[DISCUSS] FLIP-132: Temporal Table DDL

[Leonard Xu]

Hello  everyone,

Currently, user can correlate temporal table by temporal table join to enrich their fact table, a temporal table can be a changing history table which tracks the changes(e.g. database changelog) or a changing dimensioned table which materializes the changes(e.g. database table). For changing dimensioned table, Flink uses DDL to define a temporal table and visits the temporal table data by looking up the external system’s table. For changing history table, Flink uses temporal table function to define a parameterized view of changing history table and then accesses the data of view, but Temporal Table Function can only be called via Table API or YAML which is pretty inconvenient for FLINK SQL users. If we can support temporal table DDL, user no longer needs temporal table function, and they can visit temporal table easily in pure SQL world.

Flink SQL obtains the ability to interpret changelog after FLIP-95[1], changelog is natural temporal table which contains all versioned data of the the origin database table. Support temporal table on changelog would help user visit specific version of the original database table, this will enrich Flink temporal table join scenario much.

The community have already had a related discussion thread[2] and reach some consensus. Based on that, I prepare the FLIP-132[3] to support Temporal Table DDL, please see the FLIP doc for more details.

Any comment is welcome, I'm looking  forward to your feedback.
 
Best
Leonard
[1] https://cwiki.apache.org/confluence/display/FLINK/FLIP-95%3A+New+TableSource+and+TableSink+interfaces <https://cwiki.apache.org/confluence/display/FLINK/FLIP-95:+New+TableSource+and+TableSink+interfaces>
[2] http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-FLINK-16824-Creating-Temporal-Table-Function-via-DDL-td40333.html <http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-FLINK-16824-Creating-Temporal-Table-Function-via-DDL-td40333.html>
[3] https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>

[Konstantin Knauf-4]

Hi Leonard,

Thank you for preparing this FLIP. I have a few questions.

1) A  "Versioned Temporal Table DDL on source" can only be joined on the
PRIMARY KEY attribute, correct?

2) The "processing time temporal table join" example under "how to
correlate a temporal table" does not make sense to me. Isn't it the time
attribute in the ORDER BY clause of the VIEW definition that defines
whether a event-time or processing time temporal table join is used? Of
course, the time characteristic of the lookup time attribute should match
the time characteristics of the versioning attribute.

3) A "Versioned Temporal Table DDL on source" is always versioned on
operation_time regardless of the lookup table attribute (event-time or
processing time attribute), correct? Or does the lookup table attribute
define the versioning?

Thanks,

Konstantin


On Tue, Jul 28, 2020 at 9:56 AM Leonard Xu <[hidden email]> wrote:

> Hello  everyone,
>
> Currently, user can correlate temporal table by temporal table join to
> enrich their fact table, a temporal table can be a changing history table
> which tracks the changes(e.g. database changelog) or a changing dimensioned
> table which materializes the changes(e.g. database table). For changing
> dimensioned table, Flink uses DDL to define a temporal table and visits the
> temporal table data by looking up the external system’s table. For changing
> history table, Flink uses temporal table function to define a parameterized
> view of changing history table and then accesses the data of view, but
> Temporal Table Function can only be called via Table API or YAML which is
> pretty inconvenient for FLINK SQL users. If we can support temporal table
> DDL, user no longer needs temporal table function, and they can visit
> temporal table easily in pure SQL world.
>
> Flink SQL obtains the ability to interpret changelog after FLIP-95[1],
> changelog is natural temporal table which contains all versioned data of
> the the origin database table. Support temporal table on changelog would
> help user visit specific version of the original database table, this will
> enrich Flink temporal table join scenario much.
>
> The community have already had a related discussion thread[2] and reach
> some consensus. Based on that, I prepare the FLIP-132[3] to support
> Temporal Table DDL, please see the FLIP doc for more details.
>
> Any comment is welcome, I'm looking  forward to your feedback.
>
> Best
> Leonard
> [1]
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-95%3A+New+TableSource+and+TableSink+interfaces
> <
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-95:+New+TableSource+and+TableSink+interfaces
> >
> [2]
> http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-FLINK-16824-Creating-Temporal-Table-Function-via-DDL-td40333.html
> <
> http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-FLINK-16824-Creating-Temporal-Table-Function-via-DDL-td40333.html
> >
> [3]
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> >

--

Konstantin Knauf

https://twitter.com/snntrable

https://github.com/knaufk

[Leonard Xu]

Hi, Konstantin

>
> 1) A  "Versioned Temporal Table DDL on source" can only be joined on the
> PRIMARY KEY attribute, correct?
Yes, the PRIMARY KEY would be join key.

>
> 2) Isn't it the time attribute in the ORDER BY clause of the VIEW definition that defines
> whether a event-time or processing time temporal table join is used?

I think event-time or processing-time temporal table join depends on fact table’s time attribute in temporal join rather than from temporal table side, the event-time or processing time in temporal table is just used to split the validity period of versioned snapshot of temporal table. The  processing time attribute is not  necessary for temporal table without version, only the primary key is required, the following VIEW is also valid for temporal table without version.
CREATE VIEW latest_rates AS
SELECT currency, LAST_VALUE(rate)            -- only keep the latest version
FROM rates
GROUP BY currency;                           -- inferred primary key


>
> 3) A "Versioned Temporal Table DDL on source" is always versioned on
> operation_time regardless of the lookup table attribute (event-time or
> processing time attribute), correct?


Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the o.time value to lookup the version of the temporal table.
For fact table has the processing time attribute, it means only lookup the latest version of temporal table and we can do some optimization in implementation like only keep the latest version.


Best
Leonard

[Seth Wiesman-4]

Hi Leondard,

Thank you for pushing this, I think the updated syntax looks really good
and the semantics make sense to me.

+1

Seth

On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:

> Hi, Konstantin
>
> >
> > 1) A  "Versioned Temporal Table DDL on source" can only be joined on the
> > PRIMARY KEY attribute, correct?
> Yes, the PRIMARY KEY would be join key.
>
> >
> > 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
> definition that defines
> > whether a event-time or processing time temporal table join is used?
>
> I think event-time or processing-time temporal table join depends on fact
> table’s time attribute in temporal join rather than from temporal table
> side, the event-time or processing time in temporal table is just used to
> split the validity period of versioned snapshot of temporal table. The
> processing time attribute is not  necessary for temporal table without
> version, only the primary key is required, the following VIEW is also valid
> for temporal table without version.
> CREATE VIEW latest_rates AS
> SELECT currency, LAST_VALUE(rate)            -- only keep the latest
> version
> FROM rates
> GROUP BY currency;                           -- inferred primary key
>
>
> >
> > 3) A "Versioned Temporal Table DDL on source" is always versioned on
> > operation_time regardless of the lookup table attribute (event-time or
> > processing time attribute), correct?
>
>
> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the o.time
> value to lookup the version of the temporal table.
> For fact table has the processing time attribute, it means only lookup the
> latest version of temporal table and we can do some optimization in
> implementation like only keep the latest version.
>
>
> Best
> Leonard

[Fabian Hueske-2]

Hi Leonard,

Thanks for this FLIP!
Looks good from my side.

Cheers, Fabian

Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <[hidden email]
>:

> Hi Leondard,
>
> Thank you for pushing this, I think the updated syntax looks really good
> and the semantics make sense to me.
>
> +1
>
> Seth
>
> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:
>
> > Hi, Konstantin
> >
> > >
> > > 1) A  "Versioned Temporal Table DDL on source" can only be joined on
> the
> > > PRIMARY KEY attribute, correct?
> > Yes, the PRIMARY KEY would be join key.
> >
> > >
> > > 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
> > definition that defines
> > > whether a event-time or processing time temporal table join is used?
> >
> > I think event-time or processing-time temporal table join depends on fact
> > table’s time attribute in temporal join rather than from temporal table
> > side, the event-time or processing time in temporal table is just used to
> > split the validity period of versioned snapshot of temporal table. The
> > processing time attribute is not  necessary for temporal table without
> > version, only the primary key is required, the following VIEW is also
> valid
> > for temporal table without version.
> > CREATE VIEW latest_rates AS
> > SELECT currency, LAST_VALUE(rate)            -- only keep the latest
> > version
> > FROM rates
> > GROUP BY currency;                           -- inferred primary key
> >
> >
> > >
> > > 3) A "Versioned Temporal Table DDL on source" is always versioned on
> > > operation_time regardless of the lookup table attribute (event-time or
> > > processing time attribute), correct?
> >
> >
> > Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the o.time
> > value to lookup the version of the temporal table.
> > For fact table has the processing time attribute, it means only lookup
> the
> > latest version of temporal table and we can do some optimization in
> > implementation like only keep the latest version.
> >
> >
> > Best
> > Leonard
>

[Jark Wu-2]

Thanks Leonard for the great FLIP. I think it is in very good shape.
+1 to start a vote.

Best,
Jark

On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]> wrote:

> Hi Leonard,
>
> Thanks for this FLIP!
> Looks good from my side.
>
> Cheers, Fabian
>
> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
> [hidden email]
> >:
>
> > Hi Leondard,
> >
> > Thank you for pushing this, I think the updated syntax looks really good
> > and the semantics make sense to me.
> >
> > +1
> >
> > Seth
> >
> > On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:
> >
> > > Hi, Konstantin
> > >
> > > >
> > > > 1) A  "Versioned Temporal Table DDL on source" can only be joined on
> > the
> > > > PRIMARY KEY attribute, correct?
> > > Yes, the PRIMARY KEY would be join key.
> > >
> > > >
> > > > 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
> > > definition that defines
> > > > whether a event-time or processing time temporal table join is used?
> > >
> > > I think event-time or processing-time temporal table join depends on
> fact
> > > table’s time attribute in temporal join rather than from temporal table
> > > side, the event-time or processing time in temporal table is just used
> to
> > > split the validity period of versioned snapshot of temporal table. The
> > > processing time attribute is not  necessary for temporal table without
> > > version, only the primary key is required, the following VIEW is also
> > valid
> > > for temporal table without version.
> > > CREATE VIEW latest_rates AS
> > > SELECT currency, LAST_VALUE(rate)            -- only keep the latest
> > > version
> > > FROM rates
> > > GROUP BY currency;                           -- inferred primary key
> > >
> > >
> > > >
> > > > 3) A "Versioned Temporal Table DDL on source" is always versioned on
> > > > operation_time regardless of the lookup table attribute (event-time
> or
> > > > processing time attribute), correct?
> > >
> > >
> > > Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
> o.time
> > > value to lookup the version of the temporal table.
> > > For fact table has the processing time attribute, it means only lookup
> > the
> > > latest version of temporal table and we can do some optimization in
> > > implementation like only keep the latest version.
> > >
> > >
> > > Best
> > > Leonard
> >
>

[godfreyhe]

Thanks Lennard for driving this FLIP.
Looks good to me.

Best,
Godfrey

Jark Wu <[hidden email]> 于2020年8月3日周一 下午12:04写道：

> Thanks Leonard for the great FLIP. I think it is in very good shape.
> +1 to start a vote.
>
> Best,
> Jark
>
> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]> wrote:
>
> > Hi Leonard,
> >
> > Thanks for this FLIP!
> > Looks good from my side.
> >
> > Cheers, Fabian
> >
> > Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
> > [hidden email]
> > >:
> >
> > > Hi Leondard,
> > >
> > > Thank you for pushing this, I think the updated syntax looks really
> good
> > > and the semantics make sense to me.
> > >
> > > +1
> > >
> > > Seth
> > >
> > > On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:
> > >
> > > > Hi, Konstantin
> > > >
> > > > >
> > > > > 1) A  "Versioned Temporal Table DDL on source" can only be joined
> on
> > > the
> > > > > PRIMARY KEY attribute, correct?
> > > > Yes, the PRIMARY KEY would be join key.
> > > >
> > > > >
> > > > > 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
> > > > definition that defines
> > > > > whether a event-time or processing time temporal table join is
> used?
> > > >
> > > > I think event-time or processing-time temporal table join depends on
> > fact
> > > > table’s time attribute in temporal join rather than from temporal
> table
> > > > side, the event-time or processing time in temporal table is just
> used
> > to
> > > > split the validity period of versioned snapshot of temporal table.
> The
> > > > processing time attribute is not  necessary for temporal table
> without
> > > > version, only the primary key is required, the following VIEW is also
> > > valid
> > > > for temporal table without version.
> > > > CREATE VIEW latest_rates AS
> > > > SELECT currency, LAST_VALUE(rate)            -- only keep the latest
> > > > version
> > > > FROM rates
> > > > GROUP BY currency;                           -- inferred primary key
> > > >
> > > >
> > > > >
> > > > > 3) A "Versioned Temporal Table DDL on source" is always versioned
> on
> > > > > operation_time regardless of the lookup table attribute (event-time
> > or
> > > > > processing time attribute), correct?
> > > >
> > > >
> > > > Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
> > o.time
> > > > value to lookup the version of the temporal table.
> > > > For fact table has the processing time attribute, it means only
> lookup
> > > the
> > > > latest version of temporal table and we can do some optimization in
> > > > implementation like only keep the latest version.
> > > >
> > > >
> > > > Best
> > > > Leonard
> > >
> >
>

[Leonard Xu]

Thanks Konstantin,

Regarding your questions, hope my comments has address your questions and I also add a few explanation in the FLIP.

Thank you all for the feedback,

It seems everyone involved  in this thread has reached a consensus.
I will start a vote thread  later.


