database/doris
2
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

[Feature](merge-on-write)Support ignore mode for merge-on-write unique table (#21773)

Browse Source
This commit is contained in:
bobhan1
2023-09-14 18:03:51 +08:00
committed by GitHub
parent 68b13ab50f
commit 3ee89aea35
34 changed files with 479 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
be/src/exec/tablet_info.cpp
View File

@ -124,6 +124,7 @@ Status OlapTableSchemaParam::init(const POlapTableSchemaParam& pschema) {
_version = pschema.version();
_is_partial_update = pschema.partial_update();
_is_strict_mode = pschema.is_strict_mode();
_is_unique_key_ignore_mode = pschema.is_unique_key_ignore_mode();
for (auto& col : pschema.partial_update_input_columns()) {
_partial_update_input_columns.insert(col);
@ -176,9 +177,9 @@ Status OlapTableSchemaParam::init(const TOlapTableSchemaParam& tschema) {
_table_id = tschema.table_id;
_version = tschema.version;
_is_partial_update = tschema.is_partial_update;
if (tschema.__isset.is_strict_mode) {
_is_strict_mode = tschema.is_strict_mode;
}
_is_strict_mode = tschema.__isset.is_strict_mode && tschema.is_strict_mode;
_is_unique_key_ignore_mode =
tschema.__isset.is_unique_key_ignore_mode && tschema.is_unique_key_ignore_mode;
for (auto& tcolumn : tschema.partial_update_input_columns) {
_partial_update_input_columns.insert(tcolumn);
@ -246,6 +247,7 @@ void OlapTableSchemaParam::to_protobuf(POlapTableSchemaParam* pschema) const {
pschema->set_version(_version);
pschema->set_partial_update(_is_partial_update);
pschema->set_is_strict_mode(_is_strict_mode);
pschema->set_is_unique_key_ignore_mode(_is_unique_key_ignore_mode);
for (auto col : _partial_update_input_columns) {
*pschema->add_partial_update_input_columns() = col;
}

2
be/src/exec/tablet_info.h
View File

@ -90,6 +90,7 @@ public:
return _partial_update_input_columns;
}
bool is_strict_mode() const { return _is_strict_mode; }
bool is_unique_key_ignore_mode() const { return _is_unique_key_ignore_mode; }
std::string debug_string() const;
private:
@ -104,6 +105,7 @@ private:
bool _is_partial_update = false;
std::set<std::string> _partial_update_input_columns;
bool _is_strict_mode = false;
bool _is_unique_key_ignore_mode = false;
};
using OlapTableIndexTablets = TOlapTableIndexTablets;

7
be/src/http/action/stream_load.cpp
View File

@ -555,6 +555,13 @@ Status StreamLoadAction::_process_put(HttpRequest* http_req,
bool value = iequal(http_req->header(HTTP_MEMTABLE_ON_SINKNODE), "true");
request.__set_memtable_on_sink_node(value);
}
if (!http_req->header(HTTP_IGNORE_MODE).empty()) {
if (iequal(http_req->header(HTTP_IGNORE_MODE), "true")) {
request.__set_ignore_mode(true);
} else {
request.__set_ignore_mode(false);
}
}
#ifndef BE_TEST
// plan this load

1
be/src/http/http_common.h
View File

@ -58,6 +58,7 @@ static const std::string HTTP_SKIP_LINES = "skip_lines";
static const std::string HTTP_COMMENT = "comment";
static const std::string HTTP_ENABLE_PROFILE = "enable_profile";
static const std::string HTTP_PARTIAL_COLUMNS = "partial_columns";
static const std::string HTTP_IGNORE_MODE = "ignore_mode";
static const std::string HTTP_SQL = "sql";
static const std::string HTTP_TWO_PHASE_COMMIT = "two_phase_commit";
static const std::string HTTP_TXN_ID_KEY = "txn_id";

2
be/src/olap/rowset/beta_rowset_writer.cpp
View File

@ -149,7 +149,7 @@ Status BetaRowsetWriter::_generate_delete_bitmap(int32_t segment_id) {
OlapStopWatch watch;
RETURN_IF_ERROR(_context.tablet->calc_delete_bitmap(
rowset, segments, specified_rowsets, _context.mow_context->delete_bitmap,
_context.mow_context->max_version, nullptr));
_context.mow_context->max_version, nullptr, nullptr));
size_t total_rows = std::accumulate(
segments.begin(), segments.end(), 0,
[](size_t sum, const segment_v2::SegmentSharedPtr& s) { return sum += s->num_rows(); });

1
be/src/olap/rowset_builder.cpp
View File

@ -322,6 +322,7 @@ void RowsetBuilder::_build_current_tablet_schema(int64_t index_id,
_tablet_schema->set_partial_update_info(table_schema_param->is_partial_update(),
table_schema_param->partial_update_input_columns());
_tablet_schema->set_is_strict_mode(table_schema_param->is_strict_mode());
_tablet_schema->set_is_unique_key_ignore_mode(table_schema_param->is_unique_key_ignore_mode());
}
} // namespace doris

