Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/7197#discussion_r33984870
--- Diff:
unsafe/src/main/java/org/apache/spark/unsafe/types/UTF8String.java ---
@@ -106,92 +168,125 @@ public int length() {
* @param until the position after last code point, exclusive.
*/
public UTF8String substring(final int start, final int until) {
- if (until <= start || start >= bytes.length) {
+ if (until <= start || start >= numBytes) {
return UTF8String.fromBytes(new byte[0]);
}
int i = 0;
int c = 0;
- for (; i < bytes.length && c < start; i += numBytes(bytes[i])) {
+ for (; i < numBytes && c < start; i +=
numBytesForFirstByte(getByte(i))) {
c += 1;
}
int j = i;
- for (; j < bytes.length && c < until; j += numBytes(bytes[i])) {
+ for (; j < numBytes && c < until; j +=
numBytesForFirstByte(getByte(i))) {
c += 1;
}
- return UTF8String.fromBytes(Arrays.copyOfRange(bytes, i, j));
+ byte[] bytes = new byte[j - i];
+ copyMemory(base, offset + i, bytes, BYTE_ARRAY_OFFSET, j - i);
+ return UTF8String.fromBytes(bytes);
}
public boolean contains(final UTF8String substring) {
- final byte[] b = substring.getBytes();
- if (b.length == 0) {
+ if (substring.numBytes == 0) {
return true;
}
- for (int i = 0; i <= bytes.length - b.length; i++) {
- if (bytes[i] == b[0] && startsWith(b, i)) {
+ byte first = substring.getByte(0);
+ for (int i = 0; i <= numBytes - substring.numBytes; i++) {
+ if (getByte(i) == first && matchAt(substring, i)) {
return true;
}
}
return false;
}
- private boolean startsWith(final byte[] prefix, int offsetInBytes) {
- if (prefix.length + offsetInBytes > bytes.length || offsetInBytes < 0)
{
+ private byte getByte(int i) {
+ return UNSAFE.getByte(base, offset + i);
+ }
+
+ private boolean matchAt(final UTF8String s, int pos) {
+ if (s.numBytes + pos > numBytes || pos < 0) {
return false;
}
- int i = 0;
- while (i < prefix.length && prefix[i] == bytes[i + offsetInBytes]) {
- i++;
- }
- return i == prefix.length;
+ return ByteArrayMethods.arrayEquals(base, offset + pos, s.base,
s.offset, s.numBytes);
}
public boolean startsWith(final UTF8String prefix) {
- return startsWith(prefix.getBytes(), 0);
+ return matchAt(prefix, 0);
}
public boolean endsWith(final UTF8String suffix) {
- return startsWith(suffix.getBytes(), bytes.length -
suffix.getBytes().length);
+ return matchAt(suffix, numBytes - suffix.numBytes);
}
public UTF8String toUpperCase() {
- return UTF8String.fromString(toString().toUpperCase());
+ // It's always have the same number of bytes for upper case
+ byte[] buf = new byte[numBytes];
+ copyMemory(base, offset, buf, BYTE_ARRAY_OFFSET, numBytes);
+
+ for (int i = 0; i < numBytes; ){
+ int n = numBytesForFirstByte(getByte(i));
+ int code = codePointAt(i, n);
+ int upper = Character.toUpperCase(code);
+ if (upper != code) {
+ updateCodePoint(buf, BYTE_ARRAY_OFFSET + i, upper, n);
+ }
+ i += n;
+ }
+
+ return fromBytes(buf);
}
public UTF8String toLowerCase() {
- return UTF8String.fromString(toString().toLowerCase());
+ // It's always have the same number of bytes for upper case
+ byte[] buf = new byte[numBytes];
+ copyMemory(base, offset, buf, BYTE_ARRAY_OFFSET, numBytes);
+
+ for (int i = 0; i < numBytes; ){
+ int n = numBytesForFirstByte(getByte(i));
+ int code = codePointAt(i, n);
+ int lower = Character.toLowerCase(code);
+ if (lower != code) {
+ updateCodePoint(buf, BYTE_ARRAY_OFFSET + i, lower, n);
+ }
+ i += n;
+ }
+ return fromBytes(buf);
}
@Override
public String toString() {
try {
- return new String(bytes, "utf-8");
+ // this is slow
+ return new String(getBytes(), "utf-8");
} catch (UnsupportedEncodingException e) {
// Turn the exception into unchecked so we can find out about it at
runtime, but
// don't need to add lots of boilerplate code everywhere.
- PlatformDependent.throwException(e);
+ throwException(e);
return "unknown"; // we will never reach here.
}
}
@Override
public UTF8String clone() {
- return new UTF8String().set(bytes);
+ return fromBytes(getBytes());
}
@Override
public int compareTo(final UTF8String other) {
- final byte[] b = other.getBytes();
- for (int i = 0; i < bytes.length && i < b.length; i++) {
- int res = bytes[i] - b[i];
+ int len = numBytes < other.numBytes ? numBytes : other.numBytes;
+ int i = 0;
+ // TODO: compare 8 bytes as unsigned long
+ while (i < len) {
+ int res = (getByte(i) & 0xFF) - (other.getByte(i) & 0xFF);
--- End diff --
Maybe we should add a comment here to explain that this is performing
unsigned comparison.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
