It should be ensured that the obtained versions are continuous when calculate delete bitmap calculations in publish.
The remaining NOTREADY tablet in the schema change failure should be dropped.
When a rowset was deleted, the delete bitmap cannot be deleted until there are no read requests to use the rowset.
for example:
```sql
with a as (with a as (select * from t) select * from a) select * from a;
with a as (select * from t1), b as (with a as (select * from a) select * from a) select * from b;
```
Fix select table tablet not effective, table distributed by random.
If tabletID specified in query does not exist in this partition, skip scan partition.
fix shrink char column in map/struct
before we has char with specific length defined in map or struct field
we select map or struct , the char column in which one has been padding if we just insert less than specific length chars
but in mysql here just show inserted chars not padding specific length chars
---------
Co-authored-by: yiguolei <676222867@qq.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
extract slot and literal in comparison predicate. infer new one by equals predicates.
use TypeCoercion to add cast on new comparison predicate to ensure it is correct.
This reverts "[Fix](Nereids) Add cast comparison with slot reference when inferring predicate (#21171)"
commit 58f2593ba1b65713e7b3c1ed39fc84be8cc3ff2c.
Problem:
When join estimation, upper join output slot statistic ndv would go wrong
Example:
we have two table:
tableA (a1[ndv = 10.0]) tableB(b1[ndv = 0.0], b2[ndv = 10.0])
tableA left join tableB on A.a1 = B.b1. which B.b1 with ndv zero.
the problem is after join estimation, B.b2 changed to 1.0.
Reason:
When estimating outer join, we can assume it behave like inner join. But we estimation then like inner join do
Solved:
When estimation outer join, output slot would update seperatly.
File cache will return the reading data only when its writer has finished the write operation and close the segment file. However there's no need to call fdatasync when closing segment file. fdatasync takes a lot of time, and severely impact on cache performance.
