fengttt

We believe https://github.com/matrixorigin/matrixone/issues/6956 should resolve this issue.

Postpone to 0.8. As I understand, we want to make sure we never construct SQL from what a user typed in. Just switch all execution to use param binding. So...

``` select year(d_datekey) as year, s_city, p_brand, sum(lo_revenue) - sum(lo_supplycost) as profit, c_region, s_nation, p_category from lineorder join dates on lo_orderdate = d_datekey join customer on lo_custkey = c_custkey join...

According to @badboynt1 for 1G data, the query will produce 5 million rows. The main memory consumption is the hash agg. which should have used roughly 500MB memory. For 100G,...

This cannot be fixed/tested by config tweaks. Must fix the planner and executor.

#7891 is done. So this one should be done as well?

Not a bug. This is a superset bug (it generated wrong sql) and MySQL bug (it should have rejected wrong sql). That said, we may consider to add a mysql_compatibility_mode...

XuPeng did a memory profiling, it is mainly in S3Writer.WriteS3Batch. Quick look at the code shows the code need major rewrite.

> no fix. What does this mean ... You don't know how to fix or you say we should not fix?