Leave a character the document font has in that font when substituting (#76)

src/Mpdf.php

@@ -25776,7 +25776,7 @@ function SubstituteCharsNonCore(&$writehtml_a, &$writehtml_i, &$writehtml_e)
 
 		foreach ($unicode as $c => $char) {
 
-			if (($char > 127 || ($flag == 1 && $char == 32)) && $char != 173 && (!isset($this->subArrMB['a'][$char]) || ($flag == 1 && $char == 32)) && ($char < 1536 || ($char > 1791 && $char < 2304) || $char > 3455)) {
+			if ($char > 127 && $char != 173 && !isset($this->subArrMB['a'][$char]) && ($char < 1536 || ($char > 1791 && $char < 2304) || $char > 3455)) {
 				if ($flag == 0) {
 					$start = $c;
 				}
@@ -25803,7 +25803,8 @@ function SubstituteCharsNonCore(&$writehtml_a, &$writehtml_i, &$writehtml_e)
 
 		foreach ($this->backupSubsFont as $bsfctr => $bsf) {
 
-			if ($this->fonttrans[$bsf] == 'chelvetica' || $this->fonttrans[$bsf] == 'ctimes' || $this->fonttrans[$bsf] == 'ccourier') {
+			// A backup font that is a core font's name draws nothing a core font does not
+			if (isset($this->fonttrans[$bsf]) && in_array($this->fonttrans[$bsf], ['chelvetica', 'ctimes', 'ccourier'], true)) {
 				continue;
 			}
 
@@ -25880,25 +25881,30 @@ function SubstituteCharsMB(&$writehtml_a, &$writehtml_i, &$writehtml_e)
 		$ftype = '';
 		$u = [];
 		// Every emoji is outside ASCII, a keycap's U+20E3 included
-		$emojiClusters = preg_match('/[^\x00-\x7F]/', $writehtml_e) ? Emoji::clusters($unicode) : [];
+		$mayHaveEmoji = preg_match('/[^\x00-\x7F]/', $writehtml_e) === 1;
+		// Each emoji up to where the run ends, by where it starts: found as the scan reaches it, since the
+		// rest of the text is scanned again on the next pass
+		$emojiClusters = [];
 		$codepointCount = count($unicode);
 
 		// Not foreach, so the loop can jump past the rest of a multi-codepoint sequence once it has been handled
 		for ($c = 0; $c < $codepointCount; $c++) {
 			$char = $unicode[$c];
+			$cluster = $mayHaveEmoji ? Emoji::clusterAt($unicode, $c, $codepointCount) : null;
 
 			// An emoji moves whole or not at all, so its joiners and selectors go with it to the font
 			// whose GSUB joins them
-			if (isset($emojiClusters[$c])) {
+			if ($cluster !== null) {
 				// An emoji ends a SIP run; it is handled on the next pass
 				if ($flag == 2) {
 					break;
 				}
 
-				$length = $emojiClusters[$c][0];
+				$emojiClusters[$c] = $cluster;
+				$length = $cluster[0];
 				$emoji = array_slice($unicode, $c, $length);
 
-				if ($this->fontSubstitution->emojiWantsAnotherFont($emoji, $emojiClusters[$c][1])) {
+				if ($this->fontSubstitution->emojiWantsAnotherFont($emoji, $cluster[1])) {
 					if ($flag == 0) {
 						$start = $c;
 					}
@@ -25912,7 +25918,7 @@ function SubstituteCharsMB(&$writehtml_a, &$writehtml_i, &$writehtml_e)
 				continue;
 			}
 
-			if (($flag == 0 || $flag == 2) && (!$this->_charDefined($cw, $char) || ($flag == 2 && $char == 32)) && $this->checkSIP && $char > 131071) {  // Unicode Plane 2 (SIP)
+			if (($flag == 0 || $flag == 2) && !$this->_charDefined($cw, $char) && $this->checkSIP && $char > 131071) {  // Unicode Plane 2 (SIP)
 
 				if (in_array($this->FontFamily, $this->available_CJK_fonts)) {
 					return 0;
@@ -25925,9 +25931,7 @@ function SubstituteCharsMB(&$writehtml_a, &$writehtml_i, &$writehtml_e)
 				$flag = 2;
 				$u[] = $char;
 
