Give SVG Form XObjects an explicit /Resources dictionary (mirrors mpdf/mpdf#2196)

src/Writer/FormWriter.php

@@ -28,6 +28,12 @@ public function __construct(Mpdf $mpdf, BaseWriter $writer)
 
 	public function writeFormObjects() // _putformobjects
 	{
+		if (!$this->mpdf->formobjects) {
+			return;
+		}
+
+		$resources = $this->getResources();
+
 		foreach ($this->mpdf->formobjects as $file => $info) {
 
 			$this->writer->object();
@@ -38,6 +44,7 @@ public function writeFormObjects() // _putformobjects
 			$this->writer->write('/Subtype /Form');
 			$this->writer->write('/Group ' . ($this->mpdf->n + 1) . ' 0 R');
 			$this->writer->write('/BBox [' . $info['x'] . ' ' . $info['y'] . ' ' . ($info['w'] + $info['x']) . ' ' . ($info['h'] + $info['y']) . ']');
+			$this->writer->write('/Resources <<' . $resources . '>>');
 
 			if ($this->mpdf->compress) {
 				$this->writer->write('/Filter /FlateDecode');
@@ -59,4 +66,82 @@ public function writeFormObjects() // _putformobjects
 			$this->writer->write('endobj');
 		}
 	}
+
+	/**
+	 * A Form XObject with no /Resources of its own inherits the page's, which PDF 1.4 allowed
+	 * but PDF/A does not. The entries an SVG can reference are already flagged as it is drawn.
+	 *
+	 * As in _putpatterns, this hands every Form XObject the resources of every Form XObject.
+	 *
+	 * @return string
+	 */
+	private function getResources()
+	{
+		$resources = ['/ProcSet [/PDF /Text /ImageB /ImageC /ImageI]'];
+
+		$extgstates = [];
+		foreach ($this->mpdf->extgstates as $k => $extgstate) {
+			if (!isset($extgstate['fo']) || !$extgstate['fo']) {
+				continue;
+			}
+
+			if (isset($extgstate['trans'])) {
+				$extgstates[] = '/' . $extgstate['trans'] . ' ' . $extgstate['n'] . ' 0 R';
+			} else {
+				$extgstates[] = '/GS' . $k . ' ' . $extgstate['n'] . ' 0 R';
+			}
+		}
+
+		if (count($extgstates)) {
+			$resources[] = '/ExtGState <<' . implode(' ', $extgstates) . '>>';
+		}
+
+		/* -- BACKGROUNDS -- */
+		$shadings = [];
+		if (isset($this->mpdf->gradients) && count($this->mpdf->gradients) > 0) {
+			foreach ($this->mpdf->gradients as $id => $grad) {
+				// 'id' is the shading's object number, filled in by _putshaders; a gradient it
+				// does not write (a soft mask, say) is not nameable here
+				if (!isset($grad['fo']) || !$grad['fo'] || !isset($grad['id'])) {
+					continue;
+				}
+
+				$shadings[] = '/Sh' . $id . ' ' . $grad['id'] . ' 0 R';
+			}
+		}
+
+		if (count($shadings)) {
+			$resources[] = '/Shading <<' . implode(' ', $shadings) . '>>';
+		}
+		/* -- END BACKGROUNDS -- */
+
+		$fonts = [];
+		foreach ($this->mpdf->fonts as $font) {
+			if (!isset($font['fo']) || !$font['fo']) {
+				continue;
+			}
+
+			if (isset($font['type']) && $font['type'] === 'TTF') {
+				if (!$font['used']) {
+					continue;
+				}
+
+				if ($font['sip'] || $font['smp']) {
+					foreach ($font['n'] as $k => $fid) {
+						$fonts[] = '/F' . $font['subsetfontids'][$k] . ' ' . $fid . ' 0 R';
+					}
+
+					continue;
+				}
+			}
+
+			$fonts[] = '/F' . $font['i'] . ' ' . $font['n'] . ' 0 R';
+		}
+
+		if (count($fonts)) {
+			$resources[] = '/Font <<' . implode(' ', $fonts) . '>>';
+		}
+
+		return implode(' ', $resources);
+	}
 }

src/Writer/ResourceWriter.php

@@ -113,11 +113,14 @@ public function writeResources() // _putresources
 
 		$this->imageWriter->writeImages();
 
