[
https://issues.apache.org/jira/browse/HIVE-21160?focusedWorklogId=776260&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-776260
]
ASF GitHub Bot logged work on HIVE-21160:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 31/May/22 11:50
Start Date: 31/May/22 11:50
Worklog Time Spent: 10m
Work Description: kasakrisz commented on code in PR #2855:
URL: https://github.com/apache/hive/pull/2855#discussion_r885539665
##########
ql/src/java/org/apache/hadoop/hive/ql/parse/UpdateDeleteSemanticAnalyzer.java:
##########
@@ -267,9 +262,104 @@ private void reparseAndSuperAnalyze(ASTNode tree) throws
SemanticException {
}
}
- private void validateTxnManager(Table mTable) throws SemanticException {
- if (!AcidUtils.acidTableWithoutTransactions(mTable) &&
!getTxnMgr().supportsAcid()) {
- throw new SemanticException(ErrorMsg.ACID_OP_ON_NONACID_TXNMGR.getMsg());
+ private void analyzeSplitUpdate(ASTNode tree, Table mTable, ASTNode
tabNameNode) throws SemanticException {
+ operation = Context.Operation.UPDATE;
+
+ List<? extends Node> children = tree.getChildren();
+
+ ASTNode where = null;
+ int whereIndex = 2;
+ if (children.size() > whereIndex) {
+ where = (ASTNode) children.get(whereIndex);
+ assert where.getToken().getType() == HiveParser.TOK_WHERE :
+ "Expected where clause, but found " + where.getName();
+ }
+
+ Set<String> setRCols = new LinkedHashSet<>();
+// TOK_UPDATE_TABLE
+// TOK_TABNAME
+// ...
+// TOK_SET_COLUMNS_CLAUSE <- The set list from update should be the
second child (index 1)
+ assert children.size() >= 2 : "Expected update token to have at least two
children";
+ ASTNode setClause = (ASTNode) children.get(1);
+ Map<String, ASTNode> setCols = collectSetColumnsAndExpressions(setClause,
setRCols, mTable);
+ Map<Integer, ASTNode> setColExprs = new
HashMap<>(setClause.getChildCount());
+
+ List<FieldSchema> nonPartCols = mTable.getCols();
+ Map<String, String> colNameToDefaultConstraint =
getColNameToDefaultValueMap(mTable);
+ List<String> values = new ArrayList<>(mTable.getCols().size());
+ StringBuilder rewrittenQueryStr = createRewrittenQueryStrBuilder();
+ rewrittenQueryStr.append("(SELECT ROW__ID");
+ for (int i = 0; i < nonPartCols.size(); i++) {
+ rewrittenQueryStr.append(',');
+ String name = nonPartCols.get(i).getName();
+ ASTNode setCol = setCols.get(name);
+ String identifier = HiveUtils.unparseIdentifier(name, this.conf);
+
+ if (setCol != null) {
+ if (setCol.getType() == HiveParser.TOK_TABLE_OR_COL &&
+ setCol.getChildCount() == 1 && setCol.getChild(0).getType() ==
HiveParser.TOK_DEFAULT_VALUE) {
+ rewrittenQueryStr.append(colNameToDefaultConstraint.get(name));
+ } else {
+ rewrittenQueryStr.append(identifier);
+ // This is one of the columns we're setting, record it's position so
we can come back
+ // later and patch it up. 0th is ROW_ID
+ setColExprs.put(i + 1, setCol);
+ }
+ } else {
+ rewrittenQueryStr.append(identifier);
+ }
+ rewrittenQueryStr.append(" AS ");
+ rewrittenQueryStr.append(identifier);
+
+ values.add("s." + identifier);
+ }
+ addPartitionColsToSelect(mTable.getPartCols(), rewrittenQueryStr);
+ addPartitionColsAsValues(mTable.getPartCols(), "s", values);
+ rewrittenQueryStr.append(" FROM
").append(getFullTableNameForSQL(tabNameNode)).append(") s\n");
+
+ appendInsertBranch(rewrittenQueryStr, null, values);
+ appendDeleteBranch(rewrittenQueryStr, null, "s",
Collections.singletonList("s.ROW__ID "));
+
+ appendSortBy(rewrittenQueryStr, Collections.singletonList("s.ROW__ID "));
+
+ ReparseResult rr = parseRewrittenQuery(rewrittenQueryStr, ctx.getCmd());
+ Context rewrittenCtx = rr.rewrittenCtx;
+ ASTNode rewrittenTree = rr.rewrittenTree;
+
+ ASTNode rewrittenInsert = new ASTSearcher().simpleBreadthFirstSearch(
+ rewrittenTree, HiveParser.TOK_FROM, HiveParser.TOK_SUBQUERY,
HiveParser.TOK_INSERT);
+
+ rewrittenCtx.setOperation(Context.Operation.UPDATE);
+ rewrittenCtx.addDestNamePrefix(1, Context.DestClausePrefix.INSERT);
+ rewrittenCtx.addDeleteOfUpdateDestNamePrefix(2,
Context.DestClausePrefix.DELETE);
+
+ if (where != null) {
+ rewrittenInsert.addChild(where);
+ }
+
+ patchProjectionForUpdate(rewrittenInsert, setColExprs);
+
+ try {
Review Comment:
I was not able to get the final rewritten query string from the AST which
has all the transformations. To do it I need a `TokenRewriteStream` instance
belongs to the transformed AST. The one which I have from `parseRewrittenQuery`
is outdated because of the transformations made on the AST.
I added `rewrittenTree.dump()` before calling `super.analyze(rewrittenTree,
rewrittenCtx);`
It prints the string representation of the AST. I admit that this is more
difficult to read but better than nothing.
Issue Time Tracking
-------------------
Worklog Id: (was: 776260)
Time Spent: 2h 50m (was: 2h 40m)
> Rewrite Update statement as Multi-insert and do Update split early
> ------------------------------------------------------------------
>
> Key: HIVE-21160
> URL: https://issues.apache.org/jira/browse/HIVE-21160
> Project: Hive
> Issue Type: Sub-task
> Components: Transactions
> Affects Versions: 3.0.0
> Reporter: Eugene Koifman
> Assignee: Krisztian Kasa
> Priority: Major
> Labels: pull-request-available
> Time Spent: 2h 50m
> Remaining Estimate: 0h
>
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