74
be/src/olap/tablet.cpp
View File

@ -107,6 +107,7 @@
#include "olap/txn_manager.h"
#include "olap/types.h"
#include "olap/utils.h"
#include "runtime/define_primitive_type.h"
#include "segment_loader.h"
#include "service/point_query_executor.h"
#include "util/bvar_helper.h"
@ -2884,6 +2885,7 @@ Status Tablet::calc_segment_delete_bitmap(RowsetSharedPtr rowset,
Version dummy_version(end_version + 1, end_version + 1);
auto rowset_schema = rowset->tablet_schema();
bool is_partial_update = rowset_schema->is_partial_update();
bool is_unique_key_ignore_mode = rowset_schema->is_unique_key_ignore_mode();
// use for partial update
PartialUpdateReadPlan read_plan_ori;
PartialUpdateReadPlan read_plan_update;
@ -2951,42 +2953,50 @@ Status Tablet::calc_segment_delete_bitmap(RowsetSharedPtr rowset,
if (st.is<KEY_NOT_FOUND>()) {
continue;
}
// sequence id smaller than the previous one, so delete current row
if (st.is<KEY_ALREADY_EXISTS>()) {
delete_bitmap->add({rowset_id, seg->id(), DeleteBitmap::TEMP_VERSION_COMMON},
row_id);
continue;
} else if (is_partial_update && rowset_writer != nullptr) {
// In publish version, record rows to be deleted for concurrent update
// For example, if version 5 and 6 update a row, but version 6 only see
// version 4 when write, and when publish version, version 5's value will
// be marked as deleted and it's update is losed.
// So here we should read version 5's columns and build a new row, which is
// consists of version 6's update columns and version 5's origin columns
// here we build 2 read plan for ori values and update values
prepare_to_read(loc, pos, &read_plan_ori);
prepare_to_read(RowLocation {rowset_id, seg->id(), row_id}, pos, &read_plan_update);
rsid_to_rowset[rowset_find->rowset_id()] = rowset_find;
++pos;
// delete bitmap will be calculate when memtable flush and
// publish. The two stages may see different versions.
// When there is sequence column, the currently imported data
// of rowset may be marked for deletion at memtablet flush or
// publish because the seq column is smaller than the previous
// rowset.
// just set 0 as a unified temporary version number, and update to
// the real version number later.
if (UNLIKELY(is_unique_key_ignore_mode)) {
if (st.is<OK>() || st.is<KEY_ALREADY_EXISTS>()) {
delete_bitmap->add({rowset_id, seg->id(), DeleteBitmap::TEMP_VERSION_COMMON},
row_id);
}
} else {
// sequence id smaller than the previous one, so delete current row
if (st.is<KEY_ALREADY_EXISTS>()) {
delete_bitmap->add({rowset_id, seg->id(), DeleteBitmap::TEMP_VERSION_COMMON},
row_id);
continue;
} else if (is_partial_update && rowset_writer != nullptr) {
// In publish version, record rows to be deleted for concurrent update
// For example, if version 5 and 6 update a row, but version 6 only see
// version 4 when write, and when publish version, version 5's value will
// be marked as deleted and it's update is losed.
// So here we should read version 5's columns and build a new row, which is
// consists of version 6's update columns and version 5's origin columns
// here we build 2 read plan for ori values and update values
prepare_to_read(loc, pos, &read_plan_ori);
prepare_to_read(RowLocation {rowset_id, seg->id(), row_id}, pos,
&read_plan_update);
rsid_to_rowset[rowset_find->rowset_id()] = rowset_find;
++pos;
// delete bitmap will be calculate when memtable flush and
// publish. The two stages may see different versions.
// When there is sequence column, the currently imported data
// of rowset may be marked for deletion at memtablet flush or
// publish because the seq column is smaller than the previous
// rowset.
// just set 0 as a unified temporary version number, and update to
// the real version number later.
delete_bitmap->add(
{loc.rowset_id, loc.segment_id, DeleteBitmap::TEMP_VERSION_COMMON},
loc.row_id);
delete_bitmap->add({rowset_id, seg->id(), DeleteBitmap::TEMP_VERSION_COMMON},
row_id);
continue;
}
// when st = ok
delete_bitmap->add(
{loc.rowset_id, loc.segment_id, DeleteBitmap::TEMP_VERSION_COMMON},
loc.row_id);
delete_bitmap->add({rowset_id, seg->id(), DeleteBitmap::TEMP_VERSION_COMMON},
row_id);
continue;
}
// when st = ok
delete_bitmap->add({loc.rowset_id, loc.segment_id, DeleteBitmap::TEMP_VERSION_COMMON},
loc.row_id);
}
remaining -= num_read;
}

2
be/src/olap/tablet_schema.cpp
View File

@ -759,6 +759,7 @@ void TabletSchema::init_from_pb(const TabletSchemaPB& schema) {
_compression_type = schema.compression_type();
_schema_version = schema.schema_version();
_is_partial_update = schema.is_partial_update();
_is_unique_key_ignore_mode = schema.is_unique_key_ignore_mode();
for (auto& col_name : schema.partial_update_input_columns()) {
_partial_update_input_columns.emplace(col_name);
}
@ -917,6 +918,7 @@ void TabletSchema::to_schema_pb(TabletSchemaPB* tablet_schema_pb) const {
tablet_schema_pb->set_compression_type(_compression_type);
tablet_schema_pb->set_version_col_idx(_version_col_idx);
tablet_schema_pb->set_is_partial_update(_is_partial_update);
tablet_schema_pb->set_is_unique_key_ignore_mode(_is_unique_key_ignore_mode);
for (auto& col : _partial_update_input_columns) {
*tablet_schema_pb->add_partial_update_input_columns() = col;
}

