ref #79: Improve parsing of inline images; Handle the scenario where …

…"EI" can be found inside the image data, right at the end of a line, and more binary data comes afterwards;

Uses the hasNoFollowingBinData() method used by PDFBox

src/main/java/org/sejda/sambox/input/ContentStreamParser.java

@@ -26,8 +26,11 @@
 import org.sejda.sambox.cos.COSBase;
 import org.sejda.sambox.cos.COSDictionary;
 import org.sejda.sambox.cos.COSName;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
 
 import java.io.IOException;
+import java.nio.ByteBuffer;
 import java.util.ArrayList;
 import java.util.Collections;
 import java.util.List;
@@ -50,6 +53,11 @@ public class ContentStreamParser extends SourceReader
     private final ContentStreamCOSParser cosParser;
     private final List<Object> tokens = new ArrayList<>();
 
+    private static final Logger LOG = LoggerFactory.getLogger(ContentStreamParser.class);
+
+    private static final int MAX_BIN_CHAR_TEST_LENGTH = 10;
+    private final byte[] binCharTestArr = new byte[MAX_BIN_CHAR_TEST_LENGTH];
+
     public ContentStreamParser(PDContentStream stream) throws IOException
     {
         this(SeekableSources.inMemorySeekableSourceFrom(stream.getContents()));
@@ -165,7 +173,7 @@ private boolean isEndOfImageFrom(long position) throws IOException
         {
             current = source().read();
             // if not a EI we restore the position and go on
-            if (current == 'I' && (isEndOfImage() || isEOF(source().peek())))
+            if (current == 'I' && (isEndOfImage() || isEOF(source().peek())) && hasNoFollowingBinData())
             {
                 return true;
             }
@@ -219,6 +227,87 @@ private boolean isEndOfImage() throws IOException
         }
     }
 
+    /**
+     * Looks up an amount of bytes if they contain only ASCII characters (no
+     * control sequences etc.), and that these ASCII characters begin with a
+     * sequence of 1-3 non-blank characters between blanks
+     *
+     * @return <code>true</code> if next bytes are probably printable ASCII
+     * characters starting with a PDF operator, otherwise <code>false</code>
+     */
+    private boolean hasNoFollowingBinData() throws IOException
+    {
+        long originalPosition = source().position();
+
+        try {
+            // as suggested in PDFBOX-1164
+            final int readBytes = source().read(ByteBuffer.wrap(binCharTestArr));
+            boolean noBinData = true;
+            int startOpIdx = -1;
+            int endOpIdx = -1;
+
+            if (readBytes > 0) 
+            {
+                for (int bIdx = 0; bIdx < readBytes; bIdx++) 
+                {
+                    final byte b = binCharTestArr[bIdx];
+                    if (b != 0 && b < 0x09 || b > 0x0a && b < 0x20 && b != 0x0d) 
+                    {
+                        // control character or > 0x7f -> we have binary data
+                        noBinData = false;
+                        break;
+                    }
+                    // find the start of a PDF operator
+                    if (startOpIdx == -1 && !(b == 0 || b == 9 || b == 0x20 || b == 0x0a || b == 0x0d)) 
+                    {
+                        startOpIdx = bIdx;
+                    } else if (startOpIdx != -1 && endOpIdx == -1 &&
+                            (b == 0 || b == 9 || b == 0x20 || b == 0x0a || b == 0x0d)) 
+                    {
+                        endOpIdx = bIdx;
+                    }
+                }
+
+                // PDFBOX-3742: just assuming that 1-3 non blanks is a PDF operator isn't enough
+                if (endOpIdx != -1 && startOpIdx != -1) 
+                {
+                    // usually, the operator here is Q, sometimes EMC (PDFBOX-2376), S (PDFBOX-3784).
+                    String s = new String(binCharTestArr, startOpIdx, endOpIdx - startOpIdx);
+                    if (!"q".equals(s) && !"Q".equals(s) && !"EMC".equals(s) && !"S".equals(s)) 
+                    {
+                        noBinData = false;
+                    }
+                }
+
+                // only if not close to eof
+                if (readBytes == MAX_BIN_CHAR_TEST_LENGTH) 
+                {
+                    // a PDF operator is 1-3 bytes long
+                    if (startOpIdx != -1 && endOpIdx == -1) 
+                    {
+                        endOpIdx = MAX_BIN_CHAR_TEST_LENGTH;
+                    }
+                    if (endOpIdx != -1 && startOpIdx != -1 && endOpIdx - startOpIdx > 3) 
+                    {
+                        noBinData = false;
+                    }
+                }
+            }
+
+            if (!noBinData) 
+            {
+                LOG.warn("ignoring 'EI' assumed to be in the middle of inline image at stream offset " + originalPosition);
+            }
+
+            return noBinData;
+
+        } 
+        finally 
+        {
+            source().position(originalPosition);
+        }
+    }
+
     @Override
     public void close() throws IOException
     {

