* table function node enhancement
* also avoid copy for non-vec table function node
* fix table function node output slots calculation while lateral view involves subquery
Co-authored-by: cambyzju <zhuxiaoli01@baidu.com>
There are currently many types of ScanNodes in Doris. And most of the logic of these ScanNodes is the same, including:
Runtime filter
Predicate pushdown
Scanner generation and scheduling
So I intend to unify the common logic of all ScanNodes.
Different data sources only need to implement different Scanners for data access.
So that the future optimization for scan can be applied to the scan of all data sources,
while also reducing the code duplication.
This PR mainly adds 4 new class:
VScanner
All Scanners' parent class. The subclasses can inherit this class to implement specific data access methods.
VScanNode
The unified ScanNode, and is responsible for common logic including RuntimeFilter, predicate pushdown, Scanner generation and scheduling.
ScannerContext
ScannerContext is responsible for recording the execution status
of a group of Scanners corresponding to a ScanNode.
Including how many scanners are being scheduled, and maintaining
a producer-consumer blocks queue between scanners and scan nodes.
ScannerContext is also the scheduling unit of ScannerScheduler.
ScannerScheduler schedules a ScannerContext at a time,
and submits the Scanners to the scanner thread pool for data scanning.
ScannerScheduler
Unified responsible for all Scanner scheduling tasks
Test:
This work is still in progress and default is disabled.
I tested it with jmeter with 50 concurrency, but currently the scanner is just return without data.
The QPS can reach about 9000.
I can't compare it to origin implement because no data is read for now. I will test it when new olap scanner is ready.
Co-authored-by: morningman <morningman@apache.org>
1. Spark can set the timestamp precision by the following configuration:
spark.sql.parquet.outputTimestampType = INT96(NANOS), TIMESTAMP_MICROS, TIMESTAMP_MILLIS
DATETIME V1 only keeps the second precision, DATETIME V2 keeps the microsecond precision.
2. If using DECIMAL V2, the BE saves the value as decimal128, and keeps the precision of decimal as (precision=27, scale=9). DECIMAL V3 can maintain the right precision of decimal
Currently we use rapidjson to parse json document, It's fast but not fast enough compare to simdjson.And I found that the simdjson has a parsing front-end called simdjson::ondemand which will parse json when accessing fields and could strip the field token from the original document, using this feature we could reduce the cost of string copy(eg. we convert everthing to a string literal in _write_data_to_column by sprintf, I saw a hotspot from the flamegrame in this function, using simdjson::to_json_string will strip the token(a string piece) which is std::string_view and this is exactly we need).And second in _set_column_value we could iterate through the json document by for (auto field: object_val) {xxx}, this is much faster than looking up a field by it's field name like objectValue.FindMember("k1").The third optimization is the at_pointer interface simdjson provided, this could directly get the json field from original document.
In our origin design, we calc delete bitmap in publish txn, and this operation
will cost too much time as it will load segment data and lookup row key in pre
rowset and segments.And publish version task should run in order, so it'll lead
to timeout in publish_txn.
In this pr, we seperate delete_bitmap calculation to tow part, one of it will be
done in flush mem table, so this work can run parallel. And we calc final
delete_bitmap in publish_txn, get a rowset_id set that should be included and
remove rowsets that has been compacted, the rowset difference between memtable_flush
and publish_txn is really small so publish_txn become very fast.In our test,
publish_txn cost about 10ms.
Co-authored-by: yixiutt <yixiu@selectdb.com>
During load process, the same operation are performed on all replicas such as sort and aggregation,
which are resource-intensive.
Concurrent data load would consume much CPU and memory resources.
It's better to perform write process (writing data into MemTable and then data flush) on single replica
and synchronize data files to other replicas before transaction finished.
1. Fix a bug that query large column table may cause infinite loop
2. Optimize the query logic with limit, for the case where the limit value is relatively small, reduce the parallelism of the scanner, reduce unnecessary resource consumption, and increase the number of similar queries that the system can carry at the same time, and increase the query speed by more than 60%
* fix infinite loop when reading wide table
When a wide table is read, the 1st batch would be exceed raw_bytes_threshold,
so Scanner should read at least 1 row.
Actually, we should adjust batch size automatically to reduce memoery usage.
Hash join node adds three new attributes.
The following will take an SQL as an example to illustrate the meaning of these three attributes
```
select t1. a from t1 left join t2 on t1. a=t2. b;
```
1. vOutputTupleDesc：Tuple2(a'')
2. vIntermediateTupleDescList: Tuple1(a', b'<nullable>)
2. vSrcToOutputSMap: <Tuple1(a'), Tuple2(a'')>
The slot in intermediatetuple corresponds to the slot in output tuple one by one through the expr calculation of the left child in vsrctooutputsmap.
This code mainly merges the contents of two PRs:
1. [fix](vectorized) Support outer join for vectorized exec engine (https://github.com/apache/doris/pull/10323)
2. [Fix](Join) Fix the bug of outer join function under vectorization #9954
The following is the specific description of the first PR
In a vectorized scenario, the query plan will generate a new tuple for the join node.
This tuple mainly describes the output schema of the join node.
Adding this tuple mainly solves the problem that the input schema of the join node is different from the output schema.
For example:
1. The case where the null side column caused by outer join is converted to nullable.
2. The projection of the outer tuple.
The following is the specific description of the second PR
This pr mainly fixes the following problems:
1. Solve the query combined with inline view and outer join. After adding a tuple to the join operator, the position of the `tupleisnull` function is inconsistent with the row storage. Currently the vectorized `tupleisnull` will be calculated in the HashJoinNode.computeOutputTuple() function.
2. Column nullable property error problem. At present, once the outer join occurs, the column on the null-side side will be planned to be nullable in the semantic parsing stage.
For example：
```
select * from (select a as k1 from test) tmp right join b on tmp.k1=b.k1
```
At this time, the nullable property of column k1 in the `tmp` inline view should be true.
In the vectorized code, the virtual `tableRef` of tmp will be used in constructing the output tuple of HashJoinNode (specifically, the function HashJoinNode.computeOutputTuple()). So the **correctness** of the column nullable property of this tableRef is very important.
In the above case, since the tmp table needs to perform a right join with the b table, as a null-side tmp side, it is necessary to change the column attributes involved in the tmp table to nullable.
In non-vectorized code, since the virtual tableRef tmp is not used at all, it uses the `TupleIsNull` function in `outputsmp` to ensure data correctness.
That is to say, the a column of the original table test is still non-null, and it does not affect the correctness of the result.
The vectorized nullable attribute requirements are very strict.
Outer join will change the nullable attribute of the join column, thereby changing the nullable attribute of the column in the upper operator layer by layer.
Since FE has no mechanism to modify the nullable attribute in the upper operator tuple layer by layer after the analyzer.
So at present, we can only preset the attributes before the lower join as nullable in the analyzer stage in advance, so as to avoid the problem.
(At the same time, be also wrote some evasive code in order to deal with the problem of null to non-null.)
Co-authored-by: EmmyMiao87
Co-authored-by: HappenLee
Co-authored-by: morrySnow
Co-authored-by: EmmyMiao87 <522274284@qq.com>
