remove vectorAnalyze and computeOutputColumn function on Expr
remove vectorOpcode and outputColumn attribute on Expr
remove useless static LOG object on some Expr class
this PR is from @sjyango work in #32326,
wants merge #32326 into master branch, but it's draft and not maintain long time. so have this new PR.
Co-authored-by: sjyango <sjyang2022@zju.edu.cn>
This PR supports a Table Value Function called `Query`. He can push a query directly to the catalog source for execution by specifying `catalog` and `query` without parsing by Doris. Doris only receives the results returned by the query.
Currently only JDBC Catalog is supported.
Example:
```
Doris > desc function query('catalog' = 'mysql','query' = 'select count(*) as cnt from test.test');
+-------+--------+------+------+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+-------+--------+------+------+---------+-------+
| cnt | BIGINT | Yes | true | NULL | NONE |
+-------+--------+------+------+---------+-------+
Doris > select * from query('catalog' = 'mysql','query' = 'select count(*) as cnt from test.test');
+----------+
| cnt |
+----------+
| 30000000 |
+----------+
```
when create MTMV,we will save current ctl and db.
when refresh MTMV,will create an ConnectContext, and set same ctl, db to ctx
when db,ctx dropped, task will be failed.
But sometimes deleting a db does not actually have an impact, so changing it to not directly fail. If refreshing the data does cause an error, then giving the user an error message
this pr is a quick solution, but not complete.
runtime filter on NestLoopJoin suffers this bug even without PROJECT_OTHER_JOIN_CONDITION rule.
for example, when enable Min_Max Runtime filter, the target Expression is n_regionkey, but it should be "n_regionkey - 28"
explain
select n_nationkey, nrkey
from (select n_regionkey -28 nrkey, n_nationkey from nation) T
join region on nrkey > r_regionkey;
we will refactor RuntimeFilterGenerator to completely solve this issue in following pr.
we do derive stats job eager to avoid un derive stats due to merge group and optimize group
consider:
we have two groups burned by order: G1 and G2
then we have job by order derive G2, optimize group expression in G2,
derive G1, optimize group expression in G1
if G1 merged into G2, then we maybe generated job optimize group G2 before derive G1
in this case, we will do get stats from G1's child before derive G1's child stats
then we will meet NPE in CostModel.
1. we should check output set, since we will remove top project and
result output size will diff with its child output size if there are
dup slot in result list
2. should not return null, instead we should return rewritten plan
itself, because we will use return result in many place and do not
check null at all
* Support statistics
* - Fix Failed task not showing up in the task list
- Task metadata add jobName
- Fix Finished job clear time error
- Job metadata add successCount, failedCount, totalTaskCount
* add test
pick from master #33979
commit id 65fb7d43b7e838c48502d4e8a69e2541dc73aa88
This PR:
1. add a new Command type: UnsupportedCommand to handle the statement only support parse but could not execute.
2. support syntax about set and use
3. add keyword VAULT to follow legacy planner
TODO
1. support all statment syntax in Nereids
