On Wed, Feb 15, 2012 at 11:28 PM, Andreas Menke <[email protected]> wrote:
> Hi,
>
> usage of ensureBufferSize() seems to be 'asymptotic' safe (despite the
> fact that a line separator might be much bigger than DEFAULT_BUFFER_SIZE,
> see below, and that no body else use the function as he is told by
> description..).
>
> /** Increases our buffer by the {@link #DEFAULT_BUFFER_RESIZE_FACTOR}**.
> */
> private void resizeBuffer(int size) {
> //size = 0;
> if (buffer == null) {
> int nSize = getDefaultBufferSize();
> if(nSize < size )
> nSize = size;
> buffer = new byte[nSize];
> pos = 0;
> readPos = 0;
> } else {
> int nSize = buffer.length * DEFAULT_BUFFER_RESIZE_FACTOR;
> if( 0 > nSize ) // should be tested with huge data
> nSize = Integer.MAX_VALUE;
> else if (nSize < size)
> nSize = size;
> byte[] b = new byte[nSize];
> System.arraycopy(buffer, 0, b, 0, pos);
> buffer = b;
> }
> }
>
> /**
> * Ensure that the buffer has room for <code>size</code> bytes
> *
> * @param size minimum spare space required
> */
> protected void ensureBufferSize(int size){
> if ((buffer == null) || (buffer.length < pos + size)){
> resizeBuffer(pos+size);
> }
> }
>
> The test case is
>
>
> /**
> * lineSeparator much bigger than DEFAULT_BUFFER_SIZE.
> */
> @Test
> public void testDEFAULT_BUFFER_SIZE() {
> byte[] baLineSeparator = new byte[BaseNCodec_DEFAULT_**
> BUFFER_SIZE*4-3];
> Base64 b64 = new Base64(Base64_BYTES_PER_**ENCODED_BLOCK,
> baLineSeparator);
> String strOriginal = "Hello World";
> String strDecoded = new String(b64.decode(b64.encode(**
> StringUtils.getBytesUtf8(**strOriginal))));
> assertTrue("testDEFAULT_**BUFFER_SIZE",
> strOriginal.equals(strDecoded)**);
> }
> private int BaseNCodec_DEFAULT_BUFFER_SIZE = 8192;
> private int Base64_BYTES_PER_ENCODED_BLOCK = 4;
>
> It's astonishing how big the lineSeparator must be to see the error. I
> hope I have nothing overseen ('verschlimmbessern'). The code might work
> with huge data (not tested).
>
This is an extreme case indeed. I'll expand these two data points from the
test case below:
- line length is set to 4
- line separator length is 32765
The line separator length is usually 2 for CRLF.
Gary
>
>
> My personal view is, that the code is quite complicated. The loop
>
> for (int i = 0; i < inAvail; i++) {
> ensureBufferSize(encodeSize);
> modulus = (modulus+1) % BYTES_PER_UNENCODED_BLOCK;
> int b = in[inPos++];
> if (b < 0) {
> b += 256;
> }
> bitWorkArea = (bitWorkArea << 8) + b; // BITS_PER_BYTE
> if (0 == modulus) { // 3 bytes = 24 bits = 4 * 6 bits to
> extract
>
> with constant 'encodeSize' might slow down the algorithm. The estimated
> length is calculated
>
> long len = b64.getEncodedLength(**binaryData);
> if (len > maxResultSize) {
>
> but the value is not further used. The result array is doubled
>
> byte[] buf = new byte[pos - readPos];
> readResults(buf, 0, buf.length);
> return buf;
>
> which is memory wasting in case input is very large.
>
> Regards
>
> Andreas
>
> --
>
>
> Andreas Menke
>
> Diplom-Informatiker Univ.
> Software-Entwicklung
>
> Fon 0151 5081 1173
> Mail [email protected]
>
>
> ------------------------------**------------------------------**---------
> To unsubscribe, e-mail:
> dev-unsubscribe@commons.**apache.org<[email protected]>
> For additional commands, e-mail: [email protected]
>
>
--
E-Mail: [email protected] | [email protected]
JUnit in Action, 2nd Ed: <http://goog_1249600977>http://bit.ly/ECvg0
Spring Batch in Action: <http://s.apache.org/HOq>http://bit.ly/bqpbCK
Blog: http://garygregory.wordpress.com
Home: http://garygregory.com/
Tweet! http://twitter.com/GaryGregory
