Re: [PATCH] Add new lisp function length= with bytecode support

[Stefan Monnier]

> This is beginning to feel like bike-shedding.

Indeed.  I guess part of the problem is a lack of direction: is there
some benchmark or test case that drives your desire to add that bytecode?

You say that you've found a large proportion of calls to `length` where
the result is then compared to a number, so it could be sped up.  I do
not doubt that it is the case, but the question is whether this speed up
makes a difference in practice.  If it does, then we can compare the
different options.

BTW, AFAIK there are 4 options (in order of increasing performance):
- leave things as they are.
- implement length= in Elisp.
- implement length= in C (but without dedicating a bytecode to it).
- implement length= as a bytecode.


        Stefan