priyankporwal commented on a change in pull request #469: PHOENIX-5156
Consistent Global Indexes for Non-Transactional Tables
URL: https://github.com/apache/phoenix/pull/469#discussion_r282155767
##########
File path: phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java
##########
@@ -849,4 +857,59 @@ public static long
getIndexPendingDisableCount(PhoenixConnection conn, String fa
throw new IOException(e);
}
}
+
+ private static boolean containsOneOrMoreColumn(Scan scan) {
+ Map<byte[], NavigableSet<byte[]>> familyMap = scan.getFamilyMap();
+ if (familyMap == null || familyMap.isEmpty()) {
+ return false;
+ }
+ for (Map.Entry<byte[], NavigableSet<byte[]>> entry :
familyMap.entrySet()) {
+ NavigableSet<byte[]> family = entry.getValue();
+ if (family != null && !family.isEmpty()) {
+ return true;
+ }
+ }
+ return false;
+ }
+
+ private static void updateScan(Scan scan, byte[] emptyCF, byte[] emptyCQ) {
+ if (containsOneOrMoreColumn(scan)) {
+ scan.addColumn(emptyCF, emptyCQ);
+ return;
+ }
+ return;
+ }
+
+ public static void setScanAttributes(Scan scan, PTable table,
PhoenixConnection phoenixConnection) throws SQLException {
+ if (table.isTransactional() || table.isImmutableRows() ||
table.getType() != PTableType.INDEX) {
+ return;
+ }
+ PTable indexTable = table;
+ if (indexTable.getIndexType() != PTable.IndexType.GLOBAL) {
+ return;
+ }
+ String schemaName = indexTable.getParentSchemaName().getString();
+ String tableName = indexTable.getParentTableName().getString();
+ PTable dataTable = PhoenixRuntime.getTable(phoenixConnection,
SchemaUtil.getTableName(schemaName, tableName));
Review comment:
The cache would be on the client side to avoid hitting syscat on each read.
I don't think we should not use the cache. Do we really need to get dataTable?
I see it's usage for checking index really being on it (lines 919-921) and for
PHYSICAL_DATA_TABLE_NAME. Can we simply not do the check on lines 919-921 and
use tableName (from line 917)? Alternative solution - wouldn't the
GlobalIndexChecker already have the index-table properties and know the
data-table info? That way we can avoid the clients sending this needless info
to server side; and be more fool-proof as well. [Imagine a rogue client sending
a bad physical data-table name to corrupt the index]
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