Best,
Leonard

> 在 2020年8月3日，19:35，godfrey he <[hidden email]> 写道：
>
> Thanks Lennard for driving this FLIP.
> Looks good to me.
>
> Best,
> Godfrey
>
> Jark Wu <[hidden email]> 于2020年8月3日周一 下午12:04写道：
>
>> Thanks Leonard for the great FLIP. I think it is in very good shape.
>> +1 to start a vote.
>>
>> Best,
>> Jark
>>
>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]> wrote:
>>
>>> Hi Leonard,
>>>
>>> Thanks for this FLIP!
>>> Looks good from my side.
>>>
>>> Cheers, Fabian
>>>
>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>> [hidden email]
>>>> :
>>>
>>>> Hi Leondard,
>>>>
>>>> Thank you for pushing this, I think the updated syntax looks really
>> good
>>>> and the semantics make sense to me.
>>>>
>>>> +1
>>>>
>>>> Seth
>>>>
>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:
>>>>
>>>>> Hi, Konstantin
>>>>>
>>>>>>
>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be joined
>> on
>>>> the
>>>>>> PRIMARY KEY attribute, correct?
>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>
>>>>>>
>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
>>>>> definition that defines
>>>>>> whether a event-time or processing time temporal table join is
>> used?
>>>>>
>>>>> I think event-time or processing-time temporal table join depends on
>>> fact
>>>>> table’s time attribute in temporal join rather than from temporal
>> table
>>>>> side, the event-time or processing time in temporal table is just
>> used
>>> to
>>>>> split the validity period of versioned snapshot of temporal table.
>> The
>>>>> processing time attribute is not  necessary for temporal table
>> without
>>>>> version, only the primary key is required, the following VIEW is also
>>>> valid
>>>>> for temporal table without version.
>>>>> CREATE VIEW latest_rates AS
>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the latest
>>>>> version
>>>>> FROM rates
>>>>> GROUP BY currency;                           -- inferred primary key
>>>>>
>>>>>
>>>>>>
>>>>>> 3) A "Versioned Temporal Table DDL on source" is always versioned
>> on
>>>>>> operation_time regardless of the lookup table attribute (event-time
>>> or
>>>>>> processing time attribute), correct?
>>>>>
>>>>>
>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
>>> o.time
>>>>> value to lookup the version of the temporal table.
>>>>> For fact table has the processing time attribute, it means only
>> lookup
>>>> the
>>>>> latest version of temporal table and we can do some optimization in
>>>>> implementation like only keep the latest version.
>>>>>
>>>>>
>>>>> Best
>>>>> Leonard
>>>>
>>>
>>

[Timo Walther-2]

Hi Leonard,

sorry for jumping into the discussion so late. But I have two questions:

1) Naming: Is operation time a good term for this concept? If I read
"The operation time is the time when the changes happened in system." or
"The system time of DML execution in database", why don't we call it
`ChangelogTime` or `SystemTime`? Introducing another terminology of time
in Flink should be thought through.

2) Exposing it through `org.apache.flink.types.Row`: Shall we also
expose the concept of time through the user-level `Row` type? The FLIP
does not mention this explictly. I think we can keep it as an internal
concept but I just wanted to ask for clarification.

Thanks,
Timo


On 04.08.20 04:58, Leonard Xu wrote:

> Thanks Konstantin,
>
> Regarding your questions, hope my comments has address your questions and I also add a few explanation in the FLIP.
>
> Thank you all for the feedback,
>
> It seems everyone involved  in this thread has reached a consensus.
> I will start a vote thread  later.
>
>
> Best,
> Leonard
>
>
>> 在 2020年8月3日，19:35，godfrey he <[hidden email]> 写道：
>>
>> Thanks Lennard for driving this FLIP.
>> Looks good to me.
>>
>> Best,
>> Godfrey
>>
>> Jark Wu <[hidden email]> 于2020年8月3日周一 下午12:04写道：
>>
>>> Thanks Leonard for the great FLIP. I think it is in very good shape.
>>> +1 to start a vote.
>>>
>>> Best,
>>> Jark
>>>
>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]> wrote:
>>>
>>>> Hi Leonard,
>>>>
>>>> Thanks for this FLIP!
>>>> Looks good from my side.
>>>>
>>>> Cheers, Fabian
>>>>
>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>>> [hidden email]
>>>>> :
>>>>
>>>>> Hi Leondard,
>>>>>
>>>>> Thank you for pushing this, I think the updated syntax looks really
>>> good
>>>>> and the semantics make sense to me.
>>>>>
>>>>> +1
>>>>>
>>>>> Seth
>>>>>
>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:
>>>>>
>>>>>> Hi, Konstantin
>>>>>>
>>>>>>>
>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be joined
>>> on
>>>>> the
>>>>>>> PRIMARY KEY attribute, correct?
>>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>>
>>>>>>>
>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
>>>>>> definition that defines
>>>>>>> whether a event-time or processing time temporal table join is
>>> used?
>>>>>>
>>>>>> I think event-time or processing-time temporal table join depends on
>>>> fact
>>>>>> table’s time attribute in temporal join rather than from temporal
>>> table
>>>>>> side, the event-time or processing time in temporal table is just
>>> used
>>>> to
>>>>>> split the validity period of versioned snapshot of temporal table.
>>> The
>>>>>> processing time attribute is not  necessary for temporal table
>>> without
>>>>>> version, only the primary key is required, the following VIEW is also
>>>>> valid
>>>>>> for temporal table without version.
>>>>>> CREATE VIEW latest_rates AS
>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the latest
>>>>>> version
>>>>>> FROM rates
>>>>>> GROUP BY currency;                           -- inferred primary key
>>>>>>
>>>>>>
>>>>>>>
>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always versioned
>>> on
>>>>>>> operation_time regardless of the lookup table attribute (event-time
>>>> or
>>>>>>> processing time attribute), correct?
>>>>>>
>>>>>>
>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
>>>> o.time
>>>>>> value to lookup the version of the temporal table.
>>>>>> For fact table has the processing time attribute, it means only
>>> lookup
>>>>> the
>>>>>> latest version of temporal table and we can do some optimization in
>>>>>> implementation like only keep the latest version.
>>>>>>
>>>>>>
>>>>>> Best
>>>>>> Leonard
>>>>>
>>>>
>>>
>

[Leonard Xu]

Hi, Timo

Thanks for you response.

> 1) Naming: Is operation time a good term for this concept? If I read "The operation time is the time when the changes happened in system." or "The system time of DML execution in database", why don't we call it `ChangelogTime` or `SystemTime`? Introducing another terminology of time in Flink should be thought through.

I agree that we should thought through. I have considered the name `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on the name.

I proposed `operationTime` because most changelog comes from Database and we always called an action as `operation` rather than `change` in Database, the operation time is  easier to understand  for database users, but it's more like a database terminology.

For `SystemTime`, user may confuse which one does the system in `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s not a good name.

`ChangelogTime` is a pretty choice which is more unified with existed terminology `Changelog` and `ChangelogMode`, so let me use `ChangelogTime` and I’ll update the FLIP.


> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also expose the concept of time through the user-level `Row` type? The FLIP does not mention this explictly. I think we can keep it as an internal concept but I just wanted to ask for clarification.

Yes, I want to keep it as an internal concept, we have discussed that changelog time concept should be the third time concept(the other two are event-time and processing-time). It’s not easy for normal users(or to help normal users) understand the three concepts accurately, and I did not find a big enough scenario that user need to touch the changelog time for now, so I tend to do not expose the concept to users.


Best,
Leonard

>
> On 04.08.20 04:58, Leonard Xu wrote:
>> Thanks Konstantin,
>> Regarding your questions, hope my comments has address your questions and I also add a few explanation in the FLIP.
>> Thank you all for the feedback,
>> It seems everyone involved  in this thread has reached a consensus.
>> I will start a vote thread  later.
>> Best,
>> Leonard
>>> 在 2020年8月3日，19:35，godfrey he <[hidden email]> 写道：
>>>
>>> Thanks Lennard for driving this FLIP.
>>> Looks good to me.
>>>
>>> Best,
>>> Godfrey
>>>
>>> Jark Wu <[hidden email]> 于2020年8月3日周一 下午12:04写道：
>>>
>>>> Thanks Leonard for the great FLIP. I think it is in very good shape.
>>>> +1 to start a vote.
>>>>
>>>> Best,
>>>> Jark
>>>>
>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]> wrote:
>>>>
>>>>> Hi Leonard,
>>>>>
>>>>> Thanks for this FLIP!
>>>>> Looks good from my side.
>>>>>
>>>>> Cheers, Fabian
>>>>>
>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>>>> [hidden email]
>>>>>> :
>>>>>
>>>>>> Hi Leondard,
>>>>>>
>>>>>> Thank you for pushing this, I think the updated syntax looks really
>>>> good
>>>>>> and the semantics make sense to me.
>>>>>>
>>>>>> +1
>>>>>>
>>>>>> Seth
>>>>>>
>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:
>>>>>>
>>>>>>> Hi, Konstantin
>>>>>>>
>>>>>>>>
>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be joined
>>>> on
>>>>>> the
>>>>>>>> PRIMARY KEY attribute, correct?
>>>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>>>
>>>>>>>>
>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
>>>>>>> definition that defines
>>>>>>>> whether a event-time or processing time temporal table join is
>>>> used?
>>>>>>>
>>>>>>> I think event-time or processing-time temporal table join depends on
>>>>> fact
>>>>>>> table’s time attribute in temporal join rather than from temporal
>>>> table
>>>>>>> side, the event-time or processing time in temporal table is just
>>>> used
>>>>> to
>>>>>>> split the validity period of versioned snapshot of temporal table.
>>>> The
>>>>>>> processing time attribute is not  necessary for temporal table
>>>> without
>>>>>>> version, only the primary key is required, the following VIEW is also
>>>>>> valid
>>>>>>> for temporal table without version.
>>>>>>> CREATE VIEW latest_rates AS
>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the latest
>>>>>>> version
>>>>>>> FROM rates
>>>>>>> GROUP BY currency;                           -- inferred primary key
>>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always versioned
>>>> on
>>>>>>>> operation_time regardless of the lookup table attribute (event-time
>>>>> or
>>>>>>>> processing time attribute), correct?
>>>>>>>
>>>>>>>
>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
>>>>> o.time
>>>>>>> value to lookup the version of the temporal table.
>>>>>>> For fact table has the processing time attribute, it means only
>>>> lookup
>>>>>> the
>>>>>>> latest version of temporal table and we can do some optimization in
>>>>>>> implementation like only keep the latest version.
>>>>>>>
>>>>>>>
>>>>>>> Best
>>>>>>> Leonard
>>>>>>
>>>>>
>>>>
>

[Leonard Xu]

Hi, all

I’ve updated the FLIP[1] with the terminology `ChangelogTime`.

Best
Leonard
[1]  https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>

> 在 2020年8月4日，20:58，Leonard Xu <[hidden email]> 写道：
>
> Hi, Timo
>
> Thanks for you response.
>
>> 1) Naming: Is operation time a good term for this concept? If I read "The operation time is the time when the changes happened in system." or "The system time of DML execution in database", why don't we call it `ChangelogTime` or `SystemTime`? Introducing another terminology of time in Flink should be thought through.
>
> I agree that we should thought through. I have considered the name `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on the name.
>
> I proposed `operationTime` because most changelog comes from Database and we always called an action as `operation` rather than `change` in Database, the operation time is  easier to understand  for database users, but it's more like a database terminology.
>
> For `SystemTime`, user may confuse which one does the system in `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s not a good name.
>
> `ChangelogTime` is a pretty choice which is more unified with existed terminology `Changelog` and `ChangelogMode`, so let me use `ChangelogTime` and I’ll update the FLIP.
>
>
>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also expose the concept of time through the user-level `Row` type? The FLIP does not mention this explictly. I think we can keep it as an internal concept but I just wanted to ask for clarification.
>
> Yes, I want to keep it as an internal concept, we have discussed that changelog time concept should be the third time concept(the other two are event-time and processing-time). It’s not easy for normal users(or to help normal users) understand the three concepts accurately, and I did not find a big enough scenario that user need to touch the changelog time for now, so I tend to do not expose the concept to users.
>
>
> Best,
> Leonard
>
>
>>
>> On 04.08.20 04:58, Leonard Xu wrote:
>>> Thanks Konstantin,
>>> Regarding your questions, hope my comments has address your questions and I also add a few explanation in the FLIP.
>>> Thank you all for the feedback,
>>> It seems everyone involved  in this thread has reached a consensus.
>>> I will start a vote thread  later.
>>> Best,
>>> Leonard
>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email]> 写道：
>>>>
>>>> Thanks Lennard for driving this FLIP.
>>>> Looks good to me.
>>>>
>>>> Best,
>>>> Godfrey
>>>>
>>>> Jark Wu <[hidden email]> 于2020年8月3日周一 下午12:04写道：
>>>>
>>>>> Thanks Leonard for the great FLIP. I think it is in very good shape.
>>>>> +1 to start a vote.
>>>>>
>>>>> Best,
>>>>> Jark
>>>>>
>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]> wrote:
>>>>>
>>>>>> Hi Leonard,
>>>>>>
>>>>>> Thanks for this FLIP!
>>>>>> Looks good from my side.
>>>>>>
>>>>>> Cheers, Fabian
>>>>>>
>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>>>>> [hidden email]
>>>>>>> :
>>>>>>
>>>>>>> Hi Leondard,
>>>>>>>
>>>>>>> Thank you for pushing this, I think the updated syntax looks really
>>>>> good
>>>>>>> and the semantics make sense to me.
>>>>>>>
>>>>>>> +1
>>>>>>>
>>>>>>> Seth
>>>>>>>
>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]> wrote:
>>>>>>>
>>>>>>>> Hi, Konstantin
>>>>>>>>
>>>>>>>>>
>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be joined
>>>>> on
>>>>>>> the
>>>>>>>>> PRIMARY KEY attribute, correct?
>>>>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>>>>
>>>>>>>>>
>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
>>>>>>>> definition that defines
>>>>>>>>> whether a event-time or processing time temporal table join is
>>>>> used?
>>>>>>>>
>>>>>>>> I think event-time or processing-time temporal table join depends on
>>>>>> fact
>>>>>>>> table’s time attribute in temporal join rather than from temporal
>>>>> table
>>>>>>>> side, the event-time or processing time in temporal table is just
>>>>> used
>>>>>> to
>>>>>>>> split the validity period of versioned snapshot of temporal table.
>>>>> The
>>>>>>>> processing time attribute is not  necessary for temporal table
>>>>> without
>>>>>>>> version, only the primary key is required, the following VIEW is also
>>>>>>> valid
>>>>>>>> for temporal table without version.
>>>>>>>> CREATE VIEW latest_rates AS
>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the latest
>>>>>>>> version
>>>>>>>> FROM rates
>>>>>>>> GROUP BY currency;                           -- inferred primary key
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always versioned
>>>>> on
>>>>>>>>> operation_time regardless of the lookup table attribute (event-time
>>>>>> or
>>>>>>>>> processing time attribute), correct?
>>>>>>>>
>>>>>>>>
>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
>>>>>> o.time
>>>>>>>> value to lookup the version of the temporal table.
>>>>>>>> For fact table has the processing time attribute, it means only
>>>>> lookup
>>>>>>> the
>>>>>>>> latest version of temporal table and we can do some optimization in
>>>>>>>> implementation like only keep the latest version.
>>>>>>>>
>>>>>>>>
>>>>>>>> Best
>>>>>>>> Leonard
>>>>>>>
>>>>>>
>>>>>
>>
>

[Leonard Xu]

Hi, all

After a detailed offline discussion about the temporal table related concept and behavior, we had a reliable solution and rejected several alternatives.
Compared to rejected alternatives, the proposed approach is a more unified story and also friendly to user and current Flink framework.
I improved the FLIP[1] with the proposed approach and refactored the document organization to make it clear enough.

Please let me know if you have any concerns, I’m looking forward your comments.


