Use something more like white-space: pre-wrap;

BlockState.dot

@@ -7,14 +7,12 @@ digraph blockstate  {
 	start;
 	emptyBlock [label = "empty block"];
 	nonEmptyBlock [label = "non-empty block"];
-	emittedSpace [label = "emitted space"];
 	lineBreakTag [label = "line break tag"];
 	atLeastTwoLineBreakTags [label = ">=2 line break tags"];
 	emptyBlockWithAtLeastTwoPreviousLineBreakTags [label = "empty block w/ >=2 prev line break tags"];
 	beginListItem [label = "begin list item"];
 	endListItem [label = "end list item"];
 	listItemContent [label = "list item content"];
-	emittedSpaceInListItemContent [label = "emitted space in text in list item content"];
 	lineBreakTagInListItemContent [label = "line break tag in list item content"];
 	atLeastTwoLineBreakTagsInListItemContent [label = ">= 2 line break tags in list item content"];
 	preformattedStart [label = "preformatted start"];
@@ -32,27 +30,21 @@ digraph blockstate  {
 	start -> nonEmptyBlock [label = "non-whitespace"];
 	start -> preformattedStart [label = "<pre> (depth = 1)"];
 	start -> beginListItem [label = "<li>"];
-	nonEmptyBlock -> nonEmptyBlock [label = "non-whitespace"];
+	nonEmptyBlock -> nonEmptyBlock [label = "non-newline"];
 	nonEmptyBlock -> emptyBlock [label = "start/end block"];
-	nonEmptyBlock -> emittedSpace [label = "whitespace (emit space)"];
+	nonEmptyBlock -> lineBreakTag [label = "<br> or \\n (append to tmp)"];
-	nonEmptyBlock -> lineBreakTag [label = "<br> (append to tmp)"];
 	nonEmptyBlock -> beginListItem [label = "<li>"];
 	nonEmptyBlock -> endListItem [label = "</li>"];
-	emittedSpace -> nonEmptyBlock [label = "non-whitespace"];
-	emittedSpace -> emittedSpace [label = "whitespace (skip)"];
-	emittedSpace -> emptyBlock [label = "start/end block (remove 1)"];
-	emittedSpace -> lineBreakTag [label = "<br> (append to tmp)"];
-	emittedSpace -> end [label = "EOF (remove 1)"];
 	emptyBlock -> nonEmptyBlock [label = "non-whitespace (block break)"];
 	emptyBlock -> emptyBlock [label = "whitespace (skip)\n<br>\n</pre>\nstart/end block"];
 	emptyBlock -> afterPreStartTag [label = "<pre> (depth = 1)"];
 	emptyBlock -> beginListItem [label = "<li>"];
 	emptyBlock -> endListItem [label = "</li>"];
-	lineBreakTag -> lineBreakTag [label = "whitespace (skip)"];
+	lineBreakTag -> lineBreakTag [label = "whitespace (append to tmp)"];
-	lineBreakTag -> atLeastTwoLineBreakTags [label = "<br> (append to tmp)"];
+	lineBreakTag -> atLeastTwoLineBreakTags [label = "<br> or \\n (append to tmp)"];
 	lineBreakTag -> emptyBlock [label = "start/end block (clear tmp)"];
 	lineBreakTag -> nonEmptyBlock [label = "non-whitespace (emit tmp)"];
-	atLeastTwoLineBreakTags -> atLeastTwoLineBreakTags [label = "whitespace (skip)\n<br> (append to tmp)"];
+	atLeastTwoLineBreakTags -> atLeastTwoLineBreakTags [label = "whitespace or <br> (append to tmp)"];
