Content type detection of plain text documents. This detector looks at the beginning of the document input stream and considers the document to be a text document if no ASCII (ISO-Latin-1, UTF-8, etc.) control bytes are found. As a special case some control bytes (up to 2% of all characters) are als
| 37 | * @since Apache Tika 0.3 |
| 38 | */ |
| 39 | public class TextDetector implements Detector { |
| 40 | |
| 41 | /** |
| 42 | * Serial version UID |
| 43 | */ |
| 44 | private static final long serialVersionUID = 4774601079503507765L; |
| 45 | |
| 46 | /** |
| 47 | * The number of bytes from the beginning of the document stream |
| 48 | * to test for control bytes. |
| 49 | */ |
| 50 | private static final int DEFAULT_NUMBER_OF_BYTES_TO_TEST = 512; |
| 51 | |
| 52 | /** |
| 53 | * Lookup table for all the ASCII/ISO-Latin/UTF-8/etc. control bytes |
| 54 | * in the range below 0x20 (the space character). If an entry in this |
| 55 | * table is <code>true</code> then that byte is very unlikely to occur |
| 56 | * in a plain text document. |
| 57 | * <p> |
| 58 | * The contents of this lookup table are based on the following definition |
| 59 | * from section 4 of the "Content-Type Processing Model" Internet-draft |
| 60 | * (<a href="http://webblaze.cs.berkeley.edu/2009/mime-sniff/mime-sniff.txt" |
| 61 | * >draft-abarth-mime-sniff-01</a>). |
| 62 | * <pre> |
| 63 | * +-------------------------+ |
| 64 | * | Binary data byte ranges | |
| 65 | * +-------------------------+ |
| 66 | * | 0x00 -- 0x08 | |
| 67 | * | 0x0B | |
| 68 | * | 0x0E -- 0x1A | |
| 69 | * | 0x1C -- 0x1F | |
| 70 | * +-------------------------+ |
| 71 | * </pre> |
| 72 | * |
| 73 | * @see <a href="https://issues.apache.org/jira/browse/TIKA-154">TIKA-154</a> |
| 74 | */ |
| 75 | private static final boolean[] IS_CONTROL_BYTE = new boolean[0x20]; |
| 76 | |
| 77 | static { |
| 78 | Arrays.fill(IS_CONTROL_BYTE, true); |
| 79 | IS_CONTROL_BYTE[0x09] = false; // tabulator |
| 80 | IS_CONTROL_BYTE[0x0A] = false; // new line |
| 81 | IS_CONTROL_BYTE[0x0C] = false; // new page |
| 82 | IS_CONTROL_BYTE[0x0D] = false; // carriage return |
| 83 | IS_CONTROL_BYTE[0x1B] = false; // escape |
| 84 | } |
| 85 | |
| 86 | private final int bytesToTest; |
| 87 | |
| 88 | /** |
| 89 | * Constructs a {@link TextDetector} which will look at the default number |
| 90 | * of bytes from the beginning of the document. |
| 91 | */ |
| 92 | public TextDetector() { |
| 93 | this(DEFAULT_NUMBER_OF_BYTES_TO_TEST); |
| 94 | } |
| 95 | |
| 96 | /** |
nothing calls this directly
no test coverage detected
searching dependent graphs…