Best
Leonard

[1] https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>

> 在 2020年8月4日，21:25，Leonard Xu <[hidden email]> 写道：
>
> Hi, all
>
> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
>
> Best
> Leonard
> [1]  https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>
>> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:[hidden email]>> 写道：
>>
>> Hi, Timo
>>
>> Thanks for you response.
>>
>>> 1) Naming: Is operation time a good term for this concept? If I read "The operation time is the time when the changes happened in system." or "The system time of DML execution in database", why don't we call it `ChangelogTime` or `SystemTime`? Introducing another terminology of time in Flink should be thought through.
>>
>> I agree that we should thought through. I have considered the name `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on the name.
>>
>> I proposed `operationTime` because most changelog comes from Database and we always called an action as `operation` rather than `change` in Database, the operation time is  easier to understand  for database users, but it's more like a database terminology.
>>
>> For `SystemTime`, user may confuse which one does the system in `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s not a good name.
>>
>> `ChangelogTime` is a pretty choice which is more unified with existed terminology `Changelog` and `ChangelogMode`, so let me use `ChangelogTime` and I’ll update the FLIP.
>>
>>
>>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also expose the concept of time through the user-level `Row` type? The FLIP does not mention this explictly. I think we can keep it as an internal concept but I just wanted to ask for clarification.
>>
>> Yes, I want to keep it as an internal concept, we have discussed that changelog time concept should be the third time concept(the other two are event-time and processing-time). It’s not easy for normal users(or to help normal users) understand the three concepts accurately, and I did not find a big enough scenario that user need to touch the changelog time for now, so I tend to do not expose the concept to users.
>>
>>
>> Best,
>> Leonard
>>
>>
>>>
>>> On 04.08.20 04:58, Leonard Xu wrote:
>>>> Thanks Konstantin,
>>>> Regarding your questions, hope my comments has address your questions and I also add a few explanation in the FLIP.
>>>> Thank you all for the feedback,
>>>> It seems everyone involved  in this thread has reached a consensus.
>>>> I will start a vote thread  later.
>>>> Best,
>>>> Leonard
>>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:[hidden email]>> 写道：
>>>>>
>>>>> Thanks Lennard for driving this FLIP.
>>>>> Looks good to me.
>>>>>
>>>>> Best,
>>>>> Godfrey
>>>>>
>>>>> Jark Wu <[hidden email] <mailto:[hidden email]>> 于2020年8月3日周一 下午12:04写道：
>>>>>
>>>>>> Thanks Leonard for the great FLIP. I think it is in very good shape.
>>>>>> +1 to start a vote.
>>>>>>
>>>>>> Best,
>>>>>> Jark
>>>>>>
>>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email] <mailto:[hidden email]>> wrote:
>>>>>>
>>>>>>> Hi Leonard,
>>>>>>>
>>>>>>> Thanks for this FLIP!
>>>>>>> Looks good from my side.
>>>>>>>
>>>>>>> Cheers, Fabian
>>>>>>>
>>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>>>>>> [hidden email] <mailto:[hidden email]>
>>>>>>>> :
>>>>>>>
>>>>>>>> Hi Leondard,
>>>>>>>>
>>>>>>>> Thank you for pushing this, I think the updated syntax looks really
>>>>>> good
>>>>>>>> and the semantics make sense to me.
>>>>>>>>
>>>>>>>> +1
>>>>>>>>
>>>>>>>> Seth
>>>>>>>>
>>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email] <mailto:[hidden email]>> wrote:
>>>>>>>>
>>>>>>>>> Hi, Konstantin
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be joined
>>>>>> on
>>>>>>>> the
>>>>>>>>>> PRIMARY KEY attribute, correct?
>>>>>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the VIEW
>>>>>>>>> definition that defines
>>>>>>>>>> whether a event-time or processing time temporal table join is
>>>>>> used?
>>>>>>>>>
>>>>>>>>> I think event-time or processing-time temporal table join depends on
>>>>>>> fact
>>>>>>>>> table’s time attribute in temporal join rather than from temporal
>>>>>> table
>>>>>>>>> side, the event-time or processing time in temporal table is just
>>>>>> used
>>>>>>> to
>>>>>>>>> split the validity period of versioned snapshot of temporal table.
>>>>>> The
>>>>>>>>> processing time attribute is not  necessary for temporal table
>>>>>> without
>>>>>>>>> version, only the primary key is required, the following VIEW is also
>>>>>>>> valid
>>>>>>>>> for temporal table without version.
>>>>>>>>> CREATE VIEW latest_rates AS
>>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the latest
>>>>>>>>> version
>>>>>>>>> FROM rates
>>>>>>>>> GROUP BY currency;                           -- inferred primary key
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always versioned
>>>>>> on
>>>>>>>>>> operation_time regardless of the lookup table attribute (event-time
>>>>>>> or
>>>>>>>>>> processing time attribute), correct?
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
>>>>>>> o.time
>>>>>>>>> value to lookup the version of the temporal table.
>>>>>>>>> For fact table has the processing time attribute, it means only
>>>>>> lookup
>>>>>>>> the
>>>>>>>>> latest version of temporal table and we can do some optimization in
>>>>>>>>> implementation like only keep the latest version.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Best
>>>>>>>>> Leonard
>>>>>>>>
>>>>>>>
>>>>>>
>>>
>>
>

[Rui Li]

Hey Leonard,

Thanks for summarizing the document. I have one quick question. I
understand a temporal table w/o version means each row in the table only
has one version. But are we still able to track different views of such a
table through time, as rows are added/deleted to/from the table? For
example, suppose I have an append-only table source with event-time and PK,
will I be allowed to do an event-time temporal join with this table?

On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email]> wrote:

> Hi, all
>
> After a detailed offline discussion about the temporal table related
> concept and behavior, we had a reliable solution and rejected several
> alternatives.
> Compared to rejected alternatives, the proposed approach is a more unified
> story and also friendly to user and current Flink framework.
> I improved the FLIP[1] with the proposed approach and refactored the
> document organization to make it clear enough.
>
> Please let me know if you have any concerns, I’m looking forward your
> comments.
>
>
> Best
> Leonard
>
> [1]
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> >
>
>
> > 在 2020年8月4日，21:25，Leonard Xu <[hidden email]> 写道：
> >
> > Hi, all
> >
> > I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
> >
> > Best
> > Leonard
> > [1]
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> >
> >
> >> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:
> [hidden email]>> 写道：
> >>
> >> Hi, Timo
> >>
> >> Thanks for you response.
> >>
> >>> 1) Naming: Is operation time a good term for this concept? If I read
> "The operation time is the time when the changes happened in system." or
> "The system time of DML execution in database", why don't we call it
> `ChangelogTime` or `SystemTime`? Introducing another terminology of time in
> Flink should be thought through.
> >>
> >> I agree that we should thought through. I have considered the name
> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on the
> name.
> >>
> >> I proposed `operationTime` because most changelog comes from Database
> and we always called an action as `operation` rather than `change` in
> Database, the operation time is  easier to understand  for database users,
> but it's more like a database terminology.
> >>
> >> For `SystemTime`, user may confuse which one does the system in
> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s not a
> good name.
> >>
> >> `ChangelogTime` is a pretty choice which is more unified with existed
> terminology `Changelog` and `ChangelogMode`, so let me use `ChangelogTime`
> and I’ll update the FLIP.
> >>
> >>
> >>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
> expose the concept of time through the user-level `Row` type? The FLIP does
> not mention this explictly. I think we can keep it as an internal concept
> but I just wanted to ask for clarification.
> >>
> >> Yes, I want to keep it as an internal concept, we have discussed that
> changelog time concept should be the third time concept(the other two are
> event-time and processing-time). It’s not easy for normal users(or to help
> normal users) understand the three concepts accurately, and I did not find
> a big enough scenario that user need to touch the changelog time for now,
> so I tend to do not expose the concept to users.
> >>
> >>
> >> Best,
> >> Leonard
> >>
> >>
> >>>
> >>> On 04.08.20 04:58, Leonard Xu wrote:
> >>>> Thanks Konstantin,
> >>>> Regarding your questions, hope my comments has address your questions
> and I also add a few explanation in the FLIP.
> >>>> Thank you all for the feedback,
> >>>> It seems everyone involved  in this thread has reached a consensus.
> >>>> I will start a vote thread  later.
> >>>> Best,
> >>>> Leonard
> >>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:
> [hidden email]>> 写道：
> >>>>>
> >>>>> Thanks Lennard for driving this FLIP.
> >>>>> Looks good to me.
> >>>>>
> >>>>> Best,
> >>>>> Godfrey
> >>>>>
> >>>>> Jark Wu <[hidden email] <mailto:[hidden email]>> 于2020年8月3日周一
> 下午12:04写道：
> >>>>>
> >>>>>> Thanks Leonard for the great FLIP. I think it is in very good shape.
> >>>>>> +1 to start a vote.
> >>>>>>
> >>>>>> Best,
> >>>>>> Jark
> >>>>>>
> >>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]
> <mailto:[hidden email]>> wrote:
> >>>>>>
> >>>>>>> Hi Leonard,
> >>>>>>>
> >>>>>>> Thanks for this FLIP!
> >>>>>>> Looks good from my side.
> >>>>>>>
> >>>>>>> Cheers, Fabian
> >>>>>>>
> >>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
> >>>>>>> [hidden email] <mailto:[hidden email]>
> >>>>>>>> :
> >>>>>>>
> >>>>>>>> Hi Leondard,
> >>>>>>>>
> >>>>>>>> Thank you for pushing this, I think the updated syntax looks
> really
> >>>>>> good
> >>>>>>>> and the semantics make sense to me.
> >>>>>>>>
> >>>>>>>> +1
> >>>>>>>>
> >>>>>>>> Seth
> >>>>>>>>
> >>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]
> <mailto:[hidden email]>> wrote:
> >>>>>>>>
> >>>>>>>>> Hi, Konstantin
> >>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be
> joined
> >>>>>> on
> >>>>>>>> the
> >>>>>>>>>> PRIMARY KEY attribute, correct?
> >>>>>>>>> Yes, the PRIMARY KEY would be join key.
> >>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the
> VIEW
> >>>>>>>>> definition that defines
> >>>>>>>>>> whether a event-time or processing time temporal table join is
> >>>>>> used?
> >>>>>>>>>
> >>>>>>>>> I think event-time or processing-time temporal table join
> depends on
> >>>>>>> fact
> >>>>>>>>> table’s time attribute in temporal join rather than from temporal
> >>>>>> table
> >>>>>>>>> side, the event-time or processing time in temporal table is just
> >>>>>> used
> >>>>>>> to
> >>>>>>>>> split the validity period of versioned snapshot of temporal
> table.
> >>>>>> The
> >>>>>>>>> processing time attribute is not  necessary for temporal table
> >>>>>> without
> >>>>>>>>> version, only the primary key is required, the following VIEW is
> also
> >>>>>>>> valid
> >>>>>>>>> for temporal table without version.
> >>>>>>>>> CREATE VIEW latest_rates AS
> >>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the
> latest
> >>>>>>>>> version
> >>>>>>>>> FROM rates
> >>>>>>>>> GROUP BY currency;                           -- inferred primary
> key
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always
> versioned
> >>>>>> on
> >>>>>>>>>> operation_time regardless of the lookup table attribute
> (event-time
> >>>>>>> or
> >>>>>>>>>> processing time attribute), correct?
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
> >>>>>>> o.time
> >>>>>>>>> value to lookup the version of the temporal table.
> >>>>>>>>> For fact table has the processing time attribute, it means only
> >>>>>> lookup
> >>>>>>>> the
> >>>>>>>>> latest version of temporal table and we can do some optimization
> in
> >>>>>>>>> implementation like only keep the latest version.
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Best
> >>>>>>>>> Leonard
> >>>>>>>>
> >>>>>>>
> >>>>>>
> >>>
> >>
> >
>
>

--
Best regards!
Rui Li

[Leonard Xu]

> But are we still able to track different views of such a
> table through time, as rows are added/deleted to/from the table?

Yes, in fact we support temporal table from changlog which contains all possible message types(INSERT/UPDATE/DELETE).

> For
> example, suppose I have an append-only table source with event-time and PK,
> will I be allowed to do an event-time temporal join with this table?
Yes, I list some examples in the doc, the example versioned_rates3  is this case exactly.