9
be/src/olap/tablet_schema.h
View File

@ -363,8 +363,12 @@ public:
}
void set_is_strict_mode(bool is_strict_mode) { _is_strict_mode = is_strict_mode; }
bool is_strict_mode() const { return _is_strict_mode; }
std::vector<uint32_t> get_missing_cids() const { return _missing_cids; }
std::vector<uint32_t> get_update_cids() const { return _update_cids; }
void set_is_unique_key_ignore_mode(bool is_unique_key_ignore_mode) {
_is_unique_key_ignore_mode = is_unique_key_ignore_mode;
}
bool is_unique_key_ignore_mode() const { return _is_unique_key_ignore_mode; }
std::vector<uint32_t> get_missing_cids() { return _missing_cids; }
std::vector<uint32_t> get_update_cids() { return _update_cids; }
private:
friend bool operator==(const TabletSchema& a, const TabletSchema& b);
@ -411,6 +415,7 @@ private:
// to generate a new row, only available in non-strict mode
bool _can_insert_new_rows_in_partial_update = true;
bool _is_strict_mode = false;
bool _is_unique_key_ignore_mode = false;
};
bool operator==(const TabletSchema& a, const TabletSchema& b);

2
docs/en/docs/sql-manual/sql-reference/Data-Manipulation-Statements/Load/STREAM-LOAD.md
View File

@ -156,6 +156,8 @@ separated by commas.
28. comment: <version since="1.2.3" type="inline"> String type, the default value is "". </version>
29. ignore_mode: <version since="dev" type="inline"> Ignore mode, only effective when the target table is a unique table with merge-on-write enabled. When insert ignore mode is enabled, for the inserted rows, if the key of the row does not exist in the table, the row will be inserted. If the key already exists in the table, the row will be discarded. When sequence columns exists in the target table, the ignore mode can't be enabled in stream load.</version>
### Example
1. Import the data in the local file 'testData' into the table 'testTbl' in the database 'testDb', and use Label for deduplication. Specify a timeout of 100 seconds

4
docs/en/docs/sql-manual/sql-reference/Data-Manipulation-Statements/Manipulation/INSERT.md
View File

@ -35,7 +35,7 @@ INSERT
The change statement is to complete the data insertion operation.
```sql
INSERT INTO table_name
INSERT [IGNORE] INTO table_name
[ PARTITION (p1, ...) ]
[ WITH LABEL label]
[ (column [, ...]) ]
@ -44,7 +44,7 @@ INSERT INTO table_name
````
Parameters
> IGNORE: insert ignore mode, only effective when the target table is a unique table with merge-on-write enabled. When insert ignore mode is enabled, for the inserted rows, if the key of the row does not exist in the table, the row will be inserted. If the key already exists in the table, the row will be discarded. When sequence column exists in the target table, the `insert ignore` statements are forbidden.
> tablet_name: The destination table for importing data. Can be of the form `db_name.table_name`
>
> partitions: Specify the partitions to be imported, which must be partitions that exist in `table_name`. Multiple partition names are separated by commas

2
docs/zh-CN/docs/sql-manual/sql-reference/Data-Manipulation-Statements/Load/STREAM-LOAD.md
View File

@ -152,6 +152,8 @@ curl --location-trusted -u user:passwd [-H ""...] -T data.file -XPUT http://fe_h
28. comment: <version since="1.2.3" type="inline"> 字符串类型, 默认值为空. 给任务增加额外的信息. </version>
29. ignore_mode: <version since="dev" type="inline"> ignore模式，仅当目标表为开启merge-on-write的unique表时有效。开启后，对于插入的行，如果该行的key在表中不存在，则插入该行数据。如果key在表中不存在，则丢弃这行数据。当目标表中存在sequence列时stream无法开启ignore mode。</version>
### Example
1. 将本地文件'testData'中的数据导入到数据库'testDb'中'testTbl'的表，使用Label用于去重。指定超时时间为 100 秒

5
docs/zh-CN/docs/sql-manual/sql-reference/Data-Manipulation-Statements/Manipulation/INSERT.md
View File

@ -35,7 +35,7 @@ INSERT
该语句是完成数据插入操作。
```sql
INSERT INTO table_name
INSERT [IGNORE] INTO table_name
[ PARTITION (p1, ...) ]
[ WITH LABEL label]
[ (column [, ...]) ]
@ -44,7 +44,8 @@ INSERT INTO table_name
```
Parameters
> IGNORE: insert ignore模式，仅当目标表为开启merge-on-write的unique表时有效。开启后，对于插入的行，如果该行的key在表中不存在，则插入该行数据。如果key在表中不存在，则丢弃这行数据。当目标表中存在sequence列时无法通过insert ignore语句进行插入操作。
>
> tablet_name: 导入数据的目的表。可以是 `db_name.table_name` 形式
>
> partitions: 指定待导入的分区，必须是 `table_name` 中存在的分区，多个分区名称用逗号分隔

