Hi Jialin,

Thanks for your explanation. I would like to implement this issue to understand 
more about the writing interface in IoTDB.

However, due to my time reason recently, I would like to do the delete 
interface first, which I think it is easier to done it. Then I would implement 
the query interface when I get more free time.

Best,
Tsung Han Tsai
________________________________
寄件者: Jialin Qiao <qj...@mails.tsinghua.edu.cn>
寄件日期: 2019年9月3日 下午 04:06
收件者: dev@iotdb.apache.org <dev@iotdb.apache.org>
主旨: Re: Create metadata automatically

Hi,

You are right, for this statement: insert into root.sg1.d1(time, s1) values(1, 
"1").
s1 can be treated as string.

Thanks,
--
Jialin Qiao
School of Software, Tsinghua University

乔嘉林
清华大学 软件学院

> -----原始邮件-----
> 发件人: "Xiangdong Huang" <saint...@gmail.com>
> 发送时间: 2019-09-03 15:32:57 (星期二)
> 收件人: dev@iotdb.apache.org
> 抄送:
> 主题: Re: Create metadata automatically
>
> Hi,
>
> > If the value is none of the (boolean, long, double) type, then it is
> treated as string.
>
> Hm... why the string definition is not  a text which are wrapped by ".
> Otherwise, how to define a string "true"/"false".
>
> Best,
> -----------------------------------
> Xiangdong Huang
> School of Software, Tsinghua University
>
>  黄向东
> 清华大学 软件学院
>
>
> Jialin Qiao <qj...@mails.tsinghua.edu.cn> 于2019年9月3日周二 上午12:08写道:
>
> > Hi,
> >
> > Thanks for reminding, boolean type can be inferred from "true" or "false".
> >
> > Now the configuration parameters are:
> >
> > auto_create_schema=false
> > default_storage_group_level=2
> > default_boolean_encoding=RLE
> > default_long_encoding=RLE
> > default_double_encoding=GORILLA
> > default_string_encoding=PLAIN
> >
> > If the value is none of the (boolean, long, double) type, then it is
> > treated as string.
> >
> > Best,
> > --
> > Jialin Qiao
> > School of Software, Tsinghua University
> >
> > 乔嘉林
> > 清华大学 软件学院
> >
> > > -----原始邮件-----
> > > 发件人: "Xiangdong Huang" <saint...@gmail.com>
> > > 发送时间: 2019-09-03 14:52:57 (星期二)
> > > 收件人: dev@iotdb.apache.org
> > > 抄送:
> > > 主题: Re: Create metadata automatically
> > >
> > > Hi,
> > >
> > > By the way. If so, no "boolean" data type? And why there is no
> > > default_string_encoding.
> > >
> > > Best,
> > > -----------------------------------
> > > Xiangdong Huang
> > > School of Software, Tsinghua University
> > >
> > >  黄向东
> > > 清华大学 软件学院
> > >
> > >
> > > Xiangdong Huang <saint...@gmail.com> 于2019年9月2日周一 下午11:50写道:
> > >
> > > > Hi,
> > > >
> > > > I think it should be called as "schema" rather than "metadata"...
> > > >
> > > > Best,
> > > > -----------------------------------
> > > > Xiangdong Huang
> > > > School of Software, Tsinghua University
> > > >
> > > >  黄向东
> > > > 清华大学 软件学院
> > > >
> > > >
> > > > Jialin Qiao <qj...@mails.tsinghua.edu.cn> 于2019年9月2日周一 下午11:46写道:
> > > >
> > > >> Hi,
> > > >>
> > > >> I created an issue about creating metadata automatically [1].
> > > >>
> > > >> To achieve this, we can introduce four configuration parameters:
> > > >>
> > > >> auto_create_metadata=false
> > > >> default_storage_group_level=2
> > > >> default_int_encoding=RLE
> > > >> default_floating_encoding=GORILLA
> > > >>
> > > >> Supposing we receive an insert statement "insert into
> > root.sg1.d1(time,
> > > >> s1, s2) values(1, 1, 1)".
> > > >> When the "auto_create_metadata" is set to true and the user didn't
> > create
> > > >> metadata before, we can set storage group to root.sg1 (because the
> > > >> default_storage_group_level is 2) and create two timeseries
> > > >> "root.sg1.d1.s1" and "root.sg1.d1.s2".
> > > >>
> > > >> The datatype can be inferred from the value type. If it is "1", we
> > treat
> > > >> it as long. If it is a "floating number", we treat it as double.
> > > >>
> > > >> I'd like to have your suggestions.
> > > >>
> > > >> [1] https://issues.apache.org/jira/browse/IOTDB-193
> > > >>
> > > >> Best,
> > > >> --
> > > >> Jialin Qiao
> > > >> School of Software, Tsinghua University
> > > >>
> > > >> 乔嘉林
> > > >> 清华大学 软件学院
> > > >>
> > > >> > -----原始邮件-----
> > > >> > 发件人: "Jialin Qiao (Jira)" <j...@apache.org>
> > > >> > 发送时间: 2019-09-03 13:12:00 (星期二)
> > > >> > 收件人: dev@iotdb.apache.org
> > > >> > 抄送:
> > > >> > 主题: [jira] [Created] (IOTDB-193) Create metadata automatically in
> > the
> > > >> insertion
> > > >> >
> > > >> > Jialin Qiao created IOTDB-193:
> > > >> > ---------------------------------
> > > >> >
> > > >> >              Summary: Create metadata automatically in the insertion
> > > >> >                  Key: IOTDB-193
> > > >> >                  URL:
> > https://issues.apache.org/jira/browse/IOTDB-193
> > > >> >              Project: Apache IoTDB
> > > >> >           Issue Type: New Feature
> > > >> >             Reporter: Jialin Qiao
> > > >> >
> > > >> >
> > > >> > Setting storage group and creating time series can be troublesome
> > when
> > > >> using IoTDB in applications.
> > > >> >
> > > >> > Let's look at the typical scenes:
> > > >> >
> > > >> > The devices send messages to Kafka. Each message includes:
> > factoryId,
> > > >> deviceId, timestamp, s1_value, s2_value, s3_value.
> > > >> >
> > > >> > The consumer of Kafka needs to consume the messages and organize the
> > > >> data into an insert statement: "insert into
> > root.factoryId.deviceId(time,
> > > >> s1, s2, s3) values(timestamp, s1_value, s2_value, s3_value)"
> > > >> >
> > > >> > Before inserting the data into IoTDB, the consumer needs to judge
> > > >> whether the storage group and time series are created before, which
> > is not
> > > >> convenient. Typically, there are 3 ways to do that:
> > > >> >  # Caching the storage group and timeseries created before in the
> > > >> consumer
> > > >> >  # Query from IoTDB whether a storage group and timeseries exists
> > > >> >  # Directly create metadata regardless of whether created before
> > > >> >
> > > >> > Could IoTDB support creating metadata (at least time series)
> > > >> automatically when receiving an insertion?
> > > >> >
> > > >> >
> > > >> >
> > > >> >
> > > >> >
> > > >> > --
> > > >> > This message was sent by Atlassian Jira
> > > >> > (v8.3.2#803003)
> > > >>
> > > >
> >

Reply via email to