Best
Leonard

>
> On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email] <mailto:[hidden email]>> wrote:
>
>> Hi, all
>>
>> After a detailed offline discussion about the temporal table related
>> concept and behavior, we had a reliable solution and rejected several
>> alternatives.
>> Compared to rejected alternatives, the proposed approach is a more unified
>> story and also friendly to user and current Flink framework.
>> I improved the FLIP[1] with the proposed approach and refactored the
>> document organization to make it clear enough.
>>
>> Please let me know if you have any concerns, I’m looking forward your
>> comments.
>>
>>
>> Best
>> Leonard
>>
>> [1]
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> <
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>
>>
>>
>>> 在 2020年8月4日，21:25，Leonard Xu <[hidden email] <mailto:[hidden email]>> 写道：
>>>
>>> Hi, all
>>>
>>> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
>>>
>>> Best
>>> Leonard
>>> [1]
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>> <
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>
>>>
>>>> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:[hidden email]> <mailto:
>> [hidden email] <mailto:[hidden email]>>> 写道：
>>>>
>>>> Hi, Timo
>>>>
>>>> Thanks for you response.
>>>>
>>>>> 1) Naming: Is operation time a good term for this concept? If I read
>> "The operation time is the time when the changes happened in system." or
>> "The system time of DML execution in database", why don't we call it
>> `ChangelogTime` or `SystemTime`? Introducing another terminology of time in
>> Flink should be thought through.
>>>>
>>>> I agree that we should thought through. I have considered the name
>> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on the
>> name.
>>>>
>>>> I proposed `operationTime` because most changelog comes from Database
>> and we always called an action as `operation` rather than `change` in
>> Database, the operation time is  easier to understand  for database users,
>> but it's more like a database terminology.
>>>>
>>>> For `SystemTime`, user may confuse which one does the system in
>> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s not a
>> good name.
>>>>
>>>> `ChangelogTime` is a pretty choice which is more unified with existed
>> terminology `Changelog` and `ChangelogMode`, so let me use `ChangelogTime`
>> and I’ll update the FLIP.
>>>>
>>>>
>>>>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
>> expose the concept of time through the user-level `Row` type? The FLIP does
>> not mention this explictly. I think we can keep it as an internal concept
>> but I just wanted to ask for clarification.
>>>>
>>>> Yes, I want to keep it as an internal concept, we have discussed that
>> changelog time concept should be the third time concept(the other two are
>> event-time and processing-time). It’s not easy for normal users(or to help
>> normal users) understand the three concepts accurately, and I did not find
>> a big enough scenario that user need to touch the changelog time for now,
>> so I tend to do not expose the concept to users.
>>>>
>>>>
>>>> Best,
>>>> Leonard
>>>>
>>>>
>>>>>
>>>>> On 04.08.20 04:58, Leonard Xu wrote:
>>>>>> Thanks Konstantin,
>>>>>> Regarding your questions, hope my comments has address your questions
>> and I also add a few explanation in the FLIP.
>>>>>> Thank you all for the feedback,
>>>>>> It seems everyone involved  in this thread has reached a consensus.
>>>>>> I will start a vote thread  later.
>>>>>> Best,
>>>>>> Leonard
>>>>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:[hidden email]> <mailto:
>> [hidden email] <mailto:[hidden email]>>> 写道：
>>>>>>>
>>>>>>> Thanks Lennard for driving this FLIP.
>>>>>>> Looks good to me.
>>>>>>>
>>>>>>> Best,
>>>>>>> Godfrey
>>>>>>>
>>>>>>> Jark Wu <[hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>> 于2020年8月3日周一
>> 下午12:04写道：
>>>>>>>
>>>>>>>> Thanks Leonard for the great FLIP. I think it is in very good shape.
>>>>>>>> +1 to start a vote.
>>>>>>>>
>>>>>>>> Best,
>>>>>>>> Jark
>>>>>>>>
>>>>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email] <mailto:[hidden email]>
>> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>>>>>>>>
>>>>>>>>> Hi Leonard,
>>>>>>>>>
>>>>>>>>> Thanks for this FLIP!
>>>>>>>>> Looks good from my side.
>>>>>>>>>
>>>>>>>>> Cheers, Fabian
>>>>>>>>>
>>>>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>
>>>>>>>>>> :
>>>>>>>>>
>>>>>>>>>> Hi Leondard,
>>>>>>>>>>
>>>>>>>>>> Thank you for pushing this, I think the updated syntax looks
>> really
>>>>>>>> good
>>>>>>>>>> and the semantics make sense to me.
>>>>>>>>>>
>>>>>>>>>> +1
>>>>>>>>>>
>>>>>>>>>> Seth
>>>>>>>>>>
>>>>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email] <mailto:[hidden email]>
>> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>>>>>>>>>>
>>>>>>>>>>> Hi, Konstantin
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be
>> joined
>>>>>>>> on
>>>>>>>>>> the
>>>>>>>>>>>> PRIMARY KEY attribute, correct?
>>>>>>>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the
>> VIEW
>>>>>>>>>>> definition that defines
>>>>>>>>>>>> whether a event-time or processing time temporal table join is
>>>>>>>> used?
>>>>>>>>>>>
>>>>>>>>>>> I think event-time or processing-time temporal table join
>> depends on
>>>>>>>>> fact
>>>>>>>>>>> table’s time attribute in temporal join rather than from temporal
>>>>>>>> table
>>>>>>>>>>> side, the event-time or processing time in temporal table is just
>>>>>>>> used
>>>>>>>>> to
>>>>>>>>>>> split the validity period of versioned snapshot of temporal
>> table.
>>>>>>>> The
>>>>>>>>>>> processing time attribute is not  necessary for temporal table
>>>>>>>> without
>>>>>>>>>>> version, only the primary key is required, the following VIEW is
>> also
>>>>>>>>>> valid
>>>>>>>>>>> for temporal table without version.
>>>>>>>>>>> CREATE VIEW latest_rates AS
>>>>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the
>> latest
>>>>>>>>>>> version
>>>>>>>>>>> FROM rates
>>>>>>>>>>> GROUP BY currency;                           -- inferred primary
>> key
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always
>> versioned
>>>>>>>> on
>>>>>>>>>>>> operation_time regardless of the lookup table attribute
>> (event-time
>>>>>>>>> or
>>>>>>>>>>>> processing time attribute), correct?
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using the
>>>>>>>>> o.time
>>>>>>>>>>> value to lookup the version of the temporal table.
>>>>>>>>>>> For fact table has the processing time attribute, it means only
>>>>>>>> lookup
>>>>>>>>>> the
>>>>>>>>>>> latest version of temporal table and we can do some optimization
>> in
>>>>>>>>>>> implementation like only keep the latest version.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Best
>>>>>>>>>>> Leonard
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>
>>>>
>>>
>>
>>
>
> --
> Best regards!
> Rui Li

[Rui Li]

Thanks Leonard for the clarifications!

On Mon, Aug 17, 2020 at 9:17 PM Leonard Xu <[hidden email]> wrote:

>
> > But are we still able to track different views of such a
> > table through time, as rows are added/deleted to/from the table?
>
> Yes, in fact we support temporal table from changlog which contains all
> possible message types(INSERT/UPDATE/DELETE).
>
> > For
> > example, suppose I have an append-only table source with event-time and
> PK,
> > will I be allowed to do an event-time temporal join with this table?
> Yes, I list some examples in the doc, the example versioned_rates3  is
> this case exactly.
>
> Best
> Leonard
>
>
> >
> > On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email] <mailto:
> [hidden email]>> wrote:
> >
> >> Hi, all
> >>
> >> After a detailed offline discussion about the temporal table related
> >> concept and behavior, we had a reliable solution and rejected several
> >> alternatives.
> >> Compared to rejected alternatives, the proposed approach is a more
> unified
> >> story and also friendly to user and current Flink framework.
> >> I improved the FLIP[1] with the proposed approach and refactored the
> >> document organization to make it clear enough.
> >>
> >> Please let me know if you have any concerns, I’m looking forward your
> >> comments.
> >>
> >>
> >> Best
> >> Leonard
> >>
> >> [1]
> >>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> >> <
> >>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> >
> >>>
> >>
> >>
> >>> 在 2020年8月4日，21:25，Leonard Xu <[hidden email] <mailto:
> [hidden email]>> 写道：
> >>>
> >>> Hi, all
> >>>
> >>> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
> >>>
> >>> Best
> >>> Leonard
> >>> [1]
> >>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> >
> >> <
> >>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> >
> >>>
> >>>
> >>>> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:
> [hidden email]> <mailto:
> >> [hidden email] <mailto:[hidden email]>>> 写道：
> >>>>
> >>>> Hi, Timo
> >>>>
> >>>> Thanks for you response.
> >>>>
> >>>>> 1) Naming: Is operation time a good term for this concept? If I read
> >> "The operation time is the time when the changes happened in system." or
> >> "The system time of DML execution in database", why don't we call it
> >> `ChangelogTime` or `SystemTime`? Introducing another terminology of
> time in
> >> Flink should be thought through.
> >>>>
> >>>> I agree that we should thought through. I have considered the name
> >> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on the
> >> name.
> >>>>
> >>>> I proposed `operationTime` because most changelog comes from Database
> >> and we always called an action as `operation` rather than `change` in
> >> Database, the operation time is  easier to understand  for database
> users,
> >> but it's more like a database terminology.
> >>>>
> >>>> For `SystemTime`, user may confuse which one does the system in
> >> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s not a
> >> good name.
> >>>>
> >>>> `ChangelogTime` is a pretty choice which is more unified with existed
> >> terminology `Changelog` and `ChangelogMode`, so let me use
> `ChangelogTime`
> >> and I’ll update the FLIP.
> >>>>
> >>>>
> >>>>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
> >> expose the concept of time through the user-level `Row` type? The FLIP
> does
> >> not mention this explictly. I think we can keep it as an internal
> concept
> >> but I just wanted to ask for clarification.
> >>>>
> >>>> Yes, I want to keep it as an internal concept, we have discussed that
> >> changelog time concept should be the third time concept(the other two
> are
> >> event-time and processing-time). It’s not easy for normal users(or to
> help
> >> normal users) understand the three concepts accurately, and I did not
> find
> >> a big enough scenario that user need to touch the changelog time for
> now,
> >> so I tend to do not expose the concept to users.
> >>>>
> >>>>
> >>>> Best,
> >>>> Leonard
> >>>>
> >>>>
> >>>>>
> >>>>> On 04.08.20 04:58, Leonard Xu wrote:
> >>>>>> Thanks Konstantin,
> >>>>>> Regarding your questions, hope my comments has address your
> questions
> >> and I also add a few explanation in the FLIP.
> >>>>>> Thank you all for the feedback,
> >>>>>> It seems everyone involved  in this thread has reached a consensus.
> >>>>>> I will start a vote thread  later.
> >>>>>> Best,
> >>>>>> Leonard
> >>>>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:
> [hidden email]> <mailto:
> >> [hidden email] <mailto:[hidden email]>>> 写道：
> >>>>>>>
> >>>>>>> Thanks Lennard for driving this FLIP.
> >>>>>>> Looks good to me.
> >>>>>>>
> >>>>>>> Best,
> >>>>>>> Godfrey
> >>>>>>>
> >>>>>>> Jark Wu <[hidden email] <mailto:[hidden email]> <mailto:
> [hidden email] <mailto:[hidden email]>>> 于2020年8月3日周一
> >> 下午12:04写道：
> >>>>>>>
> >>>>>>>> Thanks Leonard for the great FLIP. I think it is in very good
> shape.
> >>>>>>>> +1 to start a vote.
> >>>>>>>>
> >>>>>>>> Best,
> >>>>>>>> Jark
> >>>>>>>>
> >>>>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]
> <mailto:[hidden email]>
> >> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
> >>>>>>>>
> >>>>>>>>> Hi Leonard,
> >>>>>>>>>
> >>>>>>>>> Thanks for this FLIP!
> >>>>>>>>> Looks good from my side.
> >>>>>>>>>
> >>>>>>>>> Cheers, Fabian
> >>>>>>>>>
> >>>>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
> >>>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:
> [hidden email] <mailto:[hidden email]>>
> >>>>>>>>>> :
> >>>>>>>>>
> >>>>>>>>>> Hi Leondard,
> >>>>>>>>>>
> >>>>>>>>>> Thank you for pushing this, I think the updated syntax looks
> >> really
> >>>>>>>> good
> >>>>>>>>>> and the semantics make sense to me.
> >>>>>>>>>>
> >>>>>>>>>> +1
> >>>>>>>>>>
> >>>>>>>>>> Seth
> >>>>>>>>>>
> >>>>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]
> <mailto:[hidden email]>
> >> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
> >>>>>>>>>>
> >>>>>>>>>>> Hi, Konstantin
> >>>>>>>>>>>
> >>>>>>>>>>>>
> >>>>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be
> >> joined
> >>>>>>>> on
> >>>>>>>>>> the
> >>>>>>>>>>>> PRIMARY KEY attribute, correct?
> >>>>>>>>>>> Yes, the PRIMARY KEY would be join key.
> >>>>>>>>>>>
> >>>>>>>>>>>>
> >>>>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the
> >> VIEW
> >>>>>>>>>>> definition that defines
> >>>>>>>>>>>> whether a event-time or processing time temporal table join is
> >>>>>>>> used?
> >>>>>>>>>>>
> >>>>>>>>>>> I think event-time or processing-time temporal table join
> >> depends on
> >>>>>>>>> fact
> >>>>>>>>>>> table’s time attribute in temporal join rather than from
> temporal
> >>>>>>>> table
> >>>>>>>>>>> side, the event-time or processing time in temporal table is
> just
> >>>>>>>> used
> >>>>>>>>> to
> >>>>>>>>>>> split the validity period of versioned snapshot of temporal
> >> table.
> >>>>>>>> The
> >>>>>>>>>>> processing time attribute is not  necessary for temporal table
> >>>>>>>> without
> >>>>>>>>>>> version, only the primary key is required, the following VIEW
> is
> >> also
> >>>>>>>>>> valid
> >>>>>>>>>>> for temporal table without version.
> >>>>>>>>>>> CREATE VIEW latest_rates AS
> >>>>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the
> >> latest
> >>>>>>>>>>> version
> >>>>>>>>>>> FROM rates
> >>>>>>>>>>> GROUP BY currency;                           -- inferred
> primary
> >> key
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>>>
> >>>>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always
> >> versioned
> >>>>>>>> on
> >>>>>>>>>>>> operation_time regardless of the lookup table attribute
> >> (event-time
> >>>>>>>>> or
> >>>>>>>>>>>> processing time attribute), correct?
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using
> the
> >>>>>>>>> o.time
> >>>>>>>>>>> value to lookup the version of the temporal table.
> >>>>>>>>>>> For fact table has the processing time attribute, it means only
> >>>>>>>> lookup
> >>>>>>>>>> the
> >>>>>>>>>>> latest version of temporal table and we can do some
> optimization
> >> in
> >>>>>>>>>>> implementation like only keep the latest version.
> >>>>>>>>>>>
> >>>>>>>>>>>
> >>>>>>>>>>> Best
> >>>>>>>>>>> Leonard
> >>>>>>>>>>
> >>>>>>>>>
> >>>>>>>>
> >>>>>
> >>>>
> >>>
> >>
> >>
> >
> > --
> > Best regards!
> > Rui Li
>
>

--
Best regards!
Rui Li

[Fabian Hueske-2]

Thanks for the updated FLIP Leonard!
In my opinion this was an improvement.
So +1 for this design.

I have just one remark regarding the terminology.
I find the term "Temporal Table without Version" somewhat confusing.
IMO, versions are the core principle of temporal tables and temporal tables
without versions don't make much sense to me.

What makes such a table a "Temporal" table? Isn't it just a regular table?
If I understand the proposal correctly, "Temporal Tables without Version"
can only be used in processing time temporal table joins, because this join
only requests the current version.
But all regular tables can be used in processing time (temporal) table
joins as well.
It's basically the same as a lookup join, with the only difference that the
table is maintained in Flink and not accessed in an external system (for
example via JDBC).

Are "Temporal Tables without Version" called "Temporal" because they can be
used in "processing time temporal table joins" and due to its name this
join needs to join something that's called "Temporal"?
In that case, we might want to rename "processing time temporal table
joins" into something else that does not imply a versioning.
Maybe we can call them just lookup joins to avoid introducing another term?