21
fe/fe-core/src/main/cup/sql_parser.cup
View File

@ -887,7 +887,7 @@ nonterminal ParseNode load_property;
nonterminal List<ParseNode> opt_load_property_list;
// Boolean
nonterminal Boolean opt_negative, opt_is_allow_null, opt_is_key, opt_read_only, opt_aggregate, opt_local, opt_is_auto_inc;
nonterminal Boolean opt_negative, opt_is_allow_null, opt_is_key, opt_read_only, opt_aggregate, opt_local, opt_is_auto_inc, opt_is_insert_ignore;
nonterminal String opt_from_rollup, opt_to_rollup;
nonterminal ColumnPosition opt_col_pos;
@ -3718,6 +3718,17 @@ opt_is_auto_inc ::=
RESULT = true;
:}
;
opt_is_insert_ignore ::=
{:
RESULT = false;
:}
| KW_IGNORE
{:
RESULT = true;
:}
;
opt_comment ::=
/* empty */
{:
@ -4805,16 +4816,16 @@ insert_overwrite_stmt ::=
// Insert statement
insert_stmt ::=
KW_INSERT KW_INTO insert_target:target opt_with_label:label opt_col_list:cols opt_plan_hints:hints insert_source:source
KW_INSERT opt_is_insert_ignore:is_insert_ignore KW_INTO insert_target:target opt_with_label:label opt_col_list:cols opt_plan_hints:hints insert_source:source
{:
RESULT = new NativeInsertStmt(target, label, cols, source, hints);
RESULT = new NativeInsertStmt(target, label, cols, source, hints, is_insert_ignore);
:}
// TODO(zc) add default value for SQL-2003
// | KW_INSERT KW_INTO insert_target:target KW_DEFAULT KW_VALUES
| /* used for group commit */
KW_INSERT KW_INTO INTEGER_LITERAL:table_id opt_with_label:label opt_col_list:cols opt_plan_hints:hints insert_source:source
KW_INSERT opt_is_insert_ignore:is_insert_ignore KW_INTO INTEGER_LITERAL:table_id opt_with_label:label opt_col_list:cols opt_plan_hints:hints insert_source:source
{:
RESULT = new NativeInsertStmt(table_id, label, cols, source, hints);
RESULT = new NativeInsertStmt(table_id, label, cols, source, hints, is_insert_ignore);
:}
;

3
fe/fe-core/src/main/java/org/apache/doris/analysis/DeleteStmt.java
View File

@ -201,7 +201,8 @@ public class DeleteStmt extends DdlStmt {
cols,
new InsertSource(selectStmt),
null,
isPartialUpdate);
isPartialUpdate,
false);
}
private void analyzeTargetTable(Analyzer analyzer) throws UserException {

28
fe/fe-core/src/main/java/org/apache/doris/analysis/NativeInsertStmt.java
View File

@ -166,6 +166,8 @@ public class NativeInsertStmt extends InsertStmt {
// true if be generates an insert from group commit tvf stmt and executes to load data
public boolean isInnerGroupCommit = false;
private boolean isInsertIgnore = false;
public NativeInsertStmt(InsertTarget target, String label, List<String> cols, InsertSource source,
List<String> hints) {
super(new LabelName(null, label), null, null);
@ -185,6 +187,26 @@ public class NativeInsertStmt extends InsertStmt {
this.tableId = tableId;
}
public NativeInsertStmt(long tableId, String label, List<String> cols, InsertSource source,
List<String> hints, boolean isInsertIgnore) {
this(new InsertTarget(new TableName(null, null, null), null), label, cols, source, hints, isInsertIgnore);
this.tableId = tableId;
}
public NativeInsertStmt(InsertTarget target, String label, List<String> cols, InsertSource source,
List<String> hints, boolean isInsertIgnore) {
super(new LabelName(null, label), null, null);
this.tblName = target.getTblName();
this.targetPartitionNames = target.getPartitionNames();
this.label = new LabelName(null, label);
this.queryStmt = source.getQueryStmt();
this.planHints = hints;
this.isInsertIgnore = isInsertIgnore;
this.targetColumnNames = cols;
this.isValuesOrConstantSelect = (queryStmt instanceof SelectStmt
&& ((SelectStmt) queryStmt).getTableRefs().isEmpty());
}
// Ctor for CreateTableAsSelectStmt and InsertOverwriteTableStmt
public NativeInsertStmt(TableName name, PartitionNames targetPartitionNames, LabelName label,
QueryStmt queryStmt, List<String> planHints, List<String> targetColumnNames) {
@ -199,10 +221,11 @@ public class NativeInsertStmt extends InsertStmt {
}
public NativeInsertStmt(InsertTarget target, String label, List<String> cols, InsertSource source,
List<String> hints, boolean isPartialUpdate) {
List<String> hints, boolean isPartialUpdate, boolean isInsertIgnore) {
this(target, label, cols, source, hints);
this.isPartialUpdate = isPartialUpdate;
this.partialUpdateCols.addAll(cols);
this.isInsertIgnore = isInsertIgnore;
}
public boolean isValuesOrConstantSelect() {
@ -382,7 +405,8 @@ public class NativeInsertStmt extends InsertStmt {
OlapTableSink sink = (OlapTableSink) dataSink;
TUniqueId loadId = analyzer.getContext().queryId();
int sendBatchParallelism = analyzer.getContext().getSessionVariable().getSendBatchParallelism();
sink.init(loadId, transactionId, db.getId(), timeoutSecond, sendBatchParallelism, false, false);
sink.init(loadId, transactionId, db.getId(), timeoutSecond,
sendBatchParallelism, false, false, isInsertIgnore);
}
}

3
fe/fe-core/src/main/java/org/apache/doris/analysis/UpdateStmt.java
View File

@ -125,7 +125,8 @@ public class UpdateStmt extends DdlStmt {
cols,
new InsertSource(selectStmt),
null,
isPartialUpdate);
isPartialUpdate,
false);
}
private void analyzeTargetTable(Analyzer analyzer) throws UserException {

3
fe/fe-core/src/main/java/org/apache/doris/load/loadv2/LoadingTaskPlanner.java
View File

@ -193,8 +193,9 @@ public class LoadingTaskPlanner {
List<Long> partitionIds = getAllPartitionIds();
OlapTableSink olapTableSink = new OlapTableSink(table, destTupleDesc, partitionIds,
Config.enable_single_replica_load);
olapTableSink.init(loadId, txnId, dbId, timeoutS, sendBatchParallelism, false, strictMode);
olapTableSink.init(loadId, txnId, dbId, timeoutS, sendBatchParallelism, false, strictMode, false);
olapTableSink.setPartialUpdateInputColumns(isPartialUpdate, partialUpdateInputColumns);
olapTableSink.complete(analyzer);
// 3. Plan fragment

