bug#63861: [PATCH] pp.el: New "pretty printing" code

[Stefan Monnier via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

>> The new code is in a new function `pp-region`.
>> The old code redirects to the new code if `pp-buffer-use-pp-region` is
>> non-nil, tho I'm not sure we want to bother users with such
>> a config var.  Hopefully, the new code should be good enough that users
>> don't need to choose.  Maybe I should make it a `defvar` and have it
>> default to t, so new users will complain if it's not good enough?
>
> I tried your code and it looks very slow (but looks nice once printed).
> Testing on my bookmark-alist printed in some buffer.
> Here with a slightly modified version of pp-buffer (not much faster than
> the original one):
>
> (benchmark-run-compiled 1 (pp-buffer))
> => (6.942135047 0 0.0)
> And here with your version (using pp-region):
> (benchmark-run-compiled 1 (pp-buffer))
> => (46.141411097 0 0.0)

Hmm... that's weird.  With the bookmark-alist.el file you sent me,
I get pretty much the reverse result:

    % for f in foo.el test-load-history.el test-bookmark-alist.el; do \
          for v in nil t; do \
              time src/emacs -Q --batch "$f" -l pp \
                   --eval "(progn (setq pp-buffer-use-pp-region $v) \
                             (message \"%s %s %S\" (file-name-nondirectory buffer-file-name) \
                                                   pp-buffer-use-pp-region \
                                                   (benchmark-run (pp-buffer))))"; \
      done; done
    foo.el nil (0.210123295 1 0.057426385)
    src/emacs -Q --batch "$f" -l pp --eval   0.34s user 0.04s system 99% cpu 0.382 total
    foo.el t (0.07107641199999999 0 0.0)
    src/emacs -Q --batch "$f" -l pp --eval   0.19s user 0.03s system 99% cpu 0.222 total
    test-load-history.el nil (156.07942386099998 17 1.432754161)
    src/emacs -Q --batch "$f" -l pp --eval   156.04s user 0.20s system 99% cpu 2:36.26 total
    test-load-history.el t (96.480110987 24 1.9799413479999999)
    src/emacs -Q --batch "$f" -l pp --eval   96.40s user 0.27s system 99% cpu 1:36.69 total
    test-bookmark-alist.el nil (51.211047973 8 0.6690439610000001)
    src/emacs -Q --batch "$f" -l pp --eval   51.29s user 0.11s system 99% cpu 51.401 total
    test-bookmark-alist.el t (5.110458941 6 0.468187075)
    src/emacs -Q --batch "$f" -l pp --eval   5.21s user 0.09s system 99% cpu 5.302 total
    %

This is comparing "the old pp-buffer" with "the new `pp-region`", using
the patch below.  This is not using your `tv/pp-region` code.

I find these results (mine) quite odd: they suggest that my `pp-region`
is *faster* than the old `pp-buffer` for `load-history` and `bookmark-alist`
data, which I definitely did not expect (and don't know how to explain
either).

These tests were run on a machine whose CPU's speed can vary quite
drastically depending on the load, so take those numbers with a grain of
salt, but the dynamic frequency fluctuations shouldn't cause more than
a factor of 2 difference (and according to my CPU frequency monitor
widget, the frequency was reasonably stable during the test).

> For describe variable I use a modified version of `pp` which is very
> fast (nearly instant to pretty print value above) but maybe unsafe with
> some vars, didn't have any problems though, see
> https://github.com/thierryvolpiatto/emacs-config/blob/main/describe-variable.el.

So, IIUC the numbers you cite above compare my `pp-region` to your
`tv/pp-region`, right?  And do I understand correctly that
`tv/pp-region` does not indent its output?
What was the reason for this choice?


        Stefan


PS: BTW, looking at the output of `pp` on the bookmark-data, it's not
a test where `pp-region` shines: the old pp uses up more lines, but is
more regular and arguably more readable :-(


diff --git a/lisp/emacs-lisp/lisp-mode.el b/lisp/emacs-lisp/lisp-mode.el
index d44c9d6e23d..9914ededb85 100644
--- a/lisp/emacs-lisp/lisp-mode.el
+++ b/lisp/emacs-lisp/lisp-mode.el
@@ -876,7 +876,7 @@ lisp-ppss
 2 (counting from 0).  This is important for Lisp indentation."
   (unless pos (setq pos (point)))
   (let ((pss (syntax-ppss pos)))
-    (if (nth 9 pss)
+    (if (and (not (nth 2 pss)) (nth 9 pss))
         (let ((sexp-start (car (last (nth 9 pss)))))
           (parse-partial-sexp sexp-start pos nil nil (syntax-ppss sexp-start)))
       pss)))
diff --git a/lisp/emacs-lisp/pp.el b/lisp/emacs-lisp/pp.el
index e6e3cd6c6f4..89d7325a491 100644
--- a/lisp/emacs-lisp/pp.el
+++ b/lisp/emacs-lisp/pp.el
@@ -74,31 +74,127 @@ pp-to-string
       (pp-buffer)
       (buffer-string))))
 