Thanks, Fabian

Am Di., 18. Aug. 2020 um 04:30 Uhr schrieb Rui Li <[hidden email]>:

> Thanks Leonard for the clarifications!
>
> On Mon, Aug 17, 2020 at 9:17 PM Leonard Xu <[hidden email]> wrote:
>
>>
>> > But are we still able to track different views of such a
>> > table through time, as rows are added/deleted to/from the table?
>>
>> Yes, in fact we support temporal table from changlog which contains all
>> possible message types(INSERT/UPDATE/DELETE).
>>
>> > For
>> > example, suppose I have an append-only table source with event-time and
>> PK,
>> > will I be allowed to do an event-time temporal join with this table?
>> Yes, I list some examples in the doc, the example versioned_rates3  is
>> this case exactly.
>>
>> Best
>> Leonard
>>
>>
>> >
>> > On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email] <mailto:
>> [hidden email]>> wrote:
>> >
>> >> Hi, all
>> >>
>> >> After a detailed offline discussion about the temporal table related
>> >> concept and behavior, we had a reliable solution and rejected several
>> >> alternatives.
>> >> Compared to rejected alternatives, the proposed approach is a more
>> unified
>> >> story and also friendly to user and current Flink framework.
>> >> I improved the FLIP[1] with the proposed approach and refactored the
>> >> document organization to make it clear enough.
>> >>
>> >> Please let me know if you have any concerns, I’m looking forward your
>> >> comments.
>> >>
>> >>
>> >> Best
>> >> Leonard
>> >>
>> >> [1]
>> >>
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> >> <
>> >>
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> <
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> >
>> >>>
>> >>
>> >>
>> >>> 在 2020年8月4日，21:25，Leonard Xu <[hidden email] <mailto:
>> [hidden email]>> 写道：
>> >>>
>> >>> Hi, all
>> >>>
>> >>> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
>> >>>
>> >>> Best
>> >>> Leonard
>> >>> [1]
>> >>
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> <
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> >
>> >> <
>> >>
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> <
>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>> >
>> >>>
>> >>>
>> >>>> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:
>> [hidden email]> <mailto:
>> >> [hidden email] <mailto:[hidden email]>>> 写道：
>> >>>>
>> >>>> Hi, Timo
>> >>>>
>> >>>> Thanks for you response.
>> >>>>
>> >>>>> 1) Naming: Is operation time a good term for this concept? If I read
>> >> "The operation time is the time when the changes happened in system."
>> or
>> >> "The system time of DML execution in database", why don't we call it
>> >> `ChangelogTime` or `SystemTime`? Introducing another terminology of
>> time in
>> >> Flink should be thought through.
>> >>>>
>> >>>> I agree that we should thought through. I have considered the name
>> >> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on
>> the
>> >> name.
>> >>>>
>> >>>> I proposed `operationTime` because most changelog comes from Database
>> >> and we always called an action as `operation` rather than `change` in
>> >> Database, the operation time is  easier to understand  for database
>> users,
>> >> but it's more like a database terminology.
>> >>>>
>> >>>> For `SystemTime`, user may confuse which one does the system in
>> >> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s not
>> a
>> >> good name.
>> >>>>
>> >>>> `ChangelogTime` is a pretty choice which is more unified with existed
>> >> terminology `Changelog` and `ChangelogMode`, so let me use
>> `ChangelogTime`
>> >> and I’ll update the FLIP.
>> >>>>
>> >>>>
>> >>>>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
>> >> expose the concept of time through the user-level `Row` type? The FLIP
>> does
>> >> not mention this explictly. I think we can keep it as an internal
>> concept
>> >> but I just wanted to ask for clarification.
>> >>>>
>> >>>> Yes, I want to keep it as an internal concept, we have discussed that
>> >> changelog time concept should be the third time concept(the other two
>> are
>> >> event-time and processing-time). It’s not easy for normal users(or to
>> help
>> >> normal users) understand the three concepts accurately, and I did not
>> find
>> >> a big enough scenario that user need to touch the changelog time for
>> now,
>> >> so I tend to do not expose the concept to users.
>> >>>>
>> >>>>
>> >>>> Best,
>> >>>> Leonard
>> >>>>
>> >>>>
>> >>>>>
>> >>>>> On 04.08.20 04:58, Leonard Xu wrote:
>> >>>>>> Thanks Konstantin,
>> >>>>>> Regarding your questions, hope my comments has address your
>> questions
>> >> and I also add a few explanation in the FLIP.
>> >>>>>> Thank you all for the feedback,
>> >>>>>> It seems everyone involved  in this thread has reached a consensus.
>> >>>>>> I will start a vote thread  later.
>> >>>>>> Best,
>> >>>>>> Leonard
>> >>>>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:
>> [hidden email]> <mailto:
>> >> [hidden email] <mailto:[hidden email]>>> 写道：
>> >>>>>>>
>> >>>>>>> Thanks Lennard for driving this FLIP.
>> >>>>>>> Looks good to me.
>> >>>>>>>
>> >>>>>>> Best,
>> >>>>>>> Godfrey
>> >>>>>>>
>> >>>>>>> Jark Wu <[hidden email] <mailto:[hidden email]> <mailto:
>> [hidden email] <mailto:[hidden email]>>> 于2020年8月3日周一
>> >> 下午12:04写道：
>> >>>>>>>
>> >>>>>>>> Thanks Leonard for the great FLIP. I think it is in very good
>> shape.
>> >>>>>>>> +1 to start a vote.
>> >>>>>>>>
>> >>>>>>>> Best,
>> >>>>>>>> Jark
>> >>>>>>>>
>> >>>>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]
>> <mailto:[hidden email]>
>> >> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>> >>>>>>>>
>> >>>>>>>>> Hi Leonard,
>> >>>>>>>>>
>> >>>>>>>>> Thanks for this FLIP!
>> >>>>>>>>> Looks good from my side.
>> >>>>>>>>>
>> >>>>>>>>> Cheers, Fabian
>> >>>>>>>>>
>> >>>>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>> >>>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:
>> [hidden email] <mailto:[hidden email]>>
>> >>>>>>>>>> :
>> >>>>>>>>>
>> >>>>>>>>>> Hi Leondard,
>> >>>>>>>>>>
>> >>>>>>>>>> Thank you for pushing this, I think the updated syntax looks
>> >> really
>> >>>>>>>> good
>> >>>>>>>>>> and the semantics make sense to me.
>> >>>>>>>>>>
>> >>>>>>>>>> +1
>> >>>>>>>>>>
>> >>>>>>>>>> Seth
>> >>>>>>>>>>
>> >>>>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <[hidden email]
>> <mailto:[hidden email]>
>> >> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>> >>>>>>>>>>
>> >>>>>>>>>>> Hi, Konstantin
>> >>>>>>>>>>>
>> >>>>>>>>>>>>
>> >>>>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be
>> >> joined
>> >>>>>>>> on
>> >>>>>>>>>> the
>> >>>>>>>>>>>> PRIMARY KEY attribute, correct?
>> >>>>>>>>>>> Yes, the PRIMARY KEY would be join key.
>> >>>>>>>>>>>
>> >>>>>>>>>>>>
>> >>>>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the
>> >> VIEW
>> >>>>>>>>>>> definition that defines
>> >>>>>>>>>>>> whether a event-time or processing time temporal table join
>> is
>> >>>>>>>> used?
>> >>>>>>>>>>>
>> >>>>>>>>>>> I think event-time or processing-time temporal table join
>> >> depends on
>> >>>>>>>>> fact
>> >>>>>>>>>>> table’s time attribute in temporal join rather than from
>> temporal
>> >>>>>>>> table
>> >>>>>>>>>>> side, the event-time or processing time in temporal table is
>> just
>> >>>>>>>> used
>> >>>>>>>>> to
>> >>>>>>>>>>> split the validity period of versioned snapshot of temporal
>> >> table.
>> >>>>>>>> The
>> >>>>>>>>>>> processing time attribute is not  necessary for temporal table
>> >>>>>>>> without
>> >>>>>>>>>>> version, only the primary key is required, the following VIEW
>> is
>> >> also
>> >>>>>>>>>> valid
>> >>>>>>>>>>> for temporal table without version.
>> >>>>>>>>>>> CREATE VIEW latest_rates AS
>> >>>>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the
>> >> latest
>> >>>>>>>>>>> version
>> >>>>>>>>>>> FROM rates
>> >>>>>>>>>>> GROUP BY currency;                           -- inferred
>> primary
>> >> key
>> >>>>>>>>>>>
>> >>>>>>>>>>>
>> >>>>>>>>>>>>
>> >>>>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always
>> >> versioned
>> >>>>>>>> on
>> >>>>>>>>>>>> operation_time regardless of the lookup table attribute
>> >> (event-time
>> >>>>>>>>> or
>> >>>>>>>>>>>> processing time attribute), correct?
>> >>>>>>>>>>>
>> >>>>>>>>>>>
>> >>>>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is using
>> the
>> >>>>>>>>> o.time
>> >>>>>>>>>>> value to lookup the version of the temporal table.
>> >>>>>>>>>>> For fact table has the processing time attribute, it means
>> only
>> >>>>>>>> lookup
>> >>>>>>>>>> the
>> >>>>>>>>>>> latest version of temporal table and we can do some
>> optimization
>> >> in
>> >>>>>>>>>>> implementation like only keep the latest version.
>> >>>>>>>>>>>
>> >>>>>>>>>>>
>> >>>>>>>>>>> Best
>> >>>>>>>>>>> Leonard
>> >>>>>>>>>>
>> >>>>>>>>>
>> >>>>>>>>
>> >>>>>
>> >>>>
>> >>>
>> >>
>> >>
>> >
>> > --
>> > Best regards!
>> > Rui Li
>>
>>
>
> --
> Best regards!
> Rui Li
>

[Seth Wiesman-4]

+1 to the updated design.

I agree with Fabian that the naming of "temporal table without version" is
a bit confusing but the actual semantics make sense to me. I think just
saying its a Flink managed lookup join makes sense.

Seth

On Tue, Aug 18, 2020 at 3:07 PM Fabian Hueske <[hidden email]> wrote:

