are unique enough between interfaces and the
parser is not good enough to precisely identify which subset of
classes the current invocation may involve. The search-definition
engine may then say “oh screw it, I got confused, here are all
definitions of baz in all of your bazillion classes, figure it out
yourself, you’re human after all”.

>> It apparent that we won't be able to do precise searches for identifiers
>> in most cases,
>
> It requires support from the compiler/backend, but that should always be
> available (presumably you have the compiler if you are editing the
> code). So why do you think we can't do precise searches in most cases?

Some compilers don’t expose the necessary information in a
machine-readable form, or are not sufficiently fast to produce a
result within the interval the user is prepared to wait.