Re: [PATCH] Allow inserting non-BMP characters

[Philipp Stephani <p.stephani2@gmail.com>]

[-- Attachment #1.1: Type: text/plain, Size: 1229 bytes --]

Eli Zaretskii <eliz@gnu.org> schrieb am Do., 28. Dez. 2017 um 17:29 Uhr:

> > From: Philipp Stephani <p.stephani2@gmail.com>
> > Date: Thu, 28 Dec 2017 12:31:39 +0000
> > Cc: phst@google.com, Eli Zaretskii <eliz@gnu.org>, emacs-devel@gnu.org
> >
> > Alan Third <alan@idiocy.org> schrieb am Do., 28. Dez. 2017 um 12:38 Uhr:
> >
> >  On Wed, Dec 27, 2017 at 05:41:23AM +0200, Eli Zaretskii wrote:
> >  > What about the possibility that SAFE_NALLOCA could signal an error and
> >  > longjmp to top level?  Does this code always run in the main thread,
> >  > and if so, can it allow such longjmp's?
> >
> >  I think it does always run in the main thread, however since it runs
> >  within the NSApplication run loop I’ve no idea what would happen if we
> >  did a longjmp.
> >
> > We should probably avoid longjmps here. This message is invoked by the
> window manager, which most likely
> > can't deal with longjmps.
>
> Then maybe process the input in chunks using a fixed-size buffer?
>

Maybe, but for now I've reverted to just call characterAtIndex repeatedly.
I guess in 99% of cases we insert a single code unit anyway, and if we plan
to make the code more complex we should benchmark it before.

[-- Attachment #1.2: Type: text/html, Size: 1891 bytes --]

[-- Attachment #2: 0001-Allow-inserting-non-BMP-characters.txt --]
[-- Type: text/plain, Size: 4797 bytes --]

From 383e6fe7c50c1e3a96720284fa5fc2c00610959f Mon Sep 17 00:00:00 2001
From: Philipp Stephani <phst@google.com>
Date: Mon, 25 Dec 2017 22:00:00 +0100
Subject: [PATCH] Allow inserting non-BMP characters

* src/coding.h (char_low_surrogate_p, char_high_surrogate_p)
(surrogates_to_codepoint): New functions.

* src/coding.c (decode_coding_utf_16): Use new inline functions;
remove old macros.

* src/nsterm.m (insertText:): Properly handle surrogate pairs.
---
 src/coding.c | 13 +++----------
 src/coding.h | 26 ++++++++++++++++++++++++++
 src/nsterm.m | 25 +++++++++++++++++++------
 3 files changed, 48 insertions(+), 16 deletions(-)

diff --git a/src/coding.c b/src/coding.c
index 1705838ffa..9903d87b92 100644
--- a/src/coding.c
+++ b/src/coding.c
@@ -1515,13 +1515,6 @@ encode_coding_utf_8 (struct coding_system *coding)
 /* See the above "GENERAL NOTES on `detect_coding_XXX ()' functions".
    Return true if a text is encoded in one of UTF-16 based coding systems.  */
 
-#define UTF_16_HIGH_SURROGATE_P(val) \
-  (((val) & 0xFC00) == 0xD800)
-
-#define UTF_16_LOW_SURROGATE_P(val) \
-  (((val) & 0xFC00) == 0xDC00)
-
-
 static bool
 detect_coding_utf_16 (struct coding_system *coding,
 		      struct coding_detection_info *detect_info)
@@ -1686,7 +1679,7 @@ decode_coding_utf_16 (struct coding_system *coding)
 