> Thanks for the updated FLIP Leonard!
> In my opinion this was an improvement.
> So +1 for this design.
>
> I have just one remark regarding the terminology.
> I find the term "Temporal Table without Version" somewhat confusing.
> IMO, versions are the core principle of temporal tables and temporal
> tables without versions don't make much sense to me.
>
> What makes such a table a "Temporal" table? Isn't it just a regular table?
> If I understand the proposal correctly, "Temporal Tables without Version"
> can only be used in processing time temporal table joins, because this join
> only requests the current version.
> But all regular tables can be used in processing time (temporal) table
> joins as well.
> It's basically the same as a lookup join, with the only difference that
> the table is maintained in Flink and not accessed in an external system
> (for example via JDBC).
>
> Are "Temporal Tables without Version" called "Temporal" because they can
> be used in "processing time temporal table joins" and due to its name this
> join needs to join something that's called "Temporal"?
> In that case, we might want to rename "processing time temporal table
> joins" into something else that does not imply a versioning.
> Maybe we can call them just lookup joins to avoid introducing another term?
>
> Thanks, Fabian
>
> Am Di., 18. Aug. 2020 um 04:30 Uhr schrieb Rui Li <[hidden email]>:
>
>> Thanks Leonard for the clarifications!
>>
>> On Mon, Aug 17, 2020 at 9:17 PM Leonard Xu <[hidden email]> wrote:
>>
>>>
>>> > But are we still able to track different views of such a
>>> > table through time, as rows are added/deleted to/from the table?
>>>
>>> Yes, in fact we support temporal table from changlog which contains all
>>> possible message types(INSERT/UPDATE/DELETE).
>>>
>>> > For
>>> > example, suppose I have an append-only table source with event-time
>>> and PK,
>>> > will I be allowed to do an event-time temporal join with this table?
>>> Yes, I list some examples in the doc, the example versioned_rates3  is
>>> this case exactly.
>>>
>>> Best
>>> Leonard
>>>
>>>
>>> >
>>> > On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email] <mailto:
>>> [hidden email]>> wrote:
>>> >
>>> >> Hi, all
>>> >>
>>> >> After a detailed offline discussion about the temporal table related
>>> >> concept and behavior, we had a reliable solution and rejected several
>>> >> alternatives.
>>> >> Compared to rejected alternatives, the proposed approach is a more
>>> unified
>>> >> story and also friendly to user and current Flink framework.
>>> >> I improved the FLIP[1] with the proposed approach and refactored the
>>> >> document organization to make it clear enough.
>>> >>
>>> >> Please let me know if you have any concerns, I’m looking forward your
>>> >> comments.
>>> >>
>>> >>
>>> >> Best
>>> >> Leonard
>>> >>
>>> >> [1]
>>> >>
>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>> >> <
>>> >>
>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>> <
>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>> >
>>> >>>
>>> >>
>>> >>
>>> >>> 在 2020年8月4日，21:25，Leonard Xu <[hidden email] <mailto:
>>> [hidden email]>> 写道：
>>> >>>
>>> >>> Hi, all
>>> >>>
>>> >>> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
>>> >>>
>>> >>> Best
>>> >>> Leonard
>>> >>> [1]
>>> >>
>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>> <
>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>> >
>>> >> <
>>> >>
>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>> <
>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>> >
>>> >>>
>>> >>>
>>> >>>> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:
>>> [hidden email]> <mailto:
>>> >> [hidden email] <mailto:[hidden email]>>> 写道：
>>> >>>>
>>> >>>> Hi, Timo
>>> >>>>
>>> >>>> Thanks for you response.
>>> >>>>
>>> >>>>> 1) Naming: Is operation time a good term for this concept? If I
>>> read
>>> >> "The operation time is the time when the changes happened in system."
>>> or
>>> >> "The system time of DML execution in database", why don't we call it
>>> >> `ChangelogTime` or `SystemTime`? Introducing another terminology of
>>> time in
>>> >> Flink should be thought through.
>>> >>>>
>>> >>>> I agree that we should thought through. I have considered the name
>>> >> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on
>>> the
>>> >> name.
>>> >>>>
>>> >>>> I proposed `operationTime` because most changelog comes from
>>> Database
>>> >> and we always called an action as `operation` rather than `change` in
>>> >> Database, the operation time is  easier to understand  for database
>>> users,
>>> >> but it's more like a database terminology.
>>> >>>>
>>> >>>> For `SystemTime`, user may confuse which one does the system in
>>> >> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s
>>> not a
>>> >> good name.
>>> >>>>
>>> >>>> `ChangelogTime` is a pretty choice which is more unified with
>>> existed
>>> >> terminology `Changelog` and `ChangelogMode`, so let me use
>>> `ChangelogTime`
>>> >> and I’ll update the FLIP.
>>> >>>>
>>> >>>>
>>> >>>>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
>>> >> expose the concept of time through the user-level `Row` type? The
>>> FLIP does
>>> >> not mention this explictly. I think we can keep it as an internal
>>> concept
>>> >> but I just wanted to ask for clarification.
>>> >>>>
>>> >>>> Yes, I want to keep it as an internal concept, we have discussed
>>> that
>>> >> changelog time concept should be the third time concept(the other two
>>> are
>>> >> event-time and processing-time). It’s not easy for normal users(or to
>>> help
>>> >> normal users) understand the three concepts accurately, and I did not
>>> find
>>> >> a big enough scenario that user need to touch the changelog time for
>>> now,
>>> >> so I tend to do not expose the concept to users.
>>> >>>>
>>> >>>>
>>> >>>> Best,
>>> >>>> Leonard
>>> >>>>
>>> >>>>
>>> >>>>>
>>> >>>>> On 04.08.20 04:58, Leonard Xu wrote:
>>> >>>>>> Thanks Konstantin,
>>> >>>>>> Regarding your questions, hope my comments has address your
>>> questions
>>> >> and I also add a few explanation in the FLIP.
>>> >>>>>> Thank you all for the feedback,
>>> >>>>>> It seems everyone involved  in this thread has reached a
>>> consensus.
>>> >>>>>> I will start a vote thread  later.
>>> >>>>>> Best,
>>> >>>>>> Leonard
>>> >>>>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:
>>> [hidden email]> <mailto:
>>> >> [hidden email] <mailto:[hidden email]>>> 写道：
>>> >>>>>>>
>>> >>>>>>> Thanks Lennard for driving this FLIP.
>>> >>>>>>> Looks good to me.
>>> >>>>>>>
>>> >>>>>>> Best,
>>> >>>>>>> Godfrey
>>> >>>>>>>
>>> >>>>>>> Jark Wu <[hidden email] <mailto:[hidden email]> <mailto:
>>> [hidden email] <mailto:[hidden email]>>> 于2020年8月3日周一
>>> >> 下午12:04写道：
>>> >>>>>>>
>>> >>>>>>>> Thanks Leonard for the great FLIP. I think it is in very good
>>> shape.
>>> >>>>>>>> +1 to start a vote.
>>> >>>>>>>>
>>> >>>>>>>> Best,
>>> >>>>>>>> Jark
>>> >>>>>>>>
>>> >>>>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]
>>> <mailto:[hidden email]>
>>> >> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>>> >>>>>>>>
>>> >>>>>>>>> Hi Leonard,
>>> >>>>>>>>>
>>> >>>>>>>>> Thanks for this FLIP!
>>> >>>>>>>>> Looks good from my side.
>>> >>>>>>>>>
>>> >>>>>>>>> Cheers, Fabian
>>> >>>>>>>>>
>>> >>>>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>> >>>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:
>>> [hidden email] <mailto:[hidden email]>>
>>> >>>>>>>>>> :
>>> >>>>>>>>>
>>> >>>>>>>>>> Hi Leondard,
>>> >>>>>>>>>>
>>> >>>>>>>>>> Thank you for pushing this, I think the updated syntax looks
>>> >> really
>>> >>>>>>>> good
>>> >>>>>>>>>> and the semantics make sense to me.
>>> >>>>>>>>>>
>>> >>>>>>>>>> +1
>>> >>>>>>>>>>
>>> >>>>>>>>>> Seth
>>> >>>>>>>>>>
>>> >>>>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <
>>> [hidden email] <mailto:[hidden email]>
>>> >> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>>> >>>>>>>>>>
>>> >>>>>>>>>>> Hi, Konstantin
>>> >>>>>>>>>>>
>>> >>>>>>>>>>>>
>>> >>>>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be
>>> >> joined
>>> >>>>>>>> on
>>> >>>>>>>>>> the
>>> >>>>>>>>>>>> PRIMARY KEY attribute, correct?
>>> >>>>>>>>>>> Yes, the PRIMARY KEY would be join key.
>>> >>>>>>>>>>>
>>> >>>>>>>>>>>>
>>> >>>>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the
>>> >> VIEW
>>> >>>>>>>>>>> definition that defines
>>> >>>>>>>>>>>> whether a event-time or processing time temporal table join
>>> is
>>> >>>>>>>> used?
>>> >>>>>>>>>>>
>>> >>>>>>>>>>> I think event-time or processing-time temporal table join
>>> >> depends on
>>> >>>>>>>>> fact
>>> >>>>>>>>>>> table’s time attribute in temporal join rather than from
>>> temporal
>>> >>>>>>>> table
>>> >>>>>>>>>>> side, the event-time or processing time in temporal table is
>>> just
>>> >>>>>>>> used
>>> >>>>>>>>> to
>>> >>>>>>>>>>> split the validity period of versioned snapshot of temporal
>>> >> table.
>>> >>>>>>>> The
>>> >>>>>>>>>>> processing time attribute is not  necessary for temporal
>>> table
>>> >>>>>>>> without
>>> >>>>>>>>>>> version, only the primary key is required, the following
>>> VIEW is
>>> >> also
>>> >>>>>>>>>> valid
>>> >>>>>>>>>>> for temporal table without version.
>>> >>>>>>>>>>> CREATE VIEW latest_rates AS
>>> >>>>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the
>>> >> latest
>>> >>>>>>>>>>> version
>>> >>>>>>>>>>> FROM rates
>>> >>>>>>>>>>> GROUP BY currency;                           -- inferred
>>> primary
>>> >> key
>>> >>>>>>>>>>>
>>> >>>>>>>>>>>
>>> >>>>>>>>>>>>
>>> >>>>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always
>>> >> versioned
>>> >>>>>>>> on
>>> >>>>>>>>>>>> operation_time regardless of the lookup table attribute
>>> >> (event-time
>>> >>>>>>>>> or
>>> >>>>>>>>>>>> processing time attribute), correct?
>>> >>>>>>>>>>>
>>> >>>>>>>>>>>
>>> >>>>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is
>>> using the
>>> >>>>>>>>> o.time
>>> >>>>>>>>>>> value to lookup the version of the temporal table.
>>> >>>>>>>>>>> For fact table has the processing time attribute, it means
>>> only
>>> >>>>>>>> lookup
>>> >>>>>>>>>> the
>>> >>>>>>>>>>> latest version of temporal table and we can do some
>>> optimization
>>> >> in
>>> >>>>>>>>>>> implementation like only keep the latest version.
>>> >>>>>>>>>>>
>>> >>>>>>>>>>>
>>> >>>>>>>>>>> Best
>>> >>>>>>>>>>> Leonard
>>> >>>>>>>>>>
>>> >>>>>>>>>
>>> >>>>>>>>
>>> >>>>>
>>> >>>>
>>> >>>
>>> >>
>>> >>
>>> >
>>> > --
>>> > Best regards!
>>> > Rui Li
>>>
>>>
>>
>> --
>> Best regards!
>> Rui Li
>>
>

[Leonard Xu]

Thanks Fabian and Seth for the feedback

I agree the name “temporal table without version” is less accurate because this kind of temporal table has a latest version rather than has no version.
 
How about “Latest-only temporal table” ? The related concept section updated as following:

Temporal Table: Temporal table is a table that evolves over time, rows in temporal table are associated with one or more temporal periods.
    Version: A temporal table can split into a set of versioned table snapshots, the version in table snapshots represents the valid life circle of rows, the start time and the end time of the valid period can be  assigned by users.
    Versioned temporal table: If the row in temporal table can track its history changes and visit its history versions, we called this kind of temporal table as versioned temporal table.
    Latest-only temporal table: If the row in temporal table can only track its latest version, we called this kind of temporal table as latest-only temporal table. The temporal table in lookup join can only track its latest version and thus it's also a latest-only temporal table.

Best
Leonard

> 在 2020年8月19日，04:46，Seth Wiesman <[hidden email]> 写道：
>
> +1 to the updated design.
>
> I agree with Fabian that the naming of "temporal table without version" is
> a bit confusing but the actual semantics make sense to me. I think just
> saying its a Flink managed lookup join makes sense.
>
> Seth
>
> On Tue, Aug 18, 2020 at 3:07 PM Fabian Hueske <[hidden email]> wrote:
>
>> Thanks for the updated FLIP Leonard!
>> In my opinion this was an improvement.
>> So +1 for this design.
>>
>> I have just one remark regarding the terminology.
>> I find the term "Temporal Table without Version" somewhat confusing.
>> IMO, versions are the core principle of temporal tables and temporal
>> tables without versions don't make much sense to me.
>>
>> What makes such a table a "Temporal" table? Isn't it just a regular table?
>> If I understand the proposal correctly, "Temporal Tables without Version"
>> can only be used in processing time temporal table joins, because this join
>> only requests the current version.
>> But all regular tables can be used in processing time (temporal) table
>> joins as well.
>> It's basically the same as a lookup join, with the only difference that
>> the table is maintained in Flink and not accessed in an external system
>> (for example via JDBC).
>>
>> Are "Temporal Tables without Version" called "Temporal" because they can
>> be used in "processing time temporal table joins" and due to its name this
>> join needs to join something that's called "Temporal"?
>> In that case, we might want to rename "processing time temporal table
>> joins" into something else that does not imply a versioning.
>> Maybe we can call them just lookup joins to avoid introducing another term?
>>
>> Thanks, Fabian
>>
>> Am Di., 18. Aug. 2020 um 04:30 Uhr schrieb Rui Li <[hidden email]>:
>>
>>> Thanks Leonard for the clarifications!
>>>
>>> On Mon, Aug 17, 2020 at 9:17 PM Leonard Xu <[hidden email]> wrote:
>>>
>>>>
>>>>> But are we still able to track different views of such a
>>>>> table through time, as rows are added/deleted to/from the table?
>>>>
>>>> Yes, in fact we support temporal table from changlog which contains all
>>>> possible message types(INSERT/UPDATE/DELETE).
>>>>
>>>>> For
>>>>> example, suppose I have an append-only table source with event-time
>>>> and PK,
>>>>> will I be allowed to do an event-time temporal join with this table?
>>>> Yes, I list some examples in the doc, the example versioned_rates3  is
>>>> this case exactly.
>>>>
>>>> Best
>>>> Leonard
>>>>
>>>>
>>>>>
>>>>> On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email] <mailto:
>>>> [hidden email]>> wrote:
>>>>>
>>>>>> Hi, all
>>>>>>
>>>>>> After a detailed offline discussion about the temporal table related
>>>>>> concept and behavior, we had a reliable solution and rejected several
>>>>>> alternatives.
>>>>>> Compared to rejected alternatives, the proposed approach is a more
>>>> unified
>>>>>> story and also friendly to user and current Flink framework.
>>>>>> I improved the FLIP[1] with the proposed approach and refactored the
>>>>>> document organization to make it clear enough.
>>>>>>
>>>>>> Please let me know if you have any concerns, I’m looking forward your
>>>>>> comments.
>>>>>>
>>>>>>
>>>>>> Best
>>>>>> Leonard
>>>>>>
>>>>>> [1]
>>>>>>
>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>>>>> <
>>>>>>
>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>>> <
>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>> 在 2020年8月4日，21:25，Leonard Xu <[hidden email] <mailto:
>>>> [hidden email]>> 写道：
>>>>>>>
>>>>>>> Hi, all
>>>>>>>
>>>>>>> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
>>>>>>>
>>>>>>> Best
>>>>>>> Leonard
>>>>>>> [1]
>>>>>>
>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>>> <
>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>>>>
>>>>>> <
>>>>>>
>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>>> <
>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:
>>>> [hidden email]> <mailto:
>>>>>> [hidden email] <mailto:[hidden email]>>> 写道：
>>>>>>>>
>>>>>>>> Hi, Timo
>>>>>>>>
>>>>>>>> Thanks for you response.
>>>>>>>>
>>>>>>>>> 1) Naming: Is operation time a good term for this concept? If I
>>>> read
>>>>>> "The operation time is the time when the changes happened in system."
>>>> or
>>>>>> "The system time of DML execution in database", why don't we call it
>>>>>> `ChangelogTime` or `SystemTime`? Introducing another terminology of
>>>> time in
>>>>>> Flink should be thought through.
>>>>>>>>
>>>>>>>> I agree that we should thought through. I have considered the name
>>>>>> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on
>>>> the
>>>>>> name.
>>>>>>>>
>>>>>>>> I proposed `operationTime` because most changelog comes from
>>>> Database
>>>>>> and we always called an action as `operation` rather than `change` in
>>>>>> Database, the operation time is  easier to understand  for database
>>>> users,
>>>>>> but it's more like a database terminology.
>>>>>>>>
>>>>>>>> For `SystemTime`, user may confuse which one does the system in
>>>>>> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s
>>>> not a
>>>>>> good name.
>>>>>>>>
>>>>>>>> `ChangelogTime` is a pretty choice which is more unified with
>>>> existed
>>>>>> terminology `Changelog` and `ChangelogMode`, so let me use
>>>> `ChangelogTime`
>>>>>> and I’ll update the FLIP.
>>>>>>>>
>>>>>>>>
>>>>>>>>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
>>>>>> expose the concept of time through the user-level `Row` type? The
>>>> FLIP does
>>>>>> not mention this explictly. I think we can keep it as an internal
>>>> concept
>>>>>> but I just wanted to ask for clarification.
>>>>>>>>
>>>>>>>> Yes, I want to keep it as an internal concept, we have discussed
>>>> that
>>>>>> changelog time concept should be the third time concept(the other two
>>>> are
>>>>>> event-time and processing-time). It’s not easy for normal users(or to
>>>> help
>>>>>> normal users) understand the three concepts accurately, and I did not
>>>> find
>>>>>> a big enough scenario that user need to touch the changelog time for
>>>> now,
>>>>>> so I tend to do not expose the concept to users.
>>>>>>>>
>>>>>>>>
>>>>>>>> Best,
>>>>>>>> Leonard
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 04.08.20 04:58, Leonard Xu wrote:
>>>>>>>>>> Thanks Konstantin,
>>>>>>>>>> Regarding your questions, hope my comments has address your
>>>> questions
>>>>>> and I also add a few explanation in the FLIP.
>>>>>>>>>> Thank you all for the feedback,
>>>>>>>>>> It seems everyone involved  in this thread has reached a
>>>> consensus.
>>>>>>>>>> I will start a vote thread  later.
>>>>>>>>>> Best,
>>>>>>>>>> Leonard
>>>>>>>>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:
>>>> [hidden email]> <mailto:
>>>>>> [hidden email] <mailto:[hidden email]>>> 写道：
>>>>>>>>>>>
>>>>>>>>>>> Thanks Lennard for driving this FLIP.
>>>>>>>>>>> Looks good to me.
>>>>>>>>>>>
>>>>>>>>>>> Best,
>>>>>>>>>>> Godfrey
>>>>>>>>>>>
>>>>>>>>>>> Jark Wu <[hidden email] <mailto:[hidden email]> <mailto:
>>>> [hidden email] <mailto:[hidden email]>>> 于2020年8月3日周一
>>>>>> 下午12:04写道：
>>>>>>>>>>>
>>>>>>>>>>>> Thanks Leonard for the great FLIP. I think it is in very good
>>>> shape.
>>>>>>>>>>>> +1 to start a vote.
>>>>>>>>>>>>
>>>>>>>>>>>> Best,
>>>>>>>>>>>> Jark
>>>>>>>>>>>>
>>>>>>>>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]
>>>> <mailto:[hidden email]>
>>>>>> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> Hi Leonard,
>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks for this FLIP!
>>>>>>>>>>>>> Looks good from my side.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Cheers, Fabian
>>>>>>>>>>>>>
>>>>>>>>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>>>>>>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:
>>>> [hidden email] <mailto:[hidden email]>>
>>>>>>>>>>>>>> :
>>>>>>>>>>>>>
>>>>>>>>>>>>>> Hi Leondard,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thank you for pushing this, I think the updated syntax looks
>>>>>> really
>>>>>>>>>>>> good
>>>>>>>>>>>>>> and the semantics make sense to me.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> +1
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Seth
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <
>>>> [hidden email] <mailto:[hidden email]>
>>>>>> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Hi, Konstantin
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be
>>>>>> joined
>>>>>>>>>>>> on
>>>>>>>>>>>>>> the
>>>>>>>>>>>>>>>> PRIMARY KEY attribute, correct?
>>>>>>>>>>>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the
>>>>>> VIEW
>>>>>>>>>>>>>>> definition that defines
>>>>>>>>>>>>>>>> whether a event-time or processing time temporal table join
>>>> is
>>>>>>>>>>>> used?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> I think event-time or processing-time temporal table join
>>>>>> depends on
>>>>>>>>>>>>> fact
>>>>>>>>>>>>>>> table’s time attribute in temporal join rather than from
>>>> temporal
>>>>>>>>>>>> table
>>>>>>>>>>>>>>> side, the event-time or processing time in temporal table is
>>>> just
>>>>>>>>>>>> used
>>>>>>>>>>>>> to
>>>>>>>>>>>>>>> split the validity period of versioned snapshot of temporal
>>>>>> table.
>>>>>>>>>>>> The
>>>>>>>>>>>>>>> processing time attribute is not  necessary for temporal
>>>> table
>>>>>>>>>>>> without
>>>>>>>>>>>>>>> version, only the primary key is required, the following
>>>> VIEW is
>>>>>> also
>>>>>>>>>>>>>> valid
>>>>>>>>>>>>>>> for temporal table without version.
>>>>>>>>>>>>>>> CREATE VIEW latest_rates AS
>>>>>>>>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the
>>>>>> latest
>>>>>>>>>>>>>>> version
>>>>>>>>>>>>>>> FROM rates
>>>>>>>>>>>>>>> GROUP BY currency;                           -- inferred
>>>> primary
>>>>>> key
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always
>>>>>> versioned
>>>>>>>>>>>> on
>>>>>>>>>>>>>>>> operation_time regardless of the lookup table attribute
>>>>>> (event-time
>>>>>>>>>>>>> or
>>>>>>>>>>>>>>>> processing time attribute), correct?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is
>>>> using the
>>>>>>>>>>>>> o.time
>>>>>>>>>>>>>>> value to lookup the version of the temporal table.
>>>>>>>>>>>>>>> For fact table has the processing time attribute, it means
>>>> only
>>>>>>>>>>>> lookup
>>>>>>>>>>>>>> the
>>>>>>>>>>>>>>> latest version of temporal table and we can do some
>>>> optimization
>>>>>> in
>>>>>>>>>>>>>>> implementation like only keep the latest version.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Best
>>>>>>>>>>>>>>> Leonard
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>> --
>>>>> Best regards!
>>>>> Rui Li
>>>>
>>>>
>>>
>>> --
>>> Best regards!
>>> Rui Li
>>>
>>