 	atLeastTwoLineBreakTags -> nonEmptyBlock [label = "non-whitespace (emit tmp)"];
 	atLeastTwoLineBreakTags -> emptyBlockWithAtLeastTwoPreviousLineBreakTags [label = "start/end block"];
 	emptyBlockWithAtLeastTwoPreviousLineBreakTags -> emptyBlockWithAtLeastTwoPreviousLineBreakTags [label = "whitespace (skip)\n<br>\n</pre>\nstart/end block"];
@@ -70,23 +62,15 @@ digraph blockstate  {
 	listItemContent -> listItemContent [label = "non-whitespace"];
 	listItemContent -> beginListItem [label = "<li> (line break)"];
 	listItemContent -> lineBreakTagInListItemContent [label = "<br> (append to tmp)"];
-	listItemContent -> emittedSpaceInListItemContent [label = "whitespace (emit space)"];
 	listItemContent -> emptyBlock [label = "start/end block"];
 	listItemContent -> endListItem [label = "</li>"];
-	emittedSpaceInListItemContent -> emittedSpaceInListItemContent [label = "whitespace (skip)"];
+	lineBreakTagInListItemContent -> lineBreakTagInListItemContent [label = "whitespace (append to tmp)"];
-	emittedSpaceInListItemContent -> listItemContent [label = "non-whitespace"];
-	emittedSpaceInListItemContent -> end [label = "EOF (remove 1)"];
-	emittedSpaceInListItemContent -> emptyBlock [label = "start/end block (remove 1)"];
-	emittedSpaceInListItemContent -> beginListItem [label = "<li> (remove 1, line break)"];
-	emittedSpaceInListItemContent -> lineBreakTagInListItemContent [label = "<br> (append to tmp)"];
-	emittedSpaceInListItemContent -> endListItem [label = "</li> (remove 1)"];
-	lineBreakTagInListItemContent -> lineBreakTagInListItemContent [label = "whitespace (skip)"];
 	lineBreakTagInListItemContent -> emptyBlock [label = "start/end block (clear tmp)"];
 	lineBreakTagInListItemContent -> beginListItem [label = "<li> (emit tmp, line break)"];
 	lineBreakTagInListItemContent -> listItemContent [label = "non-whitespace (emit tmp)"];
-	lineBreakTagInListItemContent -> atLeastTwoLineBreakTagsInListItemContent [label = "<br> (append to tmp)"];
+	lineBreakTagInListItemContent -> atLeastTwoLineBreakTagsInListItemContent [label = "<br> or \\n (append to tmp)"];
 	lineBreakTagInListItemContent -> endListItem [label = "</li> (clear tmp)"];
-	atLeastTwoLineBreakTagsInListItemContent -> atLeastTwoLineBreakTagsInListItemContent [label = "<br> (append to tmp)\nwhitespace (skip)"];
+	atLeastTwoLineBreakTagsInListItemContent -> atLeastTwoLineBreakTagsInListItemContent [label = "whitespace or <br> (append to tmp)"];
 	atLeastTwoLineBreakTagsInListItemContent -> beginListItem [label = "<li> (emit tmp, line break)"];
 	atLeastTwoLineBreakTagsInListItemContent -> emptyBlockWithAtLeastTwoPreviousLineBreakTags [label = "start/end block"];
 	atLeastTwoLineBreakTagsInListItemContent -> listItemContent [label = "non-whitespace (emit tmp)"];

