Re: Performance issue w/ `cl-loop`s `collect...into`

[Tianxiang Xiong]

Since it appears there's no performance degradation, can we merge this?

If others would like to do some testing, that'd be great.

On Mon, Apr 9, 2018 at 8:33 AM, Tianxiang Xiong <address@hidden> wrote:

I also figured out what the `(progn ... t)` pattern does. Here's a macroexpansion w/ it:

(macroexpand-1 '(cl-loop for i in '(1 2 3) collect i))

(cl-block nil

  (let*

      ((--cl-var--

        '(1 2 3))

       (i nil)

       (--cl-var-- nil))

    (while

        (consp --cl-var--)

      (setq i

            (car --cl-var--))

      (push i --cl-var--)

      (setq --cl-var--

            (cdr --cl-var--)))

    (nreverse --cl-var--)))

And one w/out it:

(macroexpand-1 '(cl-loop for i in '(1 2 3) collect i))

(cl-block nil

  (let*

      ((--cl-var--

        '(1 2 3))

       (i nil)

       (--cl-var-- nil))

    (while

        (and

         (consp --cl-var--)

         (progn

           (setq i

                 (car --cl-var--))

           (push i --cl-var--)))

      (setq --cl-var--

            (cdr --cl-var--)))

    (nreverse --cl-var--)))

Apparently `cl--loop-build-ands` uses this pattern to determine whether clauses go into an `and` or not 🤷. Of course 🙄!

On Mon, Apr 9, 2018 at 8:28 AM, Tianxiang Xiong <address@hidden> wrote:

Here's the result of `benchmark-run` on the new code:

(benchmark-run 10 (cl-loop for i in (number-sequence 1 1E6)

                     collect i))

;; => (10.488984668 6 3.555157208999999)

and old code:

(benchmark-run 10 (cl-loop for i in (number-sequence 1 1E6)

                     collect i))

;; => (14.876455789000001 25 2.328459104999999)

So there actually seems to be an improvement due to less GC.

On Mon, Apr 9, 2018 at 5:22 AM, Basil L. Contovounesios <address@hidden> wrote:

Tianxiang Xiong <address@hidden> writes:

> Is there a function to easily time operations in Emacs Lisp? Something
> like Clojure's `core/time`?

There are the macros benchmark-run and benchmark-run-compiled, as
mentioned under (info "(elisp) Profiling").

--
Basil