bug#61726: [PATCH] Eglot: Support positionEncoding capability

[João Távora]

On Fri, Feb 24, 2023 at 11:38 AM Eli Zaretskii <eliz@gnu.org> wrote:
> > `eglot-lsp-abiding-column' allocates a new string!
>
> If that is a concern, eglot.el could use a private temporary buffer
> into which the encoded text is inserted, eliminating the need to call
> 'length'.  The impact of that in performance should be measured, of
> course, to make sure it doesn't make code slower; it will definitely
> improve the GC pressure aspect.

Indeed, maybe something like this (untested):

diff --git a/lisp/progmodes/eglot.el b/lisp/progmodes/eglot.el
index e20d209332d..78e0f9c1f0c 100644
--- a/lisp/progmodes/eglot.el
+++ b/lisp/progmodes/eglot.el
@@ -1454,11 +1454,15 @@ eglot-current-column-function
 (defun eglot-lsp-abiding-column (&optional lbp)
   "Calculate current COLUMN as defined by the LSP spec.
 LBP defaults to `line-beginning-position'."
-  (/ (- (length (encode-coding-region (or lbp (line-beginning-position))
-                                      ;; Fix github#860
-                                      (min (point) (point-max)) 'utf-16 t))
-        2)
-     2))
+  (let ((measure (with-current-buffer (get-buffer-create "
*eglot-utf16-measure*")
+                   (erase-buffer)
+                   (current-buffer))))
+    (/ (- (encode-coding-region (or lbp (line-beginning-position))
+                                ;; Fix github#860
+                                (min (point) (point-max)) 'utf-16
+                                measure)
+          2)
+       2)))

 (defun eglot--pos-to-lsp-position (&optional pos)
   "Convert point POS to LSP position."