Sources/HTMLStreamer/BlockState.swift

@@ -38,9 +38,6 @@ extension BlockStateMachine {
             break
         case .nonEmptyBlock:
             blockState = .emptyBlock
-        case .emittedSpace:
-            blockState = .emptyBlock
-            removeChar()
         case .lineBreakTag:
             blockState = .emptyBlock
             temporaryBuffer = ""
@@ -54,9 +51,6 @@ extension BlockStateMachine {
             blockState = .emptyBlock
         case .listItemContent:
             blockState = .emptyBlock
-        case .emittedSpaceInListItemContent:
-            blockState = .emptyBlock
-            removeChar()
         case .lineBreakTagInListItemContent:
             blockState = .emptyBlock
             temporaryBuffer = ""
@@ -104,22 +98,19 @@ extension BlockStateMachine {
                 return true
             }
         case .nonEmptyBlock:
-            if isWhitespace {
+            if isNewline {
-                blockState = .emittedSpace
+                blockState = .lineBreakTag
-                append(" ")
+                temporaryBuffer.append("\n")
                 return false
             } else {
                 return true
             }
-        case .emittedSpace:
-            if isWhitespace {
-                return false
-            } else {
-                blockState = .nonEmptyBlock
-                return true
-            }
         case .lineBreakTag:
             if isWhitespace {
+                if isNewline {
+                    blockState = .atLeastTwoLineBreakTags
+                }
+                temporaryBuffer.unicodeScalars.append(char)
                 return false
             } else {
                 blockState = .nonEmptyBlock
@@ -129,6 +120,7 @@ extension BlockStateMachine {
             }
         case .atLeastTwoLineBreakTags:
             if isWhitespace {
+                temporaryBuffer.unicodeScalars.append(char)
                 return false
             } else {
                 blockState = .nonEmptyBlock
@@ -162,22 +154,19 @@ extension BlockStateMachine {
                 return true
             }
         case .listItemContent:
-            if isWhitespace {
+            if isNewline {
-                blockState = .emittedSpaceInListItemContent
+                blockState = .lineBreakTagInListItemContent
-                append(" ")
+                temporaryBuffer.append("\n")
                 return false
             } else {
                 return true
             }
-        case .emittedSpaceInListItemContent:
-            if isWhitespace {
-                return false
-            } else {
-                blockState = .listItemContent
-                return true
-            }
         case .lineBreakTagInListItemContent:
             if isWhitespace {
+                if isNewline {
+                    blockState = .atLeastTwoLineBreakTagsInListItemContent
+                }
+                temporaryBuffer.unicodeScalars.append(char)
                 return false
             } else {
                 blockState = .listItemContent
@@ -187,6 +176,7 @@ extension BlockStateMachine {
             }
         case .atLeastTwoLineBreakTagsInListItemContent:
             if isWhitespace {
+                temporaryBuffer.unicodeScalars.append(char)
                 return false
             } else {
                 blockState = .listItemContent
@@ -311,9 +301,6 @@ extension BlockStateMachine {
         case .nonEmptyBlock:
             blockState = .lineBreakTag
             temporaryBuffer.append(lineBreak)
-        case .emittedSpace:
-            blockState = .lineBreakTag
-            temporaryBuffer.append(lineBreak)
         case .lineBreakTag:
             blockState = .atLeastTwoLineBreakTags
             temporaryBuffer.append(lineBreak)
@@ -329,9 +316,6 @@ extension BlockStateMachine {
         case .listItemContent:
             blockState = .lineBreakTagInListItemContent
             temporaryBuffer.append(lineBreak)
-        case .emittedSpaceInListItemContent:
-            blockState = .lineBreakTagInListItemContent
-            temporaryBuffer.append(lineBreak)
         case .lineBreakTagInListItemContent:
             blockState = .atLeastTwoLineBreakTagsInListItemContent
             temporaryBuffer.append(lineBreak)
@@ -374,8 +358,6 @@ extension BlockStateMachine {
             blockState = .afterPreStartTag(depth: 1)
         case .nonEmptyBlock:
             fatalError("unreachable")
-        case .emittedSpace:
-            fatalError("unreachable")
         case .lineBreakTag:
             fatalError("unreachable")
         case .atLeastTwoLineBreakTags:
@@ -388,8 +370,6 @@ extension BlockStateMachine {
             fatalError("unreachable")
         case .listItemContent:
             fatalError("unreachable")
-        case .emittedSpaceInListItemContent:
-            fatalError("unreachable")
         case .lineBreakTagInListItemContent:
             fatalError("unreachable")
         case .atLeastTwoLineBreakTagsInListItemContent:
@@ -423,8 +403,6 @@ extension BlockStateMachine {
             break
         case .nonEmptyBlock:
             fatalError("unreachable")
-        case .emittedSpace:
-            fatalError("unreachable")
         case .lineBreakTag:
             fatalError("unreachable")
         case .atLeastTwoLineBreakTags:
@@ -437,8 +415,6 @@ extension BlockStateMachine {
             fatalError("unreachable")
         case .listItemContent:
             fatalError("unreachable")
-        case .emittedSpaceInListItemContent:
-            fatalError("unreachable")
         case .lineBreakTagInListItemContent:
             fatalError("unreachable")
         case .atLeastTwoLineBreakTagsInListItemContent:
@@ -501,10 +477,6 @@ extension BlockStateMachine {
         case .listItemContent:
             blockState = .beginListItem
             append(lineBreak)
-        case .emittedSpaceInListItemContent:
-            blockState = .beginListItem
-            removeChar()
-            append(lineBreak)
         case .lineBreakTagInListItemContent:
             blockState = .beginListItem
             append(temporaryBuffer)
@@ -528,9 +500,6 @@ extension BlockStateMachine {
             blockState = .endListItem
         case .listItemContent:
             blockState = .endListItem
-        case .emittedSpaceInListItemContent:
-            blockState = .endListItem
-            removeChar()
         case .lineBreakTagInListItemContent:
             blockState = .endListItem
             temporaryBuffer = ""
@@ -544,10 +513,6 @@ extension BlockStateMachine {
     