[Fabian Hueske-2]

Hi Leonard,

Not sure if "Latest-only temporal table" is much better than "temporal
table without version".
Isn't a table that has only the latest version just a regular table?
Why invent a fancy name for something that is the standard behavior of
(non-temporal) tables in all database systems?

Couldn't we just say that a table with a processing time attribute can be
joined with the current version of any table (not calling it a "Latest-only
temporal table") using `FOR SYSTEM_TIME AS OF`?
As I suggested before, it might make sense to choose a different name for
this join instead of using a fancy name for the tables such that the name
fits the name of the join.

Cheers,
Fabian

Am Mi., 19. Aug. 2020 um 04:24 Uhr schrieb Leonard Xu <[hidden email]>:

> Thanks Fabian and Seth for the feedback
>
> I agree the name “temporal table without version” is less accurate because
> this kind of temporal table has a latest version rather than has no version.
>
> How about “Latest-only temporal table” ? The related concept section
> updated as following:
>
> *Temporal Table:* Temporal table is a table that evolves over time, rows
> in temporal table are associated with one or more temporal periods.
>
> *    Version:* A temporal table can split into a set of versioned table
> snapshots, the *version* in table snapshots represents the valid life
> circle of rows, the start time and the end time of the valid period can be
>  assigned by users.
>
> *    Versioned temporal table*: If the row in temporal table can track
> its history changes and visit its history versions, we called this kind of
> temporal table as versioned temporal table.
>
> *    Latest-only temporal table**:* If the row in temporal table can only
> track its latest version, we called this kind of temporal table as
> latest-only temporal table. The temporal table in lookup join can only
> track its latest version and thus it's also a latest-only temporal table.
>
> Best
> Leonard
>
> 在 2020年8月19日，04:46，Seth Wiesman <[hidden email]> 写道：
>
> +1 to the updated design.
>
> I agree with Fabian that the naming of "temporal table without version" is
> a bit confusing but the actual semantics make sense to me. I think just
> saying its a Flink managed lookup join makes sense.
>
> Seth
>
> On Tue, Aug 18, 2020 at 3:07 PM Fabian Hueske <[hidden email]> wrote:
>
> Thanks for the updated FLIP Leonard!
> In my opinion this was an improvement.
> So +1 for this design.
>
> I have just one remark regarding the terminology.
> I find the term "Temporal Table without Version" somewhat confusing.
> IMO, versions are the core principle of temporal tables and temporal
> tables without versions don't make much sense to me.
>
> What makes such a table a "Temporal" table? Isn't it just a regular table?
> If I understand the proposal correctly, "Temporal Tables without Version"
> can only be used in processing time temporal table joins, because this join
> only requests the current version.
> But all regular tables can be used in processing time (temporal) table
> joins as well.
> It's basically the same as a lookup join, with the only difference that
> the table is maintained in Flink and not accessed in an external system
> (for example via JDBC).
>
> Are "Temporal Tables without Version" called "Temporal" because they can
> be used in "processing time temporal table joins" and due to its name this
> join needs to join something that's called "Temporal"?
> In that case, we might want to rename "processing time temporal table
> joins" into something else that does not imply a versioning.
> Maybe we can call them just lookup joins to avoid introducing another term?
>
> Thanks, Fabian
>
> Am Di., 18. Aug. 2020 um 04:30 Uhr schrieb Rui Li <[hidden email]>:
>
> Thanks Leonard for the clarifications!
>
> On Mon, Aug 17, 2020 at 9:17 PM Leonard Xu <[hidden email]> wrote:
>
>
> But are we still able to track different views of such a
> table through time, as rows are added/deleted to/from the table?
>
>
> Yes, in fact we support temporal table from changlog which contains all
> possible message types(INSERT/UPDATE/DELETE).
>
> For
> example, suppose I have an append-only table source with event-time
>
> and PK,
>
> will I be allowed to do an event-time temporal join with this table?
>
> Yes, I list some examples in the doc, the example versioned_rates3  is
> this case exactly.
>
> Best
> Leonard
>
>
>
> On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email] <mailto:
>
> [hidden email]>> wrote:
>
>
> Hi, all
>
> After a detailed offline discussion about the temporal table related
> concept and behavior, we had a reliable solution and rejected several
> alternatives.
> Compared to rejected alternatives, the proposed approach is a more
>
> unified
>
> story and also friendly to user and current Flink framework.
> I improved the FLIP[1] with the proposed approach and refactored the
> document organization to make it clear enough.
>
> Please let me know if you have any concerns, I’m looking forward your
> comments.
>
>
> Best
> Leonard
>
> [1]
>
>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>
> <
>
>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>
>
>
>
>
> 在 2020年8月4日，21:25，Leonard Xu <[hidden email] <mailto:
>
> [hidden email]>> 写道：
>
>
> Hi, all
>
> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
>
> Best
> Leonard
> [1]
>
>
>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>
>
> <
>
>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
> <
>
> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL
>
>
>
>
> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:
>
> [hidden email]> <mailto:
>
> [hidden email] <mailto:[hidden email]>>> 写道：
>
>
> Hi, Timo
>
> Thanks for you response.
>
> 1) Naming: Is operation time a good term for this concept? If I
>
> read
>
> "The operation time is the time when the changes happened in system."
>
> or
>
> "The system time of DML execution in database", why don't we call it
> `ChangelogTime` or `SystemTime`? Introducing another terminology of
>
> time in
>
> Flink should be thought through.
>
>
> I agree that we should thought through. I have considered the name
>
> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on
>
> the
>
> name.
>
>
> I proposed `operationTime` because most changelog comes from
>
> Database
>
> and we always called an action as `operation` rather than `change` in
> Database, the operation time is  easier to understand  for database
>
> users,
>
> but it's more like a database terminology.
>
>
> For `SystemTime`, user may confuse which one does the system in
>
> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s
>
> not a
>
> good name.
>
>
> `ChangelogTime` is a pretty choice which is more unified with
>
> existed
>
> terminology `Changelog` and `ChangelogMode`, so let me use
>
> `ChangelogTime`
>
> and I’ll update the FLIP.
>
>
>
> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
>
> expose the concept of time through the user-level `Row` type? The
>
> FLIP does
>
> not mention this explictly. I think we can keep it as an internal
>
> concept
>
> but I just wanted to ask for clarification.
>
>
> Yes, I want to keep it as an internal concept, we have discussed
>
> that
>
> changelog time concept should be the third time concept(the other two
>
> are
>
> event-time and processing-time). It’s not easy for normal users(or to
>
> help
>
> normal users) understand the three concepts accurately, and I did not
>
> find
>
> a big enough scenario that user need to touch the changelog time for
>
> now,
>
> so I tend to do not expose the concept to users.
>
>
>
> Best,
> Leonard
>
>
>
> On 04.08.20 04:58, Leonard Xu wrote:
>
> Thanks Konstantin,
> Regarding your questions, hope my comments has address your
>
> questions
>
> and I also add a few explanation in the FLIP.
>
> Thank you all for the feedback,
> It seems everyone involved  in this thread has reached a
>
> consensus.
>
> I will start a vote thread  later.
> Best,
> Leonard
>
> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:
>
> [hidden email]> <mailto:
>
> [hidden email] <mailto:[hidden email]>>> 写道：
>
>
> Thanks Lennard for driving this FLIP.
> Looks good to me.
>
> Best,
> Godfrey
>
> Jark Wu <[hidden email] <mailto:[hidden email]> <mailto:
>
> [hidden email] <mailto:[hidden email]>>> 于2020年8月3日周一
>
> 下午12:04写道：
>
>
> Thanks Leonard for the great FLIP. I think it is in very good
>
> shape.
>
> +1 to start a vote.
>
> Best,
> Jark
>
> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email]
>
> <mailto:[hidden email]>
>
> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>
>
> Hi Leonard,
>
> Thanks for this FLIP!
> Looks good from my side.
>
> Cheers, Fabian
>
> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
> [hidden email] <mailto:[hidden email]> <mailto:
>
> [hidden email] <mailto:[hidden email]>>
>
> :
>
>
> Hi Leondard,
>
> Thank you for pushing this, I think the updated syntax looks
>
> really
>
> good
>
> and the semantics make sense to me.
>
> +1
>
> Seth
>
> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <
>
> [hidden email] <mailto:[hidden email]>
>
> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
>
>
> Hi, Konstantin
>
>
> 1) A  "Versioned Temporal Table DDL on source" can only be
>
> joined
>
> on
>
> the
>
> PRIMARY KEY attribute, correct?
>
> Yes, the PRIMARY KEY would be join key.
>
>
> 2) Isn't it the time attribute in the ORDER BY clause of the
>
> VIEW
>
> definition that defines
>
> whether a event-time or processing time temporal table join
>
> is
>
> used?
>
>
> I think event-time or processing-time temporal table join
>
> depends on
>
> fact
>
> table’s time attribute in temporal join rather than from
>
> temporal
>
> table
>
> side, the event-time or processing time in temporal table is
>
> just
>
> used
>
> to
>
> split the validity period of versioned snapshot of temporal
>
> table.
>
> The
>
> processing time attribute is not  necessary for temporal
>
> table
>
> without
>
> version, only the primary key is required, the following
>
> VIEW is
>
> also
>
> valid
>
> for temporal table without version.
> CREATE VIEW latest_rates AS
> SELECT currency, LAST_VALUE(rate)            -- only keep the
>
> latest
>
> version
> FROM rates
> GROUP BY currency;                           -- inferred
>
> primary
>
> key
>
>
>
>
> 3) A "Versioned Temporal Table DDL on source" is always
>
> versioned
>
> on
>
> operation_time regardless of the lookup table attribute
>
> (event-time
>
> or
>
> processing time attribute), correct?
>
>
>
> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is
>
> using the
>
> o.time
>
> value to lookup the version of the temporal table.
> For fact table has the processing time attribute, it means
>
> only
>
> lookup
>
> the
>
> latest version of temporal table and we can do some
>
> optimization
>
> in
>
> implementation like only keep the latest version.
>
>
> Best
> Leonard
>
>
>
>
>
>
>
>
>
>
> --
> Best regards!
> Rui Li
>
>
>
>
> --
> Best regards!
> Rui Li
>
>
>
>

[Leonard Xu]

Thanks Fabian


> Isn't a table that has only the latest version just a regular table?
> Why invent a fancy name for something that is the standard behavior of (non-temporal) tables in all database systems?
Yes, just to help user distinguish the difference between versioned temporal table and  latest-only temporal table.

(1)The table after `FOR SYSTEM_TIME AS OF` must be a temporal table,
(2) All Flink table can be placed after `FOR SYSTEM_TIME AS OF` in processing-time temporal join

IMO, all Flink tables are temporal table, versioned temporal table is a kind of temporal table which can visit its history versions,
and the other temporal table which can only track its current version(you called regular table).

So the other regular table is temporal table, the version is the most important concept in temporal table, I’d like to rename this kind of temporal table to help user understand it.
Database system has a similar “current table” in system-versioned temporal table.


> Couldn't we just say that a table with a processing time attribute can be joined with the current version of any table (not calling it a "Latest-only temporal table") using `FOR SYSTEM_TIME AS OF`?
The "any table" is temporal table, and the current version is equal to latest version.