-				// elseif (($flag == 0 || $flag==1) && $char != 173 && !$this->_charDefined($cw,$char) && ($char<1423 ||  ($char>3583 && $char < 11263))) {
-
-			} elseif (($flag == 0 || $flag == 1) && $char != 173 && (!$this->_charDefined($cw, $char) || ($flag == 1 && $char == 32)) && ($char < 1536 || ($char > 1791 && $char < 2304) || $char > 3455)) {
+			} elseif (($flag == 0 || $flag == 1) && $char != 173 && !$this->_charDefined($cw, $char) && ($char < 1536 || ($char > 1791 && $char < 2304) || $char > 3455)) {
 
 				if ($flag == 0) {
 					$start = $c;

src/Unicode/Emoji.php

@@ -253,13 +253,16 @@ public static function isTag($codepoint)
 	}
 
 	/**
+	 * The emoji at one place in a run, for a caller walking the run itself that would otherwise find
+	 * every emoji in it with clusters() before it knows how far it will look
+	 *
 	 * @param int[] $codepoints The run
-	 * @param int   $i          Where in it to look
+	 * @param int   $i          Where in it to look, which is not inside an emoji that starts earlier
 	 * @param int   $count      How long it is
 	 *
 	 * @return array|null [length, presentation] of the emoji starting at $i, or null where none does
 	 */
-	private static function clusterAt(array $codepoints, $i, $count)
+	public static function clusterAt(array $codepoints, $i, $count)
 	{
 		$first = $codepoints[$i];
 

tests/Mpdf/EmojiSubstitutionTest.php

@@ -89,10 +89,37 @@ public function testAKeycapGoesToTheFontThatHasAllOfIt()
 	 */
 	public function testAFlagAndASkinToneStayWithTheEmojiTheyQualify()
 	{
-		$pieces = $this->drawn([0x1F1E6, 0x1F1FA, 0x20, 0x1F44D, 0x1F3FD], ['coloremoji']);
+		$pieces = $this->drawn([0x1F1E6, 0x1F1FA, 0x1F44D, 0x1F3FD], ['coloremoji']);
 
+		$this->assertCount(1, $pieces);
+		$this->assertSame('coloremoji', $pieces[0][0]);
+		$this->assertCount(2, $pieces[0][1], 'the flag and the thumb');
+	}
+
+	/**
+	 * A space between two emoji is the document font's, which has one, and not carried into the run
+	 * of the font the emoji go to: an emoji font's space can be as wide as an emoji, as Noto Color
+	 * Emoji's is. The recorder leaves out a piece that is only a space, so the space shows as the two
+	 * runs either side of it.
+	 */
+	public function testASpaceBetweenEmojiStaysInTheDocumentFont()
+	{
+		$pieces = $this->drawn([0x1F1E6, 0x1F1FA, 0x20, 0x1F44D], ['coloremoji']);
+
+		$this->assertCount(2, $pieces, 'a run either side of the space');
 		$this->assertSame('coloremoji', $pieces[0][0]);
-		$this->assertCount(3, $pieces[0][1], 'the flag, the space the run carries, and the thumb');
+		$this->assertCount(1, $pieces[0][1], 'the flag, without the space');
+		$this->assertSame(['coloremoji', [0x1F44D]], $pieces[1]);
+	}
+
+	/**
+	 * A space in a core font's text is the core font's too, and ends the run a backup font draws
+	 */
+	public function testASpaceBetweenWordsACoreFontLacksStaysInTheCoreFont()
+	{
+		$pieces = $this->drawn([0x61, 0x416, 0x416, 0x20, 0x416, 0x62], ['dejavusans'], ['default_font' => 'chelvetica']);
+
+		$this->assertSame([['chelvetica', [0x61]], ['dejavusans', [0x416, 0x416]], ['dejavusans', [0x416]], ['chelvetica', [0x62]]], $pieces);
 	}
 
 	/**