Thread: Use boolean array for nulls parameters
Hi, When I review the [1], I find that the tuple's nulls array use char type. However there are many places use boolean array to repsent the nulls array, so I think we can replace the char type nulls array to boolean type. This change will break the SPI_xxx API, I'm not sure whether this chagnges cause other problems or not. Any thought? [1] - https://www.postgresql.org/message-id/flat/CA+HiwqGkfJfYdeq5vHPh6eqPKjSbfpDDY+j-kXYFePQedtSLeg@mail.gmail.com -- Regrads, Japin Li. ChengDu WenWu Information Technology Co.,Ltd.
Attachment
I personally don't see any benefit in this change. The focus shouldn't be on fixing things that aren't broken. Perhaps, there is more value in using bitmap data type to keep track of NULL values, which is typical storage vs performance debate, and IMHO, it's better to err on using slightly more storage for much better performance. IIRC, the bitmap idea has previously discussed been rejected too.
On Tue, Jan 19, 2021 at 7:07 PM japin <japinli@hotmail.com> wrote:
Hi,
When I review the [1], I find that the tuple's nulls array use char type.
However there are many places use boolean array to repsent the nulls array,
so I think we can replace the char type nulls array to boolean type. This
change will break the SPI_xxx API, I'm not sure whether this chagnges cause
other problems or not. Any thought?
[1] - https://www.postgresql.org/message-id/flat/CA+HiwqGkfJfYdeq5vHPh6eqPKjSbfpDDY+j-kXYFePQedtSLeg@mail.gmail.com
--
Regrads,
Japin Li.
ChengDu WenWu Information Technology Co.,Ltd.
Highgo Software (Canada/China/Pakistan)
URL : www.highgo.ca
ADDR: 10318 WHALLEY BLVD, Surrey, BC
CELL:+923335449950 EMAIL: mailto:hamid.akhtar@highgo.ca
URL : www.highgo.ca
ADDR: 10318 WHALLEY BLVD, Surrey, BC
CELL:+923335449950 EMAIL: mailto:hamid.akhtar@highgo.ca
SKYPE: engineeredvirus
japin <japinli@hotmail.com> writes: > When I review the [1], I find that the tuple's nulls array use char type. > However there are many places use boolean array to repsent the nulls array, > so I think we can replace the char type nulls array to boolean type. This > change will break the SPI_xxx API, I'm not sure whether this chagnges cause > other problems or not. Any thought? We have always considered that changing the APIs of published SPI interfaces is a non-starter. The entire reason those calls still exist at all is for the benefit of third-party extensions. regards, tom lane
On Tue, 19 Jan 2021 at 23:45, Tom Lane <tgl@sss.pgh.pa.us> wrote: > japin <japinli@hotmail.com> writes: >> When I review the [1], I find that the tuple's nulls array use char type. >> However there are many places use boolean array to repsent the nulls array, >> so I think we can replace the char type nulls array to boolean type. This >> change will break the SPI_xxx API, I'm not sure whether this chagnges cause >> other problems or not. Any thought? > > We have always considered that changing the APIs of published SPI > interfaces is a non-starter. The entire reason those calls still > exist at all is for the benefit of third-party extensions. > Thanks for your clarify. I agree that we should keep the APIs stable, maybe we can modify this in someday when the APIs must be changed. -- Regrads, Japin Li. ChengDu WenWu Information Technology Co.,Ltd.