       if (surrogate)
 	{
-	  if (! UTF_16_LOW_SURROGATE_P (c))
+	  if (! char_low_surrogate_p (c))
 	    {
 	      if (endian == utf_16_big_endian)
 		c1 = surrogate >> 8, c2 = surrogate & 0xFF;
@@ -1694,7 +1687,7 @@ decode_coding_utf_16 (struct coding_system *coding)
 		c1 = surrogate & 0xFF, c2 = surrogate >> 8;
 	      *charbuf++ = c1;
 	      *charbuf++ = c2;
-	      if (UTF_16_HIGH_SURROGATE_P (c))
+	      if (char_high_surrogate_p (c))
 		CODING_UTF_16_SURROGATE (coding) = surrogate = c;
 	      else
 		*charbuf++ = c;
@@ -1708,7 +1701,7 @@ decode_coding_utf_16 (struct coding_system *coding)
 	}
       else
 	{
-	  if (UTF_16_HIGH_SURROGATE_P (c))
+	  if (char_high_surrogate_p (c))
 	    CODING_UTF_16_SURROGATE (coding) = surrogate = c;
 	  else
 	    {
diff --git a/src/coding.h b/src/coding.h
index 66d125b07e..b111ce70c6 100644
--- a/src/coding.h
+++ b/src/coding.h
@@ -662,6 +662,32 @@ struct coding_system
 /* Note that this encodes utf-8, not utf-8-emacs, so it's not a no-op.  */
 #define ENCODE_UTF_8(str) code_convert_string_norecord (str, Qutf_8, true)
 
+/* Return true if C is a low surrogate.  */
+
+INLINE bool
+char_low_surrogate_p (int c)
+{
+  return 0xDC00 <= c && c <= 0xDFFF;
+}
+
+/* Return true if C is a high surrogate.  */
+
+INLINE bool
+char_high_surrogate_p (int c)
+{
+  return 0xD800 <= c && c <= 0xDBFF;
+}
+
+/* Return the Unicode code point for the given UTF-16 surrogates.  */
+
+INLINE int
+surrogates_to_codepoint (int low, int high)
+{
+  eassert (char_low_surrogate_p (low));
+  eassert (char_high_surrogate_p (high));
+  return 0x10000 + (low - 0xDC00) + ((high - 0xD800) * 0x400);
+}
+
 /* Extern declarations.  */
 extern Lisp_Object code_conversion_save (bool, bool);
 extern bool encode_coding_utf_8 (struct coding_system *);
diff --git a/src/nsterm.m b/src/nsterm.m
index 07ac8f978f..0c0b3fd9b5 100644
--- a/src/nsterm.m
+++ b/src/nsterm.m
@@ -6283,14 +6283,13 @@ flag set (this is probably a bug in the OS).
          by doCommandBySelector: deleteBackward: */
 - (void)insertText: (id)aString
 {
-  int code;
-  int len = [(NSString *)aString length];
-  int i;
+  NSString *s = aString;
+  NSUInteger len = [s length];
 
   NSTRACE ("[EmacsView insertText:]");
 
   if (NS_KEYLOG)
-    NSLog (@"insertText '%@'\tlen = %d", aString, len);
+    NSLog (@"insertText '%@'\tlen = %lu", aString, (unsigned long) len);
   processingCompose = NO;
 
   if (!emacs_event)
@@ -6300,10 +6299,24 @@ - (void)insertText: (id)aString
   if (workingText != nil)
     [self deleteWorkingText];
 
+  /* It might be preferable to use getCharacters:range: below,
+     cf. https://developer.apple.com/library/content/documentation/Cocoa/Conceptual/CocoaPerformance/Articles/StringDrawing.html#//apple_ref/doc/uid/TP40001445-112378.
+     However, we probably can't use SAFE_NALLOCA here because it might
+     exit nonlocally.  */
+
   /* now insert the string as keystrokes */
-  for (i =0; i<len; i++)
+  for (NSUInteger i = 0; i < len; i++)
     {
-      code = [aString characterAtIndex: i];
+      NSUInteger code = [s characterAtIndex:i];
+      if (char_high_surrogate_p (code) && i < len - 1)
+        {
+          unichar low = [s characterAtIndex:i + 1];
+          if (char_low_surrogate_p (low))
+            {
+              code = surrogates_to_codepoint (low, code);
+              ++i;
+            }
+        }
       /* TODO: still need this? */
       if (code == 0x2DC)
         code = '~'; /* 0x7E */
-- 
2.15.1