Re: Table Partitioning Feature - Mailing list pgsql-hackers
| From | Amit Gupta |
|---|---|
| Subject | Re: Table Partitioning Feature |
| Date | |
| Msg-id | 8d79a95c0901230719i503ae925kc3ec9a528d708af3@mail.gmail.com Whole thread Raw |
| In response to | Re: Table Partitioning Feature (Emmanuel Cecchet <manu@frogthinker.org>) |
| Responses |
Re: Table Partitioning Feature
(Emmanuel Cecchet <manu@frogthinker.org>)
|
| List | pgsql-hackers |
Hi Emmanuel,
Please find my comments in-lined:
Please find my comments in-lined:
On 1/23/09, Emmanuel Cecchet <manu@frogthinker.org> wrote:
Sure.
The implementation is planned as follows:
- Partition table commands
++ An intermediate patch in Feb end
++ Final patch in mid March
- Global Index: Mid March
- Optimizer changes for partitioned table: May
We have just started with the implementation, i will post the grammar rules next week.
We are not supporting sub-partitioning - There is just one level of partitioning.
We will definitely consider to implement fast COPY after we are done with the planned tasks.
Thanks,
Amit
Amit,
You might want to put this on the http://wiki.postgresql.org/wiki/Table_partitioning wiki page.
Sure.
How does your timeline look like for this implementation?
The implementation is planned as follows:
- Partition table commands
++ An intermediate patch in Feb end
++ Final patch in mid March
- Global Index: Mid March
- Optimizer changes for partitioned table: May
I would be happy to contribute C triggers to your implementation. From what I understood in http://archives.postgresql.org/pgsql-hackers/2008-01/msg00269.php, you already have an implementation that parses the grammar and generates rules as if someone had written them. Is this code available?
We have just started with the implementation, i will post the grammar rules next week.
Regarding the use of triggers to push/move data to partitions, what if someone declares triggers on partitions? Especially if you have subpartitions, let's consider the case where there is a trigger on the parent, child and grandchild. If I do an insert in the parent, the user trigger on the parent will be executed, then the partition trigger that decides to move to the grandchild. Are we going to bypass the child trigger?
We are not supporting sub-partitioning - There is just one level of partitioning.
If we also want fast COPY operations on partitioned table, we could have an optimized implementation that could bypass triggers and move the tuple directly to the appropriate child table.
We will definitely consider to implement fast COPY after we are done with the planned tasks.
Thanks,
Amit
Thanks for this big contribution,
EmmanuelHi,------------------------------------------------------------------------
We are implementing table partitioning feature to support
- the attached commands. The syntax conforms to most of the suggestion mentioned in http://archives.postgresql.org/pgsql-hackers/2008-01/msg00413.php, barring the following:
-- Specification of partition names is optional. System will be able to generate partition names in such cases.
-- sub partitioning
We are using pgsql triggers to push/move data to appropriate partitions, but we will definitely consider moving to C language triggers as suggested by manu.
- Global non-partitioned indexes (that will extend all the partitions).
- Foreign key support for tables referring to partitioned tables.
Please feel free to post your comments and suggestions.
Thanks,
Amit
Persistent Systems
--
Emmanuel Cecchet
FTO @ Frog Thinker Open Source Development & Consulting
--
Web: http://www.frogthinker.org
email: manu@frogthinker.org
Skype: emmanuel_cecchet
pgsql-hackers by date: