Github user davies commented on a diff in the pull request:
https://github.com/apache/spark/pull/7197#discussion_r34013254
--- Diff:
unsafe/src/main/java/org/apache/spark/unsafe/types/UTF8String.java ---
@@ -106,92 +105,126 @@ 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 >= size) {
return UTF8String.fromBytes(new byte[0]);
}
int i = 0;
int c = 0;
- for (; i < bytes.length && c < start; i += numBytes(bytes[i])) {
+ for (; i < size && c < start; i += numBytes(getByte(i))) {
c += 1;
}
int j = i;
- for (; j < bytes.length && c < until; j += numBytes(bytes[i])) {
+ for (; j < size && c < until; j += numBytes(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.size == 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 <= size - substring.size; 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 long getLong(int i) {
+ return UNSAFE.getLong(base, offset + i);
+ }
+
+ private byte getByte(int i) {
+ return UNSAFE.getByte(base, offset + i);
+ }
+
+ private boolean matchAt(final UTF8String s, int pos) {
+ if (s.size + pos > size || pos < 0) {
return false;
}
+
int i = 0;
- while (i < prefix.length && prefix[i] == bytes[i + offsetInBytes]) {
- i++;
+ while (i <= s.size - 8) {
+ if (getLong(pos + i) != s.getLong(i)) {
+ return false;
+ }
+ i += 8;
}
- return i == prefix.length;
+ while (i < s.size) {
+ if (getByte(pos + i) != s.getByte(i)) {
+ return false;
+ }
+ i += 1;
+ }
+ return true;
}
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, size - suffix.size);
}
public UTF8String toUpperCase() {
+ // this is locale aware
return UTF8String.fromString(toString().toUpperCase());
}
public UTF8String toLowerCase() {
+ // this is locale aware
return UTF8String.fromString(toString().toLowerCase());
}
@Override
public String toString() {
try {
- return new String(bytes, "utf-8");
+ // this is slow
+ return new String(getBytes(), "utf-8");
--- End diff --
When we do the copy in `getBytes`, `base` could be not a `byte[]`, so we
still need the copy. There is a case that we could avoid the copy, `base` is
`byte[]`, and `offset` is not pointed to the first byte, do we still need to
optimize this case? We can do it later.
---
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]
