`date_time_v2` will check scale when constructed datatimev2:
```
LOG(FATAL) << fmt::format("Scale {} is out of bounds", scale);
```
This [PR](https://github.com/apache/doris/pull/15510) has fixed this issue, but parquet does not use constructor to create `TypeDescriptor`, leading the `scale = -1` when reading datetimev2 data.
The origin scan pools are in exec_env.
But after enable new_load_scan_node by default, the scan pool in exec_env is no longer used.
All scan task will be submitted to the scan pool in scanner_scheduler.
BTW, reorganize the scan pool into 3 kinds:
local scan pool
For olap scan node
remote scan pool
For file scan node
limited scan pool
For query which set cpu resource limit or with small limit clause
TODO:
Use bthread to unify all IO task.
Some trivial issues:
fix bug that the memtable flush size printed in log is not right
Add RuntimeProfile param in VScanner
Support new table value function `iceberg_meta("table" = "ctl.db.tbl", "query_type" = "snapshots")`
we can use the sql `select * from iceberg_meta("table" = "ctl.db.tbl", "query_type" = "snapshots")` to get snapshots info of a table. The other iceberg metadata will be supported later when needed.
One of the usage:
Before we use following sql to time travel:
`select * from ice_table FOR TIME AS OF "2022-10-10 11:11:11"`;
`select * from ice_table FOR VERSION AS OF "snapshot_id"`;
we can use the snapshots metadata to get the `committed time` or `snapshot_id`,
and then, we can use it as the time or version in time travel clause
The main purpose of this pr is to import `fileCache` for lakehouse reading remote files.
Use the local disk as the cache for reading remote file, so the next time this file is read,
the data can be obtained directly from the local disk.
In addition, this pr includes a few other minor changes
Import File Cache:
1. The imported `fileCache` is called `block_file_cache`, which uses lru replacement policy.
2. Implement a new FileRereader `CachedRemoteFilereader`, so that the logic of `file cache` is hidden under `CachedRemoteFilereader`.
Other changes:
1. Add a new interface `fs()` for `FileReader`.
2. `IOContext` adds some statistical information to count the situation of `FileCache`
Co-authored-by: Lightman <31928846+Lchangliang@users.noreply.github.com>
For outer join / right outer join / right semi join, when HashJoinNode::pull->process_data_in_hashtable outputs a block, it will output all rows of a key in the hash table into a block, and the output of a key is completed After that, it will check whether the block size exceeds the batch size, and if it exceeds, the output will be terminated.
If a key has 2000w+ rows, memory overflow will occur when the subsequent block operations on the 2000w+ rows are performed.
Now in ScannerContext::push_back_scanner_and_reschedule, _num_running_scanners-- is before _num_scheduling_ctx++.
InPipScannerContext::can_finish, we check _num_running_scanners == 0 && _num_scheduling_ctx == 0 without obtaining _transfer_lock.
In follow case, PipScannerContext::can_finish will return wrong result.
_num_running_scanners--
Check _num_running_scanners == 0 && _num_scheduling_ctx == 0` return true.
_num_scheduling_ctx++
So, we can set _num_running_scanners-- in the last of this func.
Describe your changes.
PipScannerContext::get_block_from_queue not block.
Set _num_running_scanners-- in the last of ScannerContext::push_back_scanner_and_reschedule.
A deleted file may belong to multiple data files. Each data file will read a full amount of deleted files,
so a deleted file may be read repeatedly. The deleted files can be cached, and multiple data files
can reuse the first read content.
The performance is improved by 60% in the case of single thread, and by 30% in the case of multithreading.
1. the agg function without distinct keyword should be a "merge" funcion in threePhaseAggregateWithDistinct
2. use aggregateParam.aggMode.consumeAggregateBuffer instead of aggregateParam.aggPhase.isGlobal() to indicate if a agg function is a "merge" function
3. add an AvgDistinctToSumDivCount rule to support avg(distinct xxx) in some case
4. AggregateExpression's nullable method should call inner function's nullable method.
5. add a bind slot rule to bind pattern "logicalSort(logicalHaving(logicalProject()))"
6. don't remove project node in PhysicalPlanTranslator
7. add a cast to bigint expr when count( distinct datelike type )
8. fallback to old optimizer if bitmap runtime filter is enabled.
9. fix exchange node mem leak
* [feature-wip](inverted index)inverted index api: reader
* [feature-wip](inverted index) Fulltext query syntax with MATCH/MATCH_ALL/MATCH_ALL
* [feature-wip](inverted index) Adapt to index meta
* [enhance] add more metrics
* [enhance] add fulltext match query check for column type and index parser
* [feature-wip](inverted index) Support apply inverted index in compound predicate which except leaf node of and node
**Optimize**
PR #14470 has used `Expr` to filter delete rows to match current data file,
but the rows in the delete file are [sorted by file_path then position](https://iceberg.apache.org/spec/#position-delete-files)
to optimize filtering rows while scanning, so this PR remove `Expr` and use binary search to filter delete rows.
In addition, delete files are likely to be encoded in dictionary, it's time-consuming to decode `file_path`
columns into `ColumnString`, so this PR use `ColumnDictionary` to read `file_path` column.
After testing, the performance of iceberg v2's MOR is improved by 30%+.
**Fix Bug**
Lazy-read-block may not have the filter column, if the whole group is filtered by `Expr`
and the batch_eof is generated from next batch.
Current column default value is used only for load task. But in the case of Iceberg schema change,
query task is also possible to read the default value for columns not exist in old schema.
This pr is to support default value for query task.
Manually tested the broker load and external emr regression cases.
Current bitmap index only can apply pushed down predicates which in AND conditions. When predicates in OR conditions and other complex compound conditions, it will not be pushed down to the storage layer, this leads to read more data.
Based on that situation, this pr will do:
1. this pr in order to support bitmap index apply compound predicates, query sql like:
select * from tb where a > 'hello' or b < 100;
select * from tb where a > 'hello' or b < 100 or c > 'ok';
select * from tb where (a > 'hello' or b <100) and (a < 'world' or b > 200);
select * from tb where (not a> 'hello') or b < 100;
...
above sql，column a and b and c has created bitmap_index.
2. this optimization can reduce reading data by index
3. set config enable_index_apply_compound_predicates to use this optimization
1. Fix 1 bug:
Throw null pointer exception when reading data after the reader reaches the end of file, so should return directly when `_do_lazy_read` read no data.
2. Optimize code:
Remove unused parameters.
3. Fix regression test
