Re: Aliasing EQ to EQL (was: [RFC PATCH] Per-window face support)

[Daniel Colascione]

>>> Getting rid of `eq' is just lying to programmers about object identity.
>>
>> I don't see how it would be lying to equate eq and eql for numbers.  An
>> object's identity is not the same thing as a machine-level address, and
>> whether two instances of the same number are eq is an implementation
>> detail
>> that Lisp programmers should not rely upon.
>>
>> This doesn't mean that we should equate eq and eql. Perhaps there are
>> good
>> efficiency reasons to continue to distinguish them. But these would be
>> merely pragmatic, not philosophical.
>
> BTW, if someone is tempted to measure the impact, here's the naive patch
> I've been using recently.
>
> It doesn't try to be clever: other than NILP, all uses of EQ are changed
> to use EQL.  Clearly, we could improve on that, but I think such
> improvements should be "profile-guided".
>
> In terms of code size it added 100KB (out of a 4MB stripped binary, so
> about 0.25%) which is not insignificant, but is a cost I'm willing to
> live with.

For what? This aliasing has _zero_ real world benefit. I maintain that
this is a change we should absolutely not make.