+		// Shaders are written first so that a Form XObject can name the gradients it paints
+		// with; _putpatterns still needs them resolved before it runs, as it always has
+		$this->backgroundWriter->writeShaders();
+
 		$this->formWriter->writeFormObjects();
 
 		$this->mpdf->writeImportedPagesAndResolvedObjects();
 
-		$this->backgroundWriter->writeShaders();
 		$this->backgroundWriter->writePatterns();
 
 		// Resource dictionary

tests/Mpdf/FormObjectResourcesTest.php

@@ -0,0 +1,89 @@
+<?php
+
+namespace Mpdf;
+
+class FormObjectResourcesTest extends \Yoast\PHPUnitPolyfills\TestCases\TestCase
+{
+
+	const SVG = __DIR__ . '/../data/img/form-xobject-resources.svg';
+
+	/**
+	 * The fixture paints with a gradient, a half-transparent rectangle and a line of text, so
+	 * the Form XObject names an entry from all three resource categories.
+	 */
+	public function testEveryNameTheStreamUsesIsDeclared()
+	{
+		$objects = $this->renderFormObjects(false);
+		$this->assertCount(1, $objects);
+
+		list($header, $stream) = $objects[0];
+
+		$names = $this->namesUsedBy($stream);
+		$this->assertGreaterThanOrEqual(3, count($names), 'Expected the fixture to use several resources');
+
+		foreach ($names as $name) {
+			$this->assertMatchesRegularExpression(
+				'#/' . preg_quote($name, '#') . ' [1-9]\d* 0 R#',
+				$header,
+				sprintf('/%s is painted with but not declared in the Form XObject /Resources', $name)
+			);
+		}
+	}
+
+	public function testResourcesAreWrittenForPdfA()
+	{
+		$objects = $this->renderFormObjects(true);
+		$this->assertCount(1, $objects);
+
+		list($header) = $objects[0];
+
+		$this->assertStringContainsString('/Resources <<', $header);
+	}
+
+	/**
+	 * @param  string $stream
+	 * @return string[]
+	 */
+	private function namesUsedBy($stream)
+	{
+		preg_match_all('#/([A-Za-z0-9]+)\s+(?:gs|sh)[^A-Za-z0-9]#', $stream, $states);
+		preg_match_all('#/([A-Za-z0-9]+)\s+[\d.]+\s+Tf#', $stream, $fonts);
+
+		return array_values(array_unique(array_merge($states[1], $fonts[1])));
+	}
+
+	/**
+	 * @param  bool $pdfa
+	 * @return array[] One [header, stream] pair per Form XObject mPDF wrote for the image
+	 */
+	private function renderFormObjects($pdfa)
+	{
+		$mpdf = new Mpdf();
+		$mpdf->compress = false;
+
+		if ($pdfa) {
+			$mpdf->PDFA = true;
+			$mpdf->PDFAauto = true;
+			$mpdf->PDFAversion = '3-B';
+		}
+
+		$mpdf->WriteHTML('<img src="' . self::SVG . '" style="width:40mm">');
+
+		$output = $mpdf->OutputBinaryData();
+		$mpdf->cleanup();
+
+		preg_match_all('/\d+ 0 obj\s*(.*?)\bstream\r?\n(.*?)\r?\nendstream/s', $output, $matches, PREG_SET_ORDER);
+
+		$objects = [];
+		foreach ($matches as $match) {
+			// The transparency group is written as a separate object, which is what tells a
+			// Form XObject mPDF wrote here apart from the inline groups a soft mask uses
+			if (strpos($match[1], '/Subtype /Form') !== false && preg_match('#/Group \d+ 0 R#', $match[1])) {
+				$objects[] = [$match[1], $match[2]];
+			}
+		}
+
+		return $objects;
+	}
+
+}