src/test/java/org/sejda/sambox/input/ContentStreamParserTest.java

@@ -147,4 +147,15 @@ public void nextInlineImageEISpace() throws IOException
         operator = (Operator) victim.nextParsedToken();
         assertEquals("q", operator.getName());
     }
+
+    @Test
+    public void nextInlineImageEIInStreamWithoutSpace() throws IOException {
+        victim = new ContentStreamParser(inMemorySeekableSourceFrom(
+                getClass().getResourceAsStream("/sambox/inline_image_EI_inside_stream.txt")));
+        Operator operator = (Operator) victim.nextParsedToken();
+        assertEquals("BI", operator.getName());
+        assertEquals(768, operator.getImageData().length);
+        operator = (Operator) victim.nextParsedToken();
+        assertEquals("Q", operator.getName());
+    }
 }

src/test/resources/sambox/inline_image_EI_inside_stream.txt

@@ -0,0 +1,7 @@
+BI /W 1088 /H 1071 /BPC 1 /IM true /F [/A85 /Fl] ID
+ Gb"/l4d,,%%YVZ[+g[6&U\)s4UZj0f6CT_kKGBW;W9W#p=OsRY:=9umQMC(!+::Xb3"a3>7pLQ'47i38q=sk&?dFQ&,nh@b\&[[lFmPR3TDs;r[lh+Fa")%grH$cWnGD
+ >2,kKA"^HC"g)s/i03L2YW(nq_([2o4Ea,WgTd(o.PK=FtYHi0</fDD0R-8<[XZ6H5219#kH^lZc!F-sib.icBtUXo->b@7DO%Ii*bg9YWu>lHItO1,*skb7r]GE!hR#
+ /;L7n*SO]LH3S`^nFU04O9eG<*IjA:O^7uS/1h?579Y[,XKrg=$C*:7;u2":*K&-2Dh:_c!_*'XDmgH?IQrA[#f%_a=hm1h(l-u?>nPbVL'g14k_"f:PkSc.)dXYV&EI
+ u0S_Y(39[BBVu&Q&L'[%Q]TGPgUP+X%JL5@>>K-]%<k0l5fUYXK<ORot+"MW7Po=?(h]1c4a*A.Xq$IARX#Q-s)K`+K\C=`89YFm_Msl,\:?V23Y,8ha(nn>7;cbc13R
+ +I5bX&WQ(h_8T^%b*S9<YPE+OM$Q+;.*J`..']`Q[ir`F-tu0E:@#^ck4mu>2oE43kWhF;I?F$QKf3kbJ]auKcJXpE>s+BaP)?8A.E[J@u[87kT9OOokNR==6/ppg?^" %>pa.:
+ [J]dg_Qgd//:ks4f:mr,;b9nLWuG1'0S_[6F5TT.eF3N@m!A5B#+U+>I?F\`\5m/AFW8`*XE%:l#?<[N0\(tO<WpI6dV*)"F-pHoj'R,Poc4~> EI Q