Hi,
+1 with Jialin Qiao.
I think the ongoing query refactor work will cover the following points:
1) refactor the ordinary query, by
a) changing all operators in query plans to be batch operators
b) replanning the reading of data from storage, taking into consideration
the interactions both between the sequence and overflow data and within
overflow data
2) refactor the aggregate and groupBy queries, by
a) abstracting new aggregate operators or interfaces to enable both point,
page and chunk level aggregate reading
b) replanning the reading of data from storage, taking into consideration
the interactions both between the sequence and overflow data and within
overflow data
Lastly, sorry @atoiLiu I don't quite get what you mean. `Limit&Offset` query is
already realized I think.
Best,
Lei Rui
On 12/10/2019 22:37，atoiLiu<atoi...@163.com> wrote：
hi,
Offset command is sometimes found in SQL, so hopefully it will also provide a
jump query
在 2019年12月10日，下午7:16，Jialin Qiao <qiaojia...@apache.org> 写道：
Hi,
Code refactoring is inevitable when building a large system. The read/write
of TsFile, storage engine of the server have been refactored. Now, it's
time to refactor the query engine in the server.
Currently, the query is in a tuple-at-a-time manner. In the meantime, the
interface is in chaos.
To improve the query speed, a batch-at-a-time(vectorization) iteration is
needed.
I have opened a branch with Lei Rui: f_batch_reader. The existing
interfaces are simplified and IBatchReader is added. Welcome to work on
this branch for query optimization.
Thanks,
—————————————————
Jialin Qiao
School of Software, Tsinghua University
乔嘉林
清华大学 软件学院