> As I suggested before, it might make sense to choose a different name for this join instead of using a fancy name for the tables such that the name fits the name of the join.
I think lookup join is a physical implementation of Processing-Time temporal table join, we can even offer a lookup join implementation for Event-time temporal table join if the external system support lookup history data.

Versioned temporal table can be used in Event-time temporal table join, all temporal table(including versioned temporal tale and regular table) can be used in Processing-time temporal table join,
lookup join is a kind of physical implementation of  Processing-time temporal table join which lookups the external system’s data.

If we agree regular tables are temporal table, I tend to keep latest-only temporal table to clarify the temporal and version concept in regular table.


Best
Leonard

> Cheers,
> Fabian
>
> Am Mi., 19. Aug. 2020 um 04:24 Uhr schrieb Leonard Xu <[hidden email] <mailto:[hidden email]>>:
> Thanks Fabian and Seth for the feedback
>
> I agree the name “temporal table without version” is less accurate because this kind of temporal table has a latest version rather than has no version.
>  
> How about “Latest-only temporal table” ? The related concept section updated as following:
>
> Temporal Table: Temporal table is a table that evolves over time, rows in temporal table are associated with one or more temporal periods.
>     Version: A temporal table can split into a set of versioned table snapshots, the version in table snapshots represents the valid life circle of rows, the start time and the end time of the valid period can be  assigned by users.
>     Versioned temporal table: If the row in temporal table can track its history changes and visit its history versions, we called this kind of temporal table as versioned temporal table.
>     Latest-only temporal table: If the row in temporal table can only track its latest version, we called this kind of temporal table as latest-only temporal table. The temporal table in lookup join can only track its latest version and thus it's also a latest-only temporal table.
>
> Best
> Leonard
>
>> 在 2020年8月19日，04:46，Seth Wiesman <[hidden email] <mailto:[hidden email]>> 写道：
>>
>> +1 to the updated design.
>>
>> I agree with Fabian that the naming of "temporal table without version" is
>> a bit confusing but the actual semantics make sense to me. I think just
>> saying its a Flink managed lookup join makes sense.
>>
>> Seth
>>
>> On Tue, Aug 18, 2020 at 3:07 PM Fabian Hueske <[hidden email] <mailto:[hidden email]>> wrote:
>>
>>> Thanks for the updated FLIP Leonard!
>>> In my opinion this was an improvement.
>>> So +1 for this design.
>>>
>>> I have just one remark regarding the terminology.
>>> I find the term "Temporal Table without Version" somewhat confusing.
>>> IMO, versions are the core principle of temporal tables and temporal
>>> tables without versions don't make much sense to me.
>>>
>>> What makes such a table a "Temporal" table? Isn't it just a regular table?
>>> If I understand the proposal correctly, "Temporal Tables without Version"
>>> can only be used in processing time temporal table joins, because this join
>>> only requests the current version.
>>> But all regular tables can be used in processing time (temporal) table
>>> joins as well.
>>> It's basically the same as a lookup join, with the only difference that
>>> the table is maintained in Flink and not accessed in an external system
>>> (for example via JDBC).
>>>
>>> Are "Temporal Tables without Version" called "Temporal" because they can
>>> be used in "processing time temporal table joins" and due to its name this
>>> join needs to join something that's called "Temporal"?
>>> In that case, we might want to rename "processing time temporal table
>>> joins" into something else that does not imply a versioning.
>>> Maybe we can call them just lookup joins to avoid introducing another term?
>>>
>>> Thanks, Fabian
>>>
>>> Am Di., 18. Aug. 2020 um 04:30 Uhr schrieb Rui Li <[hidden email] <mailto:[hidden email]>>:
>>>
>>>> Thanks Leonard for the clarifications!
>>>>
>>>> On Mon, Aug 17, 2020 at 9:17 PM Leonard Xu <[hidden email] <mailto:[hidden email]>> wrote:
>>>>
>>>>>
>>>>>> But are we still able to track different views of such a
>>>>>> table through time, as rows are added/deleted to/from the table?
>>>>>
>>>>> Yes, in fact we support temporal table from changlog which contains all
>>>>> possible message types(INSERT/UPDATE/DELETE).
>>>>>
>>>>>> For
>>>>>> example, suppose I have an append-only table source with event-time
>>>>> and PK,
>>>>>> will I be allowed to do an event-time temporal join with this table?
>>>>> Yes, I list some examples in the doc, the example versioned_rates3  is
>>>>> this case exactly.
>>>>>
>>>>> Best
>>>>> Leonard
>>>>>
>>>>>
>>>>>>
>>>>>> On Wed, Aug 12, 2020 at 3:31 PM Leonard Xu <[hidden email] <mailto:[hidden email]> <mailto:
>>>>> [hidden email] <mailto:[hidden email]>>> wrote:
>>>>>>
>>>>>>> Hi, all
>>>>>>>
>>>>>>> After a detailed offline discussion about the temporal table related
>>>>>>> concept and behavior, we had a reliable solution and rejected several
>>>>>>> alternatives.
>>>>>>> Compared to rejected alternatives, the proposed approach is a more
>>>>> unified
>>>>>>> story and also friendly to user and current Flink framework.
>>>>>>> I improved the FLIP[1] with the proposed approach and refactored the
>>>>>>> document organization to make it clear enough.
>>>>>>>
>>>>>>> Please let me know if you have any concerns, I’m looking forward your
>>>>>>> comments.
>>>>>>>
>>>>>>>
>>>>>>> Best
>>>>>>> Leonard
>>>>>>>
>>>>>>> [1]
>>>>>>>
>>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>>>>> <
>>>>>>>
>>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>>> <
>>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> 在 2020年8月4日，21:25，Leonard Xu <[hidden email] <mailto:[hidden email]> <mailto:
>>>>> [hidden email] <mailto:[hidden email]>>> 写道：
>>>>>>>>
>>>>>>>> Hi, all
>>>>>>>>
>>>>>>>> I’ve updated the FLIP[1] with the terminology `ChangelogTime`.
>>>>>>>>
>>>>>>>> Best
>>>>>>>> Leonard
>>>>>>>> [1]
>>>>>>>
>>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>>> <
>>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>>>>
>>>>>>> <
>>>>>>>
>>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>>> <
>>>>> https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL <https://cwiki.apache.org/confluence/display/FLINK/FLIP-132+Temporal+Table+DDL>
>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> 在 2020年8月4日，20:58，Leonard Xu <[hidden email] <mailto:[hidden email]> <mailto:
>>>>> [hidden email] <mailto:[hidden email]>> <mailto:
>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>> 写道：
>>>>>>>>>
>>>>>>>>> Hi, Timo
>>>>>>>>>
>>>>>>>>> Thanks for you response.
>>>>>>>>>
>>>>>>>>>> 1) Naming: Is operation time a good term for this concept? If I
>>>>> read
>>>>>>> "The operation time is the time when the changes happened in system."
>>>>> or
>>>>>>> "The system time of DML execution in database", why don't we call it
>>>>>>> `ChangelogTime` or `SystemTime`? Introducing another terminology of
>>>>> time in
>>>>>>> Flink should be thought through.
>>>>>>>>>
>>>>>>>>> I agree that we should thought through. I have considered the name
>>>>>>> `ChangelogTime` and `SystemTime` too, I don’t have strong opinion on
>>>>> the
>>>>>>> name.
>>>>>>>>>
>>>>>>>>> I proposed `operationTime` because most changelog comes from
>>>>> Database
>>>>>>> and we always called an action as `operation` rather than `change` in
>>>>>>> Database, the operation time is  easier to understand  for database
>>>>> users,
>>>>>>> but it's more like a database terminology.
>>>>>>>>>
>>>>>>>>> For `SystemTime`, user may confuse which one does the system in
>>>>>>> `SystemTime` represents?  Flink, Database or CDC tool.  Maybe it’s
>>>>> not a
>>>>>>> good name.
>>>>>>>>>
>>>>>>>>> `ChangelogTime` is a pretty choice which is more unified with
>>>>> existed
>>>>>>> terminology `Changelog` and `ChangelogMode`, so let me use
>>>>> `ChangelogTime`
>>>>>>> and I’ll update the FLIP.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> 2) Exposing it through `org.apache.flink.types.Row`: Shall we also
>>>>>>> expose the concept of time through the user-level `Row` type? The
>>>>> FLIP does
>>>>>>> not mention this explictly. I think we can keep it as an internal
>>>>> concept
>>>>>>> but I just wanted to ask for clarification.
>>>>>>>>>
>>>>>>>>> Yes, I want to keep it as an internal concept, we have discussed
>>>>> that
>>>>>>> changelog time concept should be the third time concept(the other two
>>>>> are
>>>>>>> event-time and processing-time). It’s not easy for normal users(or to
>>>>> help
>>>>>>> normal users) understand the three concepts accurately, and I did not
>>>>> find
>>>>>>> a big enough scenario that user need to touch the changelog time for
>>>>> now,
>>>>>>> so I tend to do not expose the concept to users.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Best,
>>>>>>>>> Leonard
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> On 04.08.20 04:58, Leonard Xu wrote:
>>>>>>>>>>> Thanks Konstantin,
>>>>>>>>>>> Regarding your questions, hope my comments has address your
>>>>> questions
>>>>>>> and I also add a few explanation in the FLIP.
>>>>>>>>>>> Thank you all for the feedback,
>>>>>>>>>>> It seems everyone involved  in this thread has reached a
>>>>> consensus.
>>>>>>>>>>> I will start a vote thread  later.
>>>>>>>>>>> Best,
>>>>>>>>>>> Leonard
>>>>>>>>>>>> 在 2020年8月3日，19:35，godfrey he <[hidden email] <mailto:[hidden email]> <mailto:
>>>>> [hidden email] <mailto:[hidden email]>> <mailto:
>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>> 写道：
>>>>>>>>>>>>
>>>>>>>>>>>> Thanks Lennard for driving this FLIP.
>>>>>>>>>>>> Looks good to me.
>>>>>>>>>>>>
>>>>>>>>>>>> Best,
>>>>>>>>>>>> Godfrey
>>>>>>>>>>>>
>>>>>>>>>>>> Jark Wu <[hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>> <mailto:
>>>>> [hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>> 于2020年8月3日周一
>>>>>>> 下午12:04写道：
>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks Leonard for the great FLIP. I think it is in very good
>>>>> shape.
>>>>>>>>>>>>> +1 to start a vote.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Best,
>>>>>>>>>>>>> Jark
>>>>>>>>>>>>>
>>>>>>>>>>>>> On Fri, 31 Jul 2020 at 17:56, Fabian Hueske <[hidden email] <mailto:[hidden email]>
>>>>> <mailto:[hidden email] <mailto:[hidden email]>>
>>>>>>> <mailto:[hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>> wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>>> Hi Leonard,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thanks for this FLIP!
>>>>>>>>>>>>>> Looks good from my side.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Cheers, Fabian
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Am Do., 30. Juli 2020 um 22:15 Uhr schrieb Seth Wiesman <
>>>>>>>>>>>>>> [hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>> <mailto:
>>>>> [hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>
>>>>>>>>>>>>>>> :
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Hi Leondard,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Thank you for pushing this, I think the updated syntax looks
>>>>>>> really
>>>>>>>>>>>>> good
>>>>>>>>>>>>>>> and the semantics make sense to me.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> +1
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Seth
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On Wed, Jul 29, 2020 at 11:36 AM Leonard Xu <
>>>>> [hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>
>>>>>>> <mailto:[hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>> wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Hi, Konstantin
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 1) A  "Versioned Temporal Table DDL on source" can only be
>>>>>>> joined
>>>>>>>>>>>>> on
>>>>>>>>>>>>>>> the
>>>>>>>>>>>>>>>>> PRIMARY KEY attribute, correct?
>>>>>>>>>>>>>>>> Yes, the PRIMARY KEY would be join key.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 2) Isn't it the time attribute in the ORDER BY clause of the
>>>>>>> VIEW
>>>>>>>>>>>>>>>> definition that defines
>>>>>>>>>>>>>>>>> whether a event-time or processing time temporal table join
>>>>> is
>>>>>>>>>>>>> used?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> I think event-time or processing-time temporal table join
>>>>>>> depends on
>>>>>>>>>>>>>> fact
>>>>>>>>>>>>>>>> table’s time attribute in temporal join rather than from
>>>>> temporal
>>>>>>>>>>>>> table
>>>>>>>>>>>>>>>> side, the event-time or processing time in temporal table is
>>>>> just
>>>>>>>>>>>>> used
>>>>>>>>>>>>>> to
>>>>>>>>>>>>>>>> split the validity period of versioned snapshot of temporal
>>>>>>> table.
>>>>>>>>>>>>> The
>>>>>>>>>>>>>>>> processing time attribute is not  necessary for temporal
>>>>> table
>>>>>>>>>>>>> without
>>>>>>>>>>>>>>>> version, only the primary key is required, the following
>>>>> VIEW is
>>>>>>> also
>>>>>>>>>>>>>>> valid
>>>>>>>>>>>>>>>> for temporal table without version.
>>>>>>>>>>>>>>>> CREATE VIEW latest_rates AS
>>>>>>>>>>>>>>>> SELECT currency, LAST_VALUE(rate)            -- only keep the
>>>>>>> latest
>>>>>>>>>>>>>>>> version
>>>>>>>>>>>>>>>> FROM rates
>>>>>>>>>>>>>>>> GROUP BY currency;                           -- inferred
>>>>> primary
>>>>>>> key
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> 3) A "Versioned Temporal Table DDL on source" is always
>>>>>>> versioned
>>>>>>>>>>>>> on
>>>>>>>>>>>>>>>>> operation_time regardless of the lookup table attribute
>>>>>>> (event-time
>>>>>>>>>>>>>> or
>>>>>>>>>>>>>>>>> processing time attribute), correct?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Yes, the semantics of `FOR SYSTEM_TIME AS OF o.time` is
>>>>> using the
>>>>>>>>>>>>>> o.time
>>>>>>>>>>>>>>>> value to lookup the version of the temporal table.
>>>>>>>>>>>>>>>> For fact table has the processing time attribute, it means
>>>>> only
>>>>>>>>>>>>> lookup
>>>>>>>>>>>>>>> the
>>>>>>>>>>>>>>>> latest version of temporal table and we can do some
>>>>> optimization
>>>>>>> in
>>>>>>>>>>>>>>>> implementation like only keep the latest version.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Best
>>>>>>>>>>>>>>>> Leonard
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> --
>>>>>> Best regards!
>>>>>> Rui Li
>>>>>
>>>>>
>>>>
>>>> --
>>>> Best regards!
>>>> Rui Li
>>>>
>>>
>