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

[João Távora]

Eli Zaretskii <eliz@gnu.org> writes:


>> But as I said this is just a nit.
>
> I made one more fix.

FTR, I think it's way worse now. The function variable has no clear
protocol: it's very odd to state its return type as number of code units
_or_ bytes _or_ code points.  A good docstring for such a variable notes
that, for any buffer position, a plugged-in function must return the
same _type_ but may return a different _value_ to match the
unit-counting strategy being used by the LSP server.

I haven't reverted to avoid a commit war, and because I'm a bit weary of
this bug.  I leave a patch for consideration and for the record, feel
free to ignore it.

João

diff --git a/lisp/progmodes/eglot.el b/lisp/progmodes/eglot.el
index dd84f545ed4..12fb72503aa 100644
--- a/lisp/progmodes/eglot.el
+++ b/lisp/progmodes/eglot.el
@@ -1453,11 +1453,20 @@ eglot--warn
 (defvar eglot-current-linepos-function #'eglot-utf-16-linepos
   "Function calculating position relative to line beginning.
 
-This is the inverse of `eglot-move-to-linepos-function' (which see).
-It is a function of no arguments returning the number of code units
-or bytes or codepoints corresponding to the current position of point,
-relative to line beginning, as expected by the function that is the
-value of `eglot-move-to-linepos-function'.")
+This is the inverse operation of
+`eglot-move-to-linepos-function' (which see).  The value is a
+nullary function that examines current `point' and computes the
+number of code units relative to line beginning in such a way
+that it matches how the LSP server also counts code units.  For
+example, if the LSP server is using UTF-16 encoding, and point is
+on the \"b\" of a line containing just \"aXbc\" where \"X\" is a
+funny looking character in the UTF-16 \"supplementary plane\",
+this function has to return 3 code points instead of 2.  If the
+server is using UTF-32 encoding, this function should return 2
+code points.
+
+Since LSP 3.17 server and client may agree on an encoding and
+Eglot will set this variable automatically.")
 
 (defun eglot-utf-8-linepos ()
   "Calculate number of UTF-8 bytes from line beginning."
@@ -1496,11 +1505,11 @@ eglot-move-to-linepos-function
   "Function to move to a position within a line reported by the LSP server.
 
 Per the LSP spec, character offsets in LSP Position objects count
-UTF-16 code units, not actual code points.  So when LSP says
-position 3 of a line containing just \"aXbc\", where X is a funny
-looking character in the UTF-16 \"supplementary plane\", it
-actually means `b', not `c'.  The default value
-`eglot-move-to-utf-16-linepos' accounts for this.
+code units, not actual code points.  The default is to use UTF-16
+encoding, so when LSP says position 3 of a line containing just
+\"aXbc\", where \"X\" is a funny looking character in the UTF-16
+\"supplementary plane\", it actually means \"b\", not \"c\".  The
+default value `eglot-move-to-utf-16-linepos' accounts for this.
 
 This variable can also be set to `eglot-move-to-utf-8-linepos' or
 `eglot-move-to-utf-32-linepos' for servers not closely following