For broadcast join, only one build fragment instance will build hash table, other fragment instances just receive and throw away build side data, this is waste of memory and cpu.
This PR improve this condition, data stream receiver tells sender that it does not need data from sender, and sender stops sending anydata to it.
1. Remove an exec node method corresponding to a span and replace it with an exec node corresponding to a span;
2. Fix some problems with tracing in pipeline.
This pr does three things:
1. add `delete_existing_files` property for outfile/export. If `delete_existing_files = true`, export/outfile will delete all files under file_path first.
2. add p2 test for export
3. modify docs
The original method signature is Block VExprContext::get_output_block_after_execute_exprs(
const std::vectorvectorized::VExprContext*& output_vexpr_ctxs, const Block& input_block,
Status& status)
It return error status as a out parameter and the block as return value. It has to check the block.rows == 0 and then check error status.
It is not conforming to the convention.
---------
Co-authored-by: yiguolei <yiguolei@gmail.com>
TabletSink and LoadChannel in BE are M: N relationship,
Every once in a while LoadChannel will randomly return its own runtime profile to a TabletSink, so usually all LoadChannel runtime profiles are saved on each TabletSink, and the timeliness of the same LoadChannel profile saved on different TabletSinks is different, and each TabletSink will periodically send fe reports all the LoadChannel profiles saved by itself, and ensures to update the latest LoadChannel profile according to the timestamp.
disallow call new method explicitly
force to use create_shared or create_unique to use shared ptr
placement new is allowed
reference https://abseil.io/tips/42 to add factory method to all class.
I think we should follow this guide because if throw exception in new method, the program will terminate.
---------
Co-authored-by: yiguolei <yiguolei@gmail.com>
Currently, there are some useless includes in the codebase. We can use a tool named include-what-you-use to optimize these includes. By using a strict include-what-you-use policy, we can get lots of benefits from it.
No check mem tracker limit and no cancel task in mem hook, only in Allocator. This helps in clearer analysis of memory issues and reduces performance loss.
PODArray/hash table/arena memory allocation will use Allocator.
Optimize mem limit exceeded log printing
Optimize compilation time
* [Feature](vectorized)(quantile_state): support vectorized quantile state functions
1. now quantile column only support not nullable
2. add up some regression test cases
3. set default enable_quantile_state_type = true
---------
Co-authored-by: spaces-x <weixiang06@meituan.com>
1. introduce a new type `VARIANT` to encapsulate dynamic generated columns for hidding the detail of types and names of newly generated columns
2. introduce a new expression `SchemaChangeExpr` for doing schema change for extensibility
The current distribution model for Doris is as follows:
OlapTableSink seperate the original Block into serveral subblocks of each node(BE) by tablets distribution and distributes subblocks to storage engine of backends, then the storage engine will seperate the subblock into multiple tablets channel and each delta writer will handle partial of the block.
This model causes blocks to be split according to tablets, and the splitting process can be a relatively heavy operation. After splitting, the blocks are distributed to different DeltaWriters (Memtables) through RPCs to TabletChannels. The distribution operation on TabletChannels is also a relatively heavy operation. If the distribution property of the table is RANDOM distribution, then we have the opportunity to distribute the blocks according to the complete block during distribution. The advantage of doing so is to reduce memory copying and improve write locality, similar to appending the entire block to the memtable.
This optimze could save 10% ~ 20% CPU cost of RANDOM distribution table load when enable load_to_single_tablet
This CL mainly changes:
Support specifying csv schema manually in s3/hdfs table valued function
s3 (
'URI' = 'https://bucket1/inventory.dat',
'ACCESS_KEY'= 'ak',
'SECRET_KEY' = 'sk',
'FORMAT' = 'csv',
'column_separator' = '|',
'csv_schema' = 'k1:int;k2:int;k3:int;k4:decimal(38,10)',
'use_path_style'='true'
)
Add new session variable dry_run_query
If set to true, the real query result will not be returned, instead, it will only return the number of returned rows.
mysql> select * from bigtable;
+--------------+
| ReturnedRows |
+--------------+
| 10000000 |
+--------------+
This can avoid large result set transmission time and focus on real execution time of query engine.
For debug and analysis purpose.
1. Make sure all sub types which STRUCT supported work correctly;
2. remove unused variable `_need_validate_data`;
3. lazy init min or max decimal to support nested DecimalV2 column validate;
Co-authored-by: cambyzju <zhuxiaoli01@baidu.com>
1、support stream load with json, csv format for map
2、fix olap convertor when compaction action in map column which has null
3、support select outToFile for map
4、add some regression-test
The background is described in this issue: #15723,
where users used Apache Druid to satisfy such lambada requirements before.
We will not make Doris dropping data not belonged to current time window automatically like Druid,
which is not flexible. We demand a ability to support mutable/immutable partition, the PR works this way:
1. Support mutable property for a partition.
2. The mutable property of a partition is passed from FE to BE in a load procedure
3. If a record's partition is immutable, we mark this row as "un selected" which will not be included in computation of 'max_filter_ratio',
so that data write to immutable partition will be neglected and not cause load failure.
Use Example:
1. Add immutable partition or modify an partition to be immutable:
- alter table test_tbl add [temporary] partition xxx values less than ('xxx') ('mutable' = 'true');
- alter table test_tbl modify partition xx set ('mutable' = 'false');
2. Write 5 records into table, two of then belongs to immutable partition
Now we reuse buffer pool for broadcast shuffle on pipeline engine. This PR ensures that a pipeline with a broadcast shuffle sink will not be scheduled if there are no available buffer in the buffer pool
Support IPV6 in Apache Doris, the main changes are:
1. enable binding to IPV6 address if network priority in config file contains an IPV6 CIDR string
2. BRPC and HTTP support binding to IPV6 address
3. BRPC and HTTP support visiting IPV6 Services
Issue Number: close#16351
Dynamic schema table is a special type of table, it's schema change with loading procedure.Now we implemented this feature mainly for semi-structure data such as JSON, since JSON is schema self-described we could extract schema info from the original documents and inference the final type infomation.This speical table could reduce manual schema change operation and easily import semi-structure data and extends it's schema automatically.
This commit support:
1、Insert + select for struct/map type
2、Json stream load for struct type
3、m[key] function for map type
How to use:
Set the fe config to create table for struct and map type
1、admin set frontend config("enable_struct_type" = "true");
2、admin set frontend config("enable_map_type" = "true");
#16547
Co-authored-by: xy720 <xuyang25@baidu.com>
Co-authored-by: amory <wangqiannan@selectdb.com>
Co-authored-by: cambyzju <zhuxiaoli01@baidu.com>
Co-authored-by: hucheng01 <hucheng01@baidu.com>
Currently, for broadcast shuffle, we serialize a block once and then send it by RPC through multiple channel. After this, we will serialize next block in the same memory for consideration of memory reuse. However, since the RPC is asynchronized, maybe the next block serialization will happen before sending the previous block.
So, in this PR, I use a ref count to identify if the serialized block can be reuse in broadcast shuffle.
BufferControlBlock may block all fragment handle threads leads to be out of work
modify include:
BufferControlBlock cancel after max timeout
StmtExcutor notify be to cancel the fragment when unexcepted occur
more details see issue #16203
