Fujii Masao <masao.fujii@gmail.com> writes:
>>> Though I'm not familiar with CREATE EXTENSION. Why did you exclude 1.0.sql
>>> from DATA? In hstore/Makefile, 1.0.sql is included. You think we should prevent
>>> old version (i.e., 1.0) of pg_stat_statements from being used in 9.2?
>>
>> I'm not sure. My feeling is that we probably don't want to ship all
>> the old scripts forever. People should install the latest version,
>> and use the upgrade scripts to get there if they have an older one.
>> So my gut feeling here is to change hstore to exclude that file rather
>> than adding it here. Any other opinions?
The problem with the hstore scripts is that you had to copy the 1.0
script, change a couple of lines, and call that 1.1, and you also had to
provide the 1.0--1.1 script file.
The solution would be to be able to create hstore 1.1 from 1.0
automatically and I sent over a very simple patch to do that, albeit
after the deadline for the current CF (that's why it's not listed).
Maybe that's simple enough to be considered? (re-attaching here)
b/contrib/hstore/Makefile | 2
b/contrib/hstore/hstore.control | 1
b/src/backend/commands/extension.c | 83 +++--
contrib/hstore/hstore--1.1.sql | 524 -------------------------------------
4 files changed, 57 insertions(+), 553 deletions(-)
> Agreed. But I wonder why VERSION option is usable in CREATE EXTENSION
> if people always should use the latest version. Maybe I'm missing something..
I think not that many people are using 9.1 in production already. Also
bear in mind that the mechanism is not made only for contrib, it makes
sense to ship in-house procedure code as an extension too.
Regards,
--
Dimitri Fontaine
http://2ndQuadrant.fr PostgreSQL : Expertise, Formation et Support