3.3查看JOIN算子执行
本节主要介绍Join, Aggregate等算子的trace的详细信息，从中可分析 出算子是否并行执行，采用具体何种方法执行等信息。
Join算子目前有：
SortMerge Join :主要用于非等值连接，如t1. a > t2.b；
Hash Join：主要用于等值连接，内部又细分为RR， One-Pass， Hybrid三 种方法，如t1.a=t2. a；
NestLoop Join :以上其他方法都失效，走嵌套循环连接；
Complex Join:主要用于包含Or的连接，Or的两端可以是Sort Merge也 可以是Hash Join；
3. 3. 1 sort merge join
BEGIN Join
cnd(0):
cnd(0): t1.i < t2. j
prepare to use sort-merge
Joining sorters created fo
inner join (T0 - T1), rows.
cnd(0) Done(time used: 0.0
等值连接走hash join。hash join有三种，一般情况下会走hybrid hash join
BEGIN Join
cnd(0):
cnd(0): lineitem.L_ORDERKEY = orders. O_ORDERKEY
prepare to use hash join
等值连接走hash join
op buffer size: 16777216, tuple width: 16. op buffer can hold 1048576 rows 算子 buffer 大小 16777216，join 字段每行 16 字节，buffer
可以装1048576行
Begin one-pass hash partitioning: divide 1500000 tuples into 6 parts. mat_buf_size = 2796200 看到'one-pass hash partitioning'代表
走 one-pass hash join
mat partition thread: divide 1500000 tuples into 6 parts, min 241835
tuples, max 258974 tuples, avg 250000 tuples.
Finish one-pass hash partitioning: divide 1500000 tuples into 6 parts. mat_buf_size = 2796200. (time used: 2.104s)
Begin one-pass hash partitioning: divide 6001215 tuples into 6 parts. mat_buf_size = 2796200 对 join 的右边划分
Finish one-pass hash partitioning: divide 6001215 tuples into 6 parts. mat_buf_size = 2796200. (time used: 5.589s)
Finish One-Pass Hash Join preparation: divided each side into 6 partitions
Hash tree is used, size: 16777216
Begin serial rowid merge-sorting: 6001215 rows
对join结果排序，这里serial可以sort可以看出join是串行的
Finish serial rowid merge-sorting: 6001215 rows. (time used: 5.806s) inner join(T0 - T1), using hash join, produced 6001215 rows.
cnd(0) Done(time used: 39.754s)
END Join(time used: 39. 755s)
BEGIN Join
prepare to use hash join
op buffer size: 16777216, tuple width: 16. op buffer can hold 1048576
<