From bea9d7216879bc80a8513d3374a931cba14f6951 Mon Sep 17 00:00:00 2001
From: Neil <nyamatongwe@gmail.com>
Date: Sat, 17 Jul 2021 10:04:20 +1000
Subject: Fix minor inconsistency with word searching results at start and end
 of document which were considered word start / end positions even when first
 or last character did not fit requirements.

---
 doc/ScintillaHistory.html  |  3 +++
 src/Document.cxx           | 14 +++++++++-----
 test/unit/testDocument.cxx |  8 ++++++++
 3 files changed, 20 insertions(+), 5 deletions(-)

diff --git a/doc/ScintillaHistory.html b/doc/ScintillaHistory.html
index f433322ea..c6261693d 100644
--- a/doc/ScintillaHistory.html
+++ b/doc/ScintillaHistory.html
@@ -580,6 +580,9 @@
 	<a href="https://sourceforge.net/p/scintilla/feature-requests/1381/">Feature #1381</a>.
 	</li>
 	<li>
+	Word searching behaves more consistently at start and end of document.
+	</li>
+	<li>
 	Add SCI_ALLOCATELINES to allocate indices to hold some number of lines.
 	This can decrease reallocation overhead when the application can count or estimate the number of lines in huge files.
 	<a href="https://sourceforge.net/p/scintilla/feature-requests/1370/">Feature #1370</a>.
diff --git a/src/Document.cxx b/src/Document.cxx
index 3a3627fdc..b5ec9a275 100644
--- a/src/Document.cxx
+++ b/src/Document.cxx
@@ -1923,23 +1923,27 @@ constexpr bool IsWordEdge(CharacterClass cc, CharacterClass ccNext) noexcept {
 bool Document::IsWordStartAt(Sci::Position pos) const {
 	if (pos >= LengthNoExcept())
 		return false;
-	if (pos > 0) {
+	if (pos >= 0) {
 		const CharacterExtracted cePos = CharacterAfter(pos);
-		const CharacterExtracted cePrev = CharacterBefore(pos);
+		// At start of document, treat as if space before so can be word start
+		const CharacterExtracted cePrev = (pos > 0) ?
+			CharacterBefore(pos) : CharacterExtracted(' ', 1);
 		return IsWordEdge(WordCharacterClass(cePos.character), WordCharacterClass(cePrev.character));
 	}
 	return true;
 }
 
 /**
- * Check that the character at the given position is a word or punctuation character and that
+ * Check that the character before the given position is a word or punctuation character and that
  * the next character is of a different character class.
  */
 bool Document::IsWordEndAt(Sci::Position pos) const {
 	if (pos <= 0)
 		return false;
-	if (pos < LengthNoExcept()) {
-		const CharacterExtracted cePos = CharacterAfter(pos);
+	if (pos <= LengthNoExcept()) {
+		// At end of document, treat as if space after so can be word end
+		const CharacterExtracted cePos = (pos < LengthNoExcept()) ?
+			CharacterAfter(pos) : CharacterExtracted(' ', 1);
 		const CharacterExtracted cePrev = CharacterBefore(pos);
 		return IsWordEdge(WordCharacterClass(cePrev.character), WordCharacterClass(cePos.character));
 	}
diff --git a/test/unit/testDocument.cxx b/test/unit/testDocument.cxx
index 0f8e399fa..809647e19 100644
--- a/test/unit/testDocument.cxx
+++ b/test/unit/testDocument.cxx
@@ -477,4 +477,12 @@ TEST_CASE("Words") {
 		REQUIRE(docOverSpace.document.IsWordAt(1, 4));
 	}
 
+	SECTION("WordsAtEnds") {
+		const DocPlus doc("a c", 0);
+		REQUIRE(doc.document.IsWordAt(0, 1));
+		REQUIRE(doc.document.IsWordAt(2, 3));
+		const DocPlus docEndSpace(" a c ", 0);
+		REQUIRE(!docEndSpace.document.IsWordAt(0, 2));
+		REQUIRE(!docEndSpace.document.IsWordAt(3, 5));
+	}
 }
-- 
cgit v1.2.3