Detect encoding in InputStream

Detect the encoding/Charset used in a Java InputStream by reading the BOM (Byte-Order-Mark)

1.12k

/**
 * Detect the encoding/Charset used in a InputStream by reading the BOM (Byte-Order-Mark)
 * @author Mattias Jiderhamn
 */
public class EncodingDetectingInputStream extends PushbackInputStream {

  /** Known Charsets and their BOMs*/
  private static final Map<Charset, int[]> BOM_CHARSET;

  /** Length of the longest known BOM */
  private static final int BOM_MAX_LENGTH;

  static {
    // Define known BOMS
    Map<Charset, int[]> bomCharset = new HashMap<Charset, int[]>();
    bomCharset.put(Charset.forName("UTF-16BE"), new int[] {0xFE, 0xFF});
    bomCharset.put(Charset.forName("UTF-16LE"), new int[] {0xFF, 0xFE});
    bomCharset.put(Charset.forName("UTF-8"),    new int[] {0xEF, 0xBB, 0xBF});

    // Additional BOMs - add as needed
//          {0x00, 0x00, 0xFE, 0xFF},
//          {0xFF, 0xFE, 0x00, 0x00},
//          {0x2B, 0x2F, 0x76, 0x38},
//          {0x2B, 0x2F, 0x76, 0x39},
//          {0x2B, 0x2F, 0x76, 0x2B},
//          {0x2B, 0x2F, 0x76, 0x2F},
//          {0xDD, 0x73, 0x66, 0x73},
//          {0x0E, 0xFE, 0xFF},
//          {0xFB, 0xEE, 0x28},

    BOM_CHARSET = Collections.unmodifiableMap(bomCharset);

    int bomMaxLength = 0;
    for(int[] bom : BOM_CHARSET.values())
      bomMaxLength = Math.max(bomMaxLength, bom.length);
    BOM_MAX_LENGTH = bomMaxLength;
  }

  /** Test if a given BOM exists in the buffer */
  private static boolean testForBOM(int[] bom, int[] bytes) {
    for (int index = 0; index < bom.length; index++) {
      if (bom[index] != bytes[index])
        return false;
    }
    return true;
  }

  //////////////////////////////////////////////////////////////////////////////////////////////////////////////////

  /** The Charset detected in the InputStream of this instance */
  private Charset detectedCharset;

  /** Get detected charset */
  public Charset getCharset() {
    return detectedCharset;
  }
  /** Constructor that will automatically detect the encoding / Charset */
  public EncodingDetectingInputStream(InputStream is) throws IOException {
    super(is, BOM_MAX_LENGTH);

    // Read the beginning of the stream - potentially containing a BOM
    final int buffer[] = new int[BOM_MAX_LENGTH];
    for(int i = 0; i < BOM_MAX_LENGTH; i++) {
      int r = read();
      if(r == -1)
        break;
      else
        buffer[i] = r;
    }

    // Detect if BOM is present
    int bomLength = 0;
    for(Charset charset : BOM_CHARSET.keySet()) {
      final int[] bom = BOM_CHARSET.get(charset);
      if(testForBOM(bom, buffer)) {
        this.detectedCharset = charset;
        bomLength = bom.length;
      }
    }

    // Push back bytes read that were not part of BOM
    for (int index = buffer.length - 1; index >= bomLength; index--) {
      if (buffer[index] != -1)
        unread(buffer[index]);
    }
  }
}