2
fe/fe-core/src/main/java/org/apache/doris/nereids/trees/plans/commands/InsertIntoTableCommand.java
View File

@ -150,7 +150,7 @@ public class InsertIntoTableCommand extends Command implements ForwardWithSync,
sink.init(ctx.queryId(), txn.getTxnId(),
physicalOlapTableSink.getDatabase().getId(),
ctx.getExecTimeout(),
ctx.getSessionVariable().getSendBatchParallelism(), false, false);
ctx.getSessionVariable().getSendBatchParallelism(), false, false, false);
sink.complete(new Analyzer(Env.getCurrentEnv(), ctx));
TransactionState state = Env.getCurrentGlobalTransactionMgr().getTransactionState(

18
fe/fe-core/src/main/java/org/apache/doris/planner/OlapTableSink.java
View File

@ -26,6 +26,7 @@ import org.apache.doris.catalog.DistributionInfo;
import org.apache.doris.catalog.Env;
import org.apache.doris.catalog.HashDistributionInfo;
import org.apache.doris.catalog.Index;
import org.apache.doris.catalog.KeysType;
import org.apache.doris.catalog.ListPartitionItem;
import org.apache.doris.catalog.MaterializedIndex;
import org.apache.doris.catalog.MaterializedIndex.IndexExtState;
@ -104,6 +105,8 @@ public class OlapTableSink extends DataSink {
private boolean isStrictMode = false;
private boolean isUniqueKeyIgnoreMode = false;
public OlapTableSink(OlapTable dstTable, TupleDescriptor tupleDescriptor, List<Long> partitionIds,
boolean singleReplicaLoad) {
this.dstTable = dstTable;
@ -113,7 +116,7 @@ public class OlapTableSink extends DataSink {
}
public void init(TUniqueId loadId, long txnId, long dbId, long loadChannelTimeoutS, int sendBatchParallelism,
boolean loadToSingleTablet, boolean isStrictMode) throws AnalysisException {
boolean loadToSingleTablet, boolean isStrictMode, boolean isUniqueKeyIgnoreMode) throws AnalysisException {
TOlapTableSink tSink = new TOlapTableSink();
tSink.setLoadId(loadId);
tSink.setTxnId(txnId);
@ -121,6 +124,7 @@ public class OlapTableSink extends DataSink {
tSink.setLoadChannelTimeoutS(loadChannelTimeoutS);
tSink.setSendBatchParallelism(sendBatchParallelism);
this.isStrictMode = isStrictMode;
this.isUniqueKeyIgnoreMode = isUniqueKeyIgnoreMode;
if (loadToSingleTablet && !(dstTable.getDefaultDistributionInfo() instanceof RandomDistributionInfo)) {
throw new AnalysisException(
"if load_to_single_tablet set to true," + " the olap table must be with random distribution");
@ -186,6 +190,17 @@ public class OlapTableSink extends DataSink {
}
tSink.setWriteSingleReplica(singleReplicaLoad);
tSink.setNodesInfo(createPaloNodesInfo());
if (isUniqueKeyIgnoreMode) {
if (dstTable.getKeysType() != KeysType.UNIQUE_KEYS || !dstTable.getEnableUniqueKeyMergeOnWrite()) {
throw new UserException("ignore mode can only be enabled if the target table is "
+ "a unique table with merge-on-write enabled.");
} else if (isPartialUpdate) {
throw new UserException("ignore mode can't be used in partial update.");
} else if (dstTable.hasSequenceCol()) {
throw new UserException("ignore mode can't be used if the target table has sequence column, "
+ "but table[" + dstTable.getName() + "] has sequnce column.");
}
}
}
@Override
@ -221,6 +236,7 @@ public class OlapTableSink extends DataSink {
schemaParam.setTableId(table.getId());
schemaParam.setVersion(table.getIndexMetaByIndexId(table.getBaseIndexId()).getSchemaVersion());
schemaParam.setIsStrictMode(isStrictMode);
schemaParam.setIsUniqueKeyIgnoreMode(isUniqueKeyIgnoreMode);
schemaParam.tuple_desc = tupleDescriptor.toThrift();
for (SlotDescriptor slotDesc : tupleDescriptor.getSlots()) {

8
fe/fe-core/src/main/java/org/apache/doris/planner/StreamLoadPlanner.java
View File

@ -256,8 +256,8 @@ public class StreamLoadPlanner {
List<Long> partitionIds = getAllPartitionIds();
OlapTableSink olapTableSink = new OlapTableSink(destTable, tupleDesc, partitionIds,
Config.enable_single_replica_load);
olapTableSink.init(loadId, taskInfo.getTxnId(), db.getId(), timeout,
taskInfo.getSendBatchParallelism(), taskInfo.isLoadToSingleTablet(), taskInfo.isStrictMode());
olapTableSink.init(loadId, taskInfo.getTxnId(), db.getId(), timeout, taskInfo.getSendBatchParallelism(),
taskInfo.isLoadToSingleTablet(), taskInfo.isStrictMode(), taskInfo.isIgnoreMode());
olapTableSink.setPartialUpdateInputColumns(isPartialUpdate, partialUpdateInputColumns);
olapTableSink.complete(analyzer);
@ -465,8 +465,8 @@ public class StreamLoadPlanner {
List<Long> partitionIds = getAllPartitionIds();
OlapTableSink olapTableSink = new OlapTableSink(destTable, tupleDesc, partitionIds,
Config.enable_single_replica_load);
olapTableSink.init(loadId, taskInfo.getTxnId(), db.getId(), timeout,
taskInfo.getSendBatchParallelism(), taskInfo.isLoadToSingleTablet(), taskInfo.isStrictMode());
olapTableSink.init(loadId, taskInfo.getTxnId(), db.getId(), timeout, taskInfo.getSendBatchParallelism(),
taskInfo.isLoadToSingleTablet(), taskInfo.isStrictMode(), taskInfo.isIgnoreMode());
olapTableSink.setPartialUpdateInputColumns(isPartialUpdate, partialUpdateInputColumns);
olapTableSink.complete(analyzer);

