Automatic Index Creation for Column Types
Hi,
I don't see indication that the capability described below exists in
Postgres (or any RDBMS), so this is likely a feature request --
For column types that are frequently defined in tables, and which are
typically indexed, it would be helpful to be able to specify in the type
definition of the column that its addition/creation into a table should
be automatically accompanied by the creation of an associated index.
Auto-index generation for a specific column type would be overridable
for specific tables, as you wouldn't always want to auto-create an index
at the time of column addition to a table. In the case of CREATE TABLE,
this would be less of a concern, and more of a concern for ALTER TABLE
ADD COLUMN.
Such a configuration would also specify the type of index.
What do you think ?
Thanks,
Charles Sheridan
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Hello
2013/8/19 Charles Sheridan <cesheri@swbell.net>
Hi,
I don't see indication that the capability described below exists in
Postgres (or any RDBMS), so this is likely a feature request --For column types that are frequently defined in tables, and which are
typically indexed, it would be helpful to be able to specify in the type
definition of the column that its addition/creation into a table should
be automatically accompanied by the creation of an associated index.Auto-index generation for a specific column type would be overridable
for specific tables, as you wouldn't always want to auto-create an index
at the time of column addition to a table. In the case of CREATE TABLE,
this would be less of a concern, and more of a concern for ALTER TABLE
ADD COLUMN.Such a configuration would also specify the type of index.
What do you think ?
It is relatively dangerous, I think. What is typical type that can be
indexed?
Regards
Pavel
Show quoted text
Thanks,
Charles Sheridan--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/**mailpref/pgsql-hackers<http://www.postgresql.org/mailpref/pgsql-hackers>
In my use, mainly arrays and hstore cols, including arrays of hstores.
These are column types that I suspect benefit from indexes more than
most column types.
Charles
Show quoted text
On 13-08-19 8:14 AM, Pavel Stehule wrote:
Hello
2013/8/19 Charles Sheridan <cesheri@swbell.net
<mailto:cesheri@swbell.net>>Hi,
I don't see indication that the capability described below exists in
Postgres (or any RDBMS), so this is likely a feature request --For column types that are frequently defined in tables, and which are
typically indexed, it would be helpful to be able to specify in
the type
definition of the column that its addition/creation into a table
should
be automatically accompanied by the creation of an associated index.Auto-index generation for a specific column type would be overridable
for specific tables, as you wouldn't always want to auto-create an
index
at the time of column addition to a table. In the case of CREATE
TABLE,
this would be less of a concern, and more of a concern for ALTER TABLE
ADD COLUMN.Such a configuration would also specify the type of index.
What do you think ?
It is relatively dangerous, I think. What is typical type that can be
indexed?Regards
Pavel
Thanks,
Charles Sheridan--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org
<mailto:pgsql-hackers@postgresql.org>)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
2013/8/19 Charles Sheridan <cesheri@swbell.net>
In my use, mainly arrays and hstore cols, including arrays of hstores.
These are column types that I suspect benefit from indexes more than most
column types.
depends on usage
when you use a arrays for time series, then you don't use a index usually
and in hstore, you should to decide between GiST or GIN index. What should
be default?
I don't think so this feature is a good idea. Maybe enhancing a domains
about possibility to specify a index type for domain.
Regards
Pavel
Show quoted text
Charles
On 13-08-19 8:14 AM, Pavel Stehule wrote:
Hello
2013/8/19 Charles Sheridan <cesheri@swbell.net>
Hi,
I don't see indication that the capability described below exists in
Postgres (or any RDBMS), so this is likely a feature request --For column types that are frequently defined in tables, and which are
typically indexed, it would be helpful to be able to specify in the type
definition of the column that its addition/creation into a table should
be automatically accompanied by the creation of an associated index.Auto-index generation for a specific column type would be overridable
for specific tables, as you wouldn't always want to auto-create an index
at the time of column addition to a table. In the case of CREATE TABLE,
this would be less of a concern, and more of a concern for ALTER TABLE
ADD COLUMN.Such a configuration would also specify the type of index.
What do you think ?
It is relatively dangerous, I think. What is typical type that can be
indexed?Regards
Pavel
Thanks,
Charles Sheridan--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
On 08/19/2013 09:10 AM, Charles Sheridan wrote:
Hi,
I don't see indication that the capability described below exists in
Postgres (or any RDBMS), so this is likely a feature request --For column types that are frequently defined in tables, and which are
typically indexed, it would be helpful to be able to specify in the type
definition of the column that its addition/creation into a table should
be automatically accompanied by the creation of an associated index.Auto-index generation for a specific column type would be overridable
for specific tables, as you wouldn't always want to auto-create an index
at the time of column addition to a table. In the case of CREATE TABLE,
this would be less of a concern, and more of a concern for ALTER TABLE
ADD COLUMN.Such a configuration would also specify the type of index.
What do you think ?
I think it's a bad idea. Indexes are not free and it would be a huge and
unwarranted assumption that they are wanted just because some column
type is used. If you want indexes it's up to you to create them. The
only exception is when they are created as an implementation artifact
for a table constraint, but even then you have to express the
constraint, it's not just assumed.
cheers
andrew
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Agree, these are significant concerns. Just to make sure the scope of
the proposal is clear, I'm referring to col types created via CREATE
TYPE, i.e. locally-defined col types of much tighter granularity than
standard col types like INTEGER. Locally, one could adopt a naming
convention for these cols to identify that they have auto-generated
indices.
Best, Charles
On 13-08-19 8:42 AM, Andrew Dunstan wrote:
On 08/19/2013 09:10 AM, Charles Sheridan wrote:
Hi,
I don't see indication that the capability described below exists in
Postgres (or any RDBMS), so this is likely a feature request --For column types that are frequently defined in tables, and which are
typically indexed, it would be helpful to be able to specify in the type
definition of the column that its addition/creation into a table should
be automatically accompanied by the creation of an associated index.Auto-index generation for a specific column type would be overridable
for specific tables, as you wouldn't always want to auto-create an index
at the time of column addition to a table. In the case of CREATE TABLE,
this would be less of a concern, and more of a concern for ALTER TABLE
ADD COLUMN.Such a configuration would also specify the type of index.
What do you think ?
I think it's a bad idea. Indexes are not free and it would be a huge
and unwarranted assumption that they are wanted just because some
column type is used. If you want indexes it's up to you to create
them. The only exception is when they are created as an implementation
artifact for a table constraint, but even then you have to express the
constraint, it's not just assumed.cheers
andrew
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers