From f831910c5cc536d3f68aa2f919d29935a187c47d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Uwe=20Kr=C3=BCger?=
 <45282134+UweKrueger@users.noreply.github.com>
Date: Mon, 4 May 2020 13:21:11 +0200
Subject: [PATCH] utf: fix string length calculation for combining characters

---
 vlib/builtin/utf8.v                      | 37 +++++++++++++++++++++++-
 vlib/v/gen/str.v                         |  4 +--
 vlib/v/tests/string_interpolation_test.v |  5 ++++
 3 files changed, 43 insertions(+), 3 deletions(-)

diff --git a/vlib/builtin/utf8.v b/vlib/builtin/utf8.v
index dbf6fb349b..545c99c23c 100644
--- a/vlib/builtin/utf8.v
+++ b/vlib/builtin/utf8.v
@@ -174,7 +174,7 @@ fn utf8_len(c byte) int {
 	return b
 }
 
-// Calculate string length for formatting, i.e. number of "characters"
+// Calculate string length for in number of codepoints
 fn utf8_str_len(s string) int {
 	mut l := 0
 	for i := 0; i < s.len; i++ {
@@ -189,6 +189,41 @@ fn utf8_str_len(s string) int {
 	return l
 }
 
+// Calculate string length for formatting, i.e. number of "characters"
+fn utf8_str_visible_length(s string) int {
+	mut l := 0
+	mut ul := 1
+	for i := 0; i < s.len; i+=ul {
+		ul = 1
+		c := s.str[i]
+		if (c & (1 << 7)) != 0 {
+			for t := byte(1 << 6); (c & t) != 0; t >>= 1 {
+				ul++
+			}
+		}
+		if i + ul > s.len { // incomplete UTF-8 sequence
+			return l
+		}
+		l++
+		// recognize combining characters
+		if c == 0xcc || c == 0xcd {
+			r := (u16(c) << 8) | s.str[i+1]
+			if r >= 0xcc80 && r < 0xcdb0 { // diacritical marks
+				l--
+			}
+		} else if c == 0xe1 || c == 0xe2 || c == 0xef {
+			r := (u32(c) << 16) | (u32(s.str[i+1]) << 8) | s.str[i+2]
+			if r >= 0xe1aab0 && r < 0xe1ac80 // diacritical marks extended
+			|| r >= 0xe1b780 && r < 0xe1b880 // diacritical marks supplement
+			|| r >= 0xe28390 && r < 0xe28480 // diacritical marks for symbols
+			|| r >= 0xefb8a0 && r < 0xefb8b0 { // half marks
+				l--
+			}
+		}
+	}
+	return l
+}
+
 // Reads an utf8 character from standard input
 pub fn utf8_getchar() int {
 	c := C.getchar()
diff --git a/vlib/v/gen/str.v b/vlib/v/gen/str.v
index 9e7a5d662a..820e4377d7 100644
--- a/vlib/v/gen/str.v
+++ b/vlib/v/gen/str.v
@@ -57,9 +57,9 @@ string _STR(const char *fmt, int nfmts, ...) {
 				if (fmt[k-4] == '*') { // %*.*s
 					int fwidth = va_arg(argptr, int);
 					if (fwidth < 0)
-						fwidth -= (s.len - utf8_str_len(s));
+						fwidth -= (s.len - utf8_str_visible_length(s));
 					else
-						fwidth += (s.len - utf8_str_len(s));
+						fwidth += (s.len - utf8_str_visible_length(s));
 					_STR_PRINT_ARG(fmt, &buf, &nbytes, &memsize, k+fwidth-4, fwidth, s.len, s.str);
 				} else { // %.*s
 					_STR_PRINT_ARG(fmt, &buf, &nbytes, &memsize, k+s.len-4, s.len, s.str);
diff --git a/vlib/v/tests/string_interpolation_test.v b/vlib/v/tests/string_interpolation_test.v
index 4a14b35c64..e62fcefb77 100644
--- a/vlib/v/tests/string_interpolation_test.v
+++ b/vlib/v/tests/string_interpolation_test.v
@@ -102,6 +102,11 @@ fn test_utf8_string_interpolation() {
 	e := '\u20AC' // Eurosign
 	// TODO: this fails with MSVC and tcc
 	// assert '100.00 $e' == '100.00 €'
+	m2 := 'Москва́' // cyrillic а́: combination of U+0430 and U+0301, UTF-8: d0 b0 cc 81
+	d := 'Antonín Dvořák' // latin á: U+00E1, UTF-8: c3 a1
+	assert ':${m2:7}:${d:-15}:' == ': Москва́:Antonín Dvořák :'
+	g := 'Πελοπόννησος'
+	assert '>${g:-13}<' == '>Πελοπόννησος <'
 }
 
 struct S {