4
fe/fe-core/src/main/java/org/apache/doris/task/LoadTaskInfo.java
View File

@ -125,6 +125,10 @@ public interface LoadTaskInfo {
return false;
}
default boolean isIgnoreMode() {
return false;
}
class ImportColumnDescs {
public List<ImportColumnDesc> descs = Lists.newArrayList();
public boolean isColumnDescsRewrited = false;

8
fe/fe-core/src/main/java/org/apache/doris/task/StreamLoadTask.java
View File

@ -84,6 +84,7 @@ public class StreamLoadTask implements LoadTaskInfo {
private List<String> hiddenColumns;
private boolean trimDoubleQuotes = false;
private boolean isPartialUpdate = false;
private boolean isIgnoreMode = false;
private int skipLines = 0;
private boolean enableProfile = false;
@ -308,6 +309,10 @@ public class StreamLoadTask implements LoadTaskInfo {
this.memtableOnSinkNode = memtableOnSinkNode;
}
public boolean isIgnoreMode() {
return isIgnoreMode;
}
public static StreamLoadTask fromTStreamLoadPutRequest(TStreamLoadPutRequest request) throws UserException {
StreamLoadTask streamLoadTask = new StreamLoadTask(request.getLoadId(), request.getTxnId(),
request.getFileType(), request.getFormatType(),
@ -444,6 +449,9 @@ public class StreamLoadTask implements LoadTaskInfo {
if (request.isSetMemtableOnSinkNode()) {
this.memtableOnSinkNode = request.isMemtableOnSinkNode();
}
if (request.isSetIgnoreMode()) {
isIgnoreMode = request.isIgnoreMode();
}
}
// used for stream load

8
fe/fe-core/src/test/java/org/apache/doris/planner/OlapTableSinkTest.java
View File

@ -107,7 +107,7 @@ public class OlapTableSinkTest {
new DataProperty(DataProperty.DEFAULT_STORAGE_MEDIUM));
dstTable.getPartitionInfo().setIsMutable(partition.getId(), true);
OlapTableSink sink = new OlapTableSink(dstTable, tuple, Lists.newArrayList(2L), false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false, false);
sink.complete(null);
LOG.info("sink is {}", sink.toThrift());
LOG.info("{}", sink.getExplainString("", TExplainLevel.NORMAL));
@ -144,7 +144,7 @@ public class OlapTableSinkTest {
};
OlapTableSink sink = new OlapTableSink(dstTable, tuple, Lists.newArrayList(p1.getId()), false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false, false);
try {
sink.complete(null);
} catch (UserException e) {
@ -169,7 +169,7 @@ public class OlapTableSinkTest {
};
OlapTableSink sink = new OlapTableSink(dstTable, tuple, Lists.newArrayList(unknownPartId), false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false, false);
sink.complete(null);
LOG.info("sink is {}", sink.toThrift());
LOG.info("{}", sink.getExplainString("", TExplainLevel.NORMAL));
@ -206,7 +206,7 @@ public class OlapTableSinkTest {
};
OlapTableSink sink = new OlapTableSink(dstTable, tuple, Lists.newArrayList(p1.getId()), false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false);
sink.init(new TUniqueId(1, 2), 3, 4, 1000, 1, false, false, false);
try {
sink.complete(null);
} catch (UserException e) {

1
gensrc/proto/descriptors.proto
View File

@ -67,5 +67,6 @@ message POlapTableSchemaParam {
optional bool partial_update = 7;
repeated string partial_update_input_columns = 8;
optional bool is_strict_mode = 9 [default = false];
optional bool is_unique_key_ignore_mode = 10 [default = false];
};

1
gensrc/proto/olap_file.proto
View File

@ -254,6 +254,7 @@ message TabletSchemaPB {
repeated string partial_update_input_columns = 21;
optional bool enable_single_replica_compaction = 22 [default=false];
optional bool skip_write_index_on_load = 23 [default=false];
optional bool is_unique_key_ignore_mode = 24 [default=false];
}
enum TabletStatePB {

1
gensrc/thrift/Descriptors.thrift
View File

@ -230,6 +230,7 @@ struct TOlapTableSchemaParam {
8: optional bool is_partial_update
9: optional list<string> partial_update_input_columns
10: optional bool is_strict_mode = false;
11: optional bool is_unique_key_ignore_mode = false;
}
struct TTabletLocation {

1
gensrc/thrift/FrontendService.thrift
View File

@ -628,6 +628,7 @@ struct TStreamLoadPutRequest {
// only valid when file type is CSV
52: optional i8 escape
53: optional bool memtable_on_sink_node;
54: optional bool ignore_mode = false
}
struct TStreamLoadPutResult {

36
regression-test/data/insert_p0/insert_ignore.out Normal file
View File

@ -0,0 +1,36 @@
-- This file is automatically generated. You should know what you did if you want to edit this
-- !origin_data --
1 kevin 18 shenzhen 400
2 bob 20 beijing 500
3 alice 22 shanghai 600
4 jack 24 hangzhou 700
5 tom 26 guanzhou 800
-- !after_insert_ignore --
1 kevin 18 shenzhen 400
2 bob 20 beijing 500
3 alice 22 shanghai 600
4 jack 24 hangzhou 700
5 tom 26 guanzhou 800
10 alex 28 shenzhen 1111
20 leo 30 beijing 2222
30 sam 32 shanghai 3333
40 Ruth 34 hangzhou 4444
50 cynthia 36 guanzhou 8000
-- !origin_data --
1 1
-- !delete_a_row --
-- !after_insert_ignore --
1 3
-- !sql --
1 10
-- !delete_a_row --
-- !after_insert_ignore --
1 1

