…avior (#38847)
https://github.com/apache/doris/pull/38847
## Proposed changes
There are two issues here. First, the results of casting are
inconsistent between FE and BE .
```
FE
mysql [(none)]>select cast('3.000' as int);
+----------------------+
| cast('3.000' as INT) |
+----------------------+
| 3 |
+----------------------+
mysql [(none)]>set debug_skip_fold_constant = true;
BE
mysql [(none)]>select cast('3.000' as int);
+----------------------+
| cast('3.000' as INT) |
+----------------------+
| NULL |
+----------------------+
```
The second issue is that casting on BE converts '3.0' to null. Here, the
casting logic for FE and BE has been unified
<!--Describe your changes.-->
## Proposed changes
Issue Number: close #xxx
<!--Describe your changes.-->
---------
Co-authored-by: Xinyi Zou <zouxinyi02@gmail.com>
pick (#39992)
BE node was killed by OOM-killer when use multi table load under high
concurrency with a large number of tables(128 concurrency and every
concurrency load 200 tables).
This pr restricts the multi tables load memory under this issue. If
memory reaches hard limit, new task will be rejected and return
directly.
pick https://github.com/apache/doris/pull/37062
1. revert https://github.com/apache/doris/pull/25097. we decide to rely
on OS. not maintain independent tzdata anymore to keep result
consistency
2. refactor timezone load. removed rwlock.
before:
```sql
mysql [optest]>select count(convert_tz(d, 'Asia/Shanghai', 'America/Los_Angeles')), count(convert_tz(dt, 'America/Los_Angeles', '+00:00')) from dates;
+-------------------------------------------------------------------------------------+--------------------------------------------------------+
| count(convert_tz(cast(d as DATETIMEV2(6)), 'Asia/Shanghai', 'America/Los_Angeles')) | count(convert_tz(dt, 'America/Los_Angeles', '+00:00')) |
+-------------------------------------------------------------------------------------+--------------------------------------------------------+
| 16000000 | 16000000 |
+-------------------------------------------------------------------------------------+--------------------------------------------------------+
1 row in set (6.88 sec)
```
now:
```sql
mysql [optest]>select count(convert_tz(d, 'Asia/Shanghai', 'America/Los_Angeles')), count(convert_tz(dt, 'America/Los_Angeles', '+00:00')) from dates;
+-------------------------------------------------------------------------------------+--------------------------------------------------------+
| count(convert_tz(cast(d as DATETIMEV2(6)), 'Asia/Shanghai', 'America/Los_Angeles')) | count(convert_tz(dt, 'America/Los_Angeles', '+00:00')) |
+-------------------------------------------------------------------------------------+--------------------------------------------------------+
| 16000000 | 16000000 |
+-------------------------------------------------------------------------------------+--------------------------------------------------------+
1 row in set (2.61 sec)
```
3. now don't support timezone offset format string like 'UTC+8', like we
already said in
https://doris.apache.org/docs/dev/query/query-variables/time-zone/#usage
4. support case-insensitive timezone parsing in nereids.
5. a bug when parse timezone using nereids. should check DST by input,
but wrongly by now before. now fixed.
doc pr: https://github.com/apache/doris-website/pull/810
Generally speaking, as long as a rowset has a version, it can be
considered not to be in a pending state. However, if the rowset was
created through ingesting binlogs, it will have a version but should
still be considered in a pending state because the ingesting txn has not
yet been committed.
This PR updates the condition for determining the pending state. If a
rowset is COMMITTED, the txn should be allowed to roll back even if a
version exists.
Cherry-pick #36551
## Proposed changes
1. return error when bloom filter allocate memory failed
2. return error when deserialize a block， it may need a lot of memory.
---------
Co-authored-by: yiguolei <yiguolei@gmail.com>