     mutating func endBlocks() {
         switch blockState {
-        case .emittedSpace:
-            removeChar()
-        case .emittedSpaceInListItemContent:
-            removeChar()
         default:
             break
         }
@@ -558,14 +523,12 @@ enum BlockState: Equatable {
     case start
     case emptyBlock
     case nonEmptyBlock
-    case emittedSpace
     case lineBreakTag
     case atLeastTwoLineBreakTags
     case emptyBlockWithAtLeastTwoPreviousLineBreakTags
     case beginListItem
     case endListItem
     case listItemContent
-    case emittedSpaceInListItemContent
     case lineBreakTagInListItemContent
     case atLeastTwoLineBreakTagsInListItemContent
     case preformattedStart(depth: Int32)
@@ -583,5 +546,5 @@ enum BlockState: Equatable {
 private func isWhitespace(_ c: UnicodeScalar) -> Bool {
     // this is not strictly correct, but checking the actual unicode properties is slow
     // and this should cover the vast majority of actual use
-    c == " " || c == "\n" || c == "\t" || c == "\u{A0}" /* NO-BREAK SPACE */
+    c == " " || c == "\n" || c == "\t"
 }

Tests/HTMLStreamerTests/AttributedStringConverterTests.swift

@@ -50,7 +50,7 @@ final class AttributedStringConverterTests: XCTestCase {
             color: color,
             paragraphStyle: .default
         )
-        var converter = AttributedStringConverter<Callbacks>(configuration: config)
+        let converter = AttributedStringConverter<Callbacks>(configuration: config)
         return converter.convert(html: html)
     }
 
@@ -391,7 +391,7 @@ final class AttributedStringConverterTests: XCTestCase {
         XCTAssertEqual(convert("<p>a\n\n</p>\n"), result)
         XCTAssertEqual(convert("<p>\n\na</p>"), result)
         XCTAssertEqual(convert("<p>\n\na</p>\n"), result)
-        let result2 = NSAttributedString(string: "a b", attributes: [
+        let result2 = NSAttributedString(string: "a\n\n\nb", attributes: [
             .font: font,
             .paragraphStyle: NSParagraphStyle.default,
             .foregroundColor: color,
@@ -521,6 +521,12 @@ final class AttributedStringConverterTests: XCTestCase {
         XCTAssertEqual(convert("  \n\t<p>a</p>"), result)
         XCTAssertEqual(convert("a\n\t"), result)
         XCTAssertEqual(convert("<p>a</p>  \n\t"), result)
+        let result2 = NSAttributedString(string: "a  ", attributes: [
+            .font: font,
+            .paragraphStyle: NSParagraphStyle.default,
+            .foregroundColor: color,
+        ])
+        XCTAssertEqual(convert("a  \n\t"), result2)
         
         let pre = NSAttributedString(string: "a", attributes: [
             .font: monospaceFont,
@@ -531,8 +537,8 @@ final class AttributedStringConverterTests: XCTestCase {
         XCTAssertEqual(convert("<pre>a</pre>  \n\t"), pre)
     }
     
-    func testWhitespaceCollapsing() {
+    func testDoesNotCollapseWhitespace() {
-        let result = NSAttributedString(string: "a b", attributes: [
+        let result = NSAttributedString(string: "a \t\nb", attributes: [
             .font: font,
             .paragraphStyle: NSParagraphStyle.default,
             .foregroundColor: color,
@@ -565,7 +571,12 @@ final class AttributedStringConverterTests: XCTestCase {
             .foregroundColor: color,
         ])
         XCTAssertEqual(convert("<ol><li>a</li>c<li>b</li></ol>"), result)
-        XCTAssertEqual(convert("<ol><li>a</li>c  <li>b</li></ol>"), result)
+        let result2 = NSAttributedString(string: "\t1.\ta\n\t\tc  \n\t2.\tb", attributes: [
+            .font: font,
+            .paragraphStyle: listParagraphStyle,
+            .foregroundColor: color,
+        ])
+        XCTAssertEqual(convert("<ol><li>a</li>c  <li>b</li></ol>"), result2)
     }
     
     func testWhitespaceCollapsingInTextBetweenListItems() {