10
regression-test/data/unique_with_mow_p0/ignore_mode.csv Normal file
View File

@ -0,0 +1,10 @@
1,"kevin",18,"shenzhen",4000
10,"alex",28,"shenzhen",1111
2,"bob",20,"beijing",5000
20,"leo",30,"beijing",2222
30,"sam",32,"shanghai",3333
3,"alice",22,"shanghai",6000
4,"jack",24,"hangzhou",7000
40,"Ruth",34,"hangzhou",4444
5,"tom",26,"guanzhou",8000
50,"cynthia",36,"guanzhou",8000
1 1 kevin 18 shenzhen 4000
2 10 alex 28 shenzhen 1111
3 2 bob 20 beijing 5000
4 20 leo 30 beijing 2222
5 30 sam 32 shanghai 3333
6 3 alice 22 shanghai 6000
7 4 jack 24 hangzhou 7000
8 40 Ruth 34 hangzhou 4444
9 5 tom 26 guanzhou 8000
10 50 cynthia 36 guanzhou 8000

2
regression-test/data/unique_with_mow_p0/ignore_mode2.csv Normal file
View File

@ -0,0 +1,2 @@
1,4000
2,5000
1 1 4000
2 2 5000

20
regression-test/data/unique_with_mow_p0/test_ignore_mode.out Normal file
View File

@ -0,0 +1,20 @@
-- This file is automatically generated. You should know what you did if you want to edit this
-- !origin_data --
1 kevin 18 shenzhen 400
2 bob 20 beijing 500
3 alice 22 shanghai 600
4 jack 24 hangzhou 700
5 tom 26 guanzhou 800
-- !after_ignore_mode_stream_load --
1 kevin 18 shenzhen 400
2 bob 20 beijing 500
3 alice 22 shanghai 600
4 jack 24 hangzhou 700
5 tom 26 guanzhou 800
10 "alex" 28 "shenzhen" 1111
20 "leo" 30 "beijing" 2222
30 "sam" 32 "shanghai" 3333
40 "Ruth" 34 "hangzhou" 4444
50 "cynthia" 36 "guanzhou" 8000

132
regression-test/suites/insert_p0/insert_ignore.groovy Normal file
View File

@ -0,0 +1,132 @@
// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied. See the License for the
// specific language governing permissions and limitations
// under the License.
suite("test_insert_ignore") {
def tableName = "test_insert_ignore1"
sql """ DROP TABLE IF EXISTS ${tableName} FORCE;"""
sql """
CREATE TABLE ${tableName} (
`id` int(11) NULL,
`name` varchar(10) NULL,
`age` int(11) NULL DEFAULT "20",
`city` varchar(10) NOT NULL DEFAULT "beijing",
`balance` decimalv3(9, 0) NULL
) ENGINE = OLAP UNIQUE KEY(`id`)
COMMENT 'OLAP' DISTRIBUTED BY HASH(`id`) BUCKETS 1
PROPERTIES (
"replication_allocation" = "tag.location.default: 1",
"storage_format" = "V2",
"enable_unique_key_merge_on_write" = "true",
"light_schema_change" = "true",
"disable_auto_compaction" = "false",
"enable_single_replica_compaction" = "false"
);
"""
sql """insert into ${tableName} values
(1,"kevin",18,"shenzhen",400),
(2,"bob",20,"beijing",500),
(3,"alice",22,"shanghai",600),
(4,"jack",24,"hangzhou",700),
(5,"tom",26,"guanzhou",800);"""
qt_origin_data "select * from ${tableName} order by id;"
// some rows are with existing keys, some are not
sql """insert ignore into ${tableName} values
(1,"kevin",18,"shenzhen",4000),
(10,"alex",28,"shenzhen",1111),
(2,"bob",20,"beijing",5000),
(20,"leo",30,"beijing",2222),
(30,"sam",32,"shanghai",3333),
(3,"alice",22,"shanghai",6000),
(4,"jack",24,"hangzhou",7000),
(40,"Ruth",34,"hangzhou",4444),
(5,"tom",26,"guanzhou",8000),
(50,"cynthia",36,"guanzhou",8000);"""
qt_after_insert_ignore "select * from ${tableName} order by id;"
sql """ DROP TABLE IF EXISTS ${tableName};"""
def tableName2 = "test_insert_ignore2"
sql """ DROP TABLE IF EXISTS ${tableName2} FORCE; """
sql """CREATE TABLE IF NOT EXISTS ${tableName2} (
`uid` BIGINT NULL,
`v1` BIGINT NULL
)
UNIQUE KEY(uid)
DISTRIBUTED BY HASH(uid) BUCKETS 1
PROPERTIES (
"enable_unique_key_merge_on_write" = "true",
"replication_num" = "1"
);"""
sql "insert into ${tableName2} values(1,1);"
qt_origin_data "select * from ${tableName2} order by uid;"
sql "insert into ${tableName2}(uid, v1, __DORIS_DELETE_SIGN__) values(1, 2, 1);"
qt_delete_a_row "select * from ${tableName2} order by uid;"
sql "insert ignore into ${tableName2} values(1,3);"
qt_after_insert_ignore "select * from ${tableName2} order by uid;"
sql "insert into ${tableName2} values(1,10);"
qt_sql "select * from ${tableName2} order by uid;"
sql "insert into ${tableName2}(uid, v1, __DORIS_DELETE_SIGN__) values(1, 1, 1);"
qt_delete_a_row "select * from ${tableName2} order by uid;"
sql "insert ignore into ${tableName2} values(1,1);"
qt_after_insert_ignore "select * from ${tableName2} order by uid;"
sql """ DROP TABLE IF EXISTS ${tableName2}; """
// test illigal cases
def tableName3 = "test_insert_ignore3"
sql """ DROP TABLE IF EXISTS ${tableName3} FORCE; """
sql """CREATE TABLE IF NOT EXISTS ${tableName3} (
`uid` BIGINT NULL,
`v1` BIGINT NULL
) UNIQUE KEY(uid)
DISTRIBUTED BY HASH(uid) BUCKETS 1
PROPERTIES (
"enable_unique_key_merge_on_write" = "false",
"replication_num" = "1"
);"""
sql "insert into ${tableName3} values(1,1);"
test {
sql "insert ignore into ${tableName3} values(1,3);"
exception "ignore mode can only be enabled if the target table is a unique table with merge-on-write enabled."
}
def tableName4 = "test_insert_ignore4"
sql """ DROP TABLE IF EXISTS ${tableName4} FORCE; """
sql """CREATE TABLE IF NOT EXISTS ${tableName4} (
`uid` BIGINT NULL,
`v1` BIGINT NULL
) UNIQUE KEY(uid)
DISTRIBUTED BY HASH(uid) BUCKETS 1
PROPERTIES (
"enable_unique_key_merge_on_write" = "true",
"replication_num" = "1",
"function_column.sequence_col" = 'v1'
);"""
sql "insert into ${tableName4} values(1,1);"
test {
sql "insert ignore into ${tableName4} values(1,3);"
exception "ignore mode can't be used if the target table has sequence column, but table[${tableName4}] has sequnce column."
}
}