+(defun pp--within-fill-column-p ()
+  "Return non-nil if point is within `fill-column'."
+  ;; Try and make it O(fill-column) rather than O(current-column),
+  ;; so as to avoid major slowdowns on long lines.
+  ;; FIXME: This doesn't account for invisible text or `display' properties :-(
+  (and (save-excursion
+         (re-search-backward
+          "^\\|\n" (max (point-min) (- (point) fill-column)) t))
+       (<= (current-column) fill-column)))
+
+(defun pp-region (beg end)
+  "Insert newlines in BEG..END to try and fit within `fill-column'.
+Presumes the current buffer contains Lisp code and has indentation properly
+configured for that.
+Designed under the assumption that the region occupies a single line,
+tho it should also work if that's not the case."
+  (interactive "r")
+  (goto-char beg)
+  (let ((end (copy-marker end t))
+        (newline (lambda ()
+                   (skip-chars-forward ")]}")
+                   (unless (save-excursion (skip-chars-forward " \t") (eolp))
+                     (insert "\n")
+                     (indent-according-to-mode)))))
+    (while (progn (forward-comment (point-max))
+                  (< (point) end))
+      (let ((beg (point))
+            ;; Whether we're in front of an element with paired delimiters.
+            ;; Can be something funky like #'(lambda ..) or ,'#s(...).
+            (paired (when (looking-at "['`,#]*[[:alpha:]]*\\([({[\"]\\)")
+                      (match-beginning 1))))
+        ;; Go to the end of the sexp.
+        (goto-char (or (scan-sexps (or paired (point)) 1) end))
+        (unless
+            (and
+             ;; The sexp is all on a single line.
+             (save-excursion (not (search-backward "\n" beg t)))
+             ;; And its end is within `fill-column'.
+             (or (pp--within-fill-column-p)
+                 ;; If the end of the sexp is beyond `fill-column',
+                 ;; try to move the sexp to its own line.
+                 (and
+                  (save-excursion
+                    (goto-char beg)
+                    (if (save-excursion (skip-chars-backward " \t({[',") (bolp))
+                        ;; The sexp was already on its own line.
+                        nil
+                      (skip-chars-backward " \t")
+                      (setq beg (copy-marker beg t))
+                      (if paired (setq paired (copy-marker paired t)))
+                      ;; We could try to undo this insertion if it
+                      ;; doesn't reduce the indentation depth, but I'm
+                      ;; not sure it's worth the trouble.
+                      (insert "\n") (indent-according-to-mode)
+                      t))
+                  ;; Check again if we moved the whole exp to a new line.
+                  (pp--within-fill-column-p))))
+          ;; The sexp is spread over several lines, and/or its end is
+          ;; (still) beyond `fill-column'.
+          (when (and paired (not (eq ?\" (char-after paired))))
+            ;; The sexp has sub-parts, so let's try and spread
+            ;; them over several lines.
+            (save-excursion
+              (goto-char beg)
+              (when (looking-at "(\\([^][()\" \t\n;']+\\)")
+                ;; Inside an expression of the form (SYM ARG1
+                ;; ARG2 ... ARGn) where SYM has a `lisp-indent-function'
+                ;; property that's a number, insert a newline after
+                ;; the corresponding ARGi, because it tends to lead to
+                ;; more natural and less indented code.
+                (let* ((sym (intern-soft (match-string 1)))
+                       (lif (and sym (get sym 'lisp-indent-function))))
+                  (if (eq lif 'defun) (setq lif 2))
+                  (when (natnump lif)
+                    (goto-char (match-end 0))
+                    (forward-sexp lif)
+                    (funcall newline)))))
+            (save-excursion
+              (pp-region (1+ paired) (1- (point)))))
+          ;; Now the sexp either ends beyond `fill-column' or is
+          ;; spread over several lines (or both).  Either way, the rest of the
+          ;; line should be moved to its own line.
+          (funcall newline))))))
+
+(defcustom pp-buffer-use-pp-region t
+  "If non-nil, `pp-buffer' uses the new `pp-region' code."
+  :type 'boolean)
+
 ;;;###autoload
 (defun pp-buffer ()
   "Prettify the current buffer with printed representation of a Lisp object."
   (interactive)
   (goto-char (point-min))
-  (while (not (eobp))
-    (cond
-     ((ignore-errors (down-list 1) t)
-      (save-excursion
-        (backward-char 1)
-        (skip-chars-backward "'`#^")
-        (when (and (not (bobp)) (memq (char-before) '(?\s ?\t ?\n)))
-          (delete-region
-           (point)
-           (progn (skip-chars-backward " \t\n") (point)))
-          (insert "\n"))))
-     ((ignore-errors (up-list 1) t)
-      (skip-syntax-forward ")")
-      (delete-region
-       (point)
-       (progn (skip-chars-forward " \t\n") (point)))
-      (insert ?\n))
-     (t (goto-char (point-max)))))
-  (goto-char (point-min))
-  (indent-sexp))
+  (if pp-buffer-use-pp-region
+      (with-syntax-table emacs-lisp-mode-syntax-table
+        (let ((fill-column (max fill-column 70))
+              (indent-line-function
+               (if (local-variable-p 'indent-line-function)
+                   indent-line-function
+                 #'lisp-indent-line)))
+          (pp-region (point-min) (point-max))))
+    (while (not (eobp))
+      (cond
+       ((ignore-errors (down-list 1) t)
+        (save-excursion
+          (backward-char 1)
+          (skip-chars-backward "'`#^")
+          (when (and (not (bobp)) (memq (char-before) '(?\s ?\t ?\n)))
+            (delete-region
+             (point)
+             (progn (skip-chars-backward " \t\n") (point)))
+            (insert "\n"))))
+       ((ignore-errors (up-list 1) t)
+        (skip-syntax-forward ")")
+        (delete-region
+         (point)
+         (progn (skip-chars-forward " \t\n") (point)))
+        (insert ?\n))
+       (t (goto-char (point-max)))))
+    (goto-char (point-min))
+    (indent-sexp)))
 
 ;;;###autoload
 (defun pp (object &optional stream)