112
regression-test/suites/unique_with_mow_p0/test_ignore_mode.groovy Normal file
View File

@ -0,0 +1,112 @@
// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied. See the License for the
// specific language governing permissions and limitations
// under the License.
suite("test_mow_table_ignore_mode") {
def tableName = "test_mow_table_ignore_mode1"
sql """ DROP TABLE IF EXISTS ${tableName} FORCE;"""
sql """
CREATE TABLE ${tableName} (
`id` int(11) NULL,
`name` varchar(10) NULL,
`age` int(11) NULL DEFAULT "20",
`city` varchar(10) NOT NULL DEFAULT "beijing",
`balance` decimalv3(9, 0) NULL
) ENGINE = OLAP UNIQUE KEY(`id`)
COMMENT 'OLAP' DISTRIBUTED BY HASH(`id`) BUCKETS 1
PROPERTIES (
"replication_allocation" = "tag.location.default: 1",
"storage_format" = "V2",
"enable_unique_key_merge_on_write" = "true",
"light_schema_change" = "true",
"disable_auto_compaction" = "false",
"enable_single_replica_compaction" = "false"
);
"""
sql """insert into ${tableName} values
(1,"kevin",18,"shenzhen",400),
(2,"bob",20,"beijing",500),
(3,"alice",22,"shanghai",600),
(4,"jack",24,"hangzhou",700),
(5,"tom",26,"guanzhou",800);"""
qt_origin_data "select * from ${tableName} order by id;"
// some rows are with existing keys, some are not
streamLoad {
table "${tableName}"
set 'column_separator', ','
set 'format', 'csv'
set 'columns', 'id,name,age,city,balance'
set 'ignore_mode', 'true'
file 'ignore_mode.csv'
time 10000 // limit inflight 10s
}
sql "sync"
qt_after_ignore_mode_stream_load "select * from ${tableName} order by id;"
sql """ DROP TABLE IF EXISTS ${tableName};"""
// test illegal case
def tableName2 = "test_mow_table_ignore_mode2"
sql """ DROP TABLE IF EXISTS ${tableName2} FORCE;"""
sql """
CREATE TABLE ${tableName2} (
`id` int(11) NULL,
`name` varchar(10) NULL,
`age` int(11) NULL DEFAULT "20",
`city` varchar(10) NOT NULL DEFAULT "beijing",
`balance` decimalv3(9, 0) NULL
) ENGINE = OLAP UNIQUE KEY(`id`)
COMMENT 'OLAP' DISTRIBUTED BY HASH(`id`) BUCKETS 1
PROPERTIES (
"replication_allocation" = "tag.location.default: 1",
"storage_format" = "V2",
"enable_unique_key_merge_on_write" = "true",
"light_schema_change" = "true",
"disable_auto_compaction" = "false",
"enable_single_replica_compaction" = "false"
);"""
sql """insert into ${tableName2} values
(1,"kevin",18,"shenzhen",400),
(2,"bob",20,"beijing",500),
(3,"alice",22,"shanghai",600),
(4,"jack",24,"hangzhou",700),
(5,"tom",26,"guanzhou",800);"""
// some rows are with existing keys, some are not
streamLoad {
table "${tableName2}"
set 'column_separator', ','
set 'format', 'csv'
set 'columns', 'id,balance'
set 'partial_columns', 'true'
set 'ignore_mode', 'true'
file 'ignore_mode.csv'
time 10000 // limit inflight 10s
check {result, exception, startTime, endTime ->
assertTrue(exception == null)
def json = parseJson(result)
assertEquals("Fail", json.Status)
assertTrue(json.Message.contains("ignore mode can't be used in partial update."))
}
}
}