Skip to content
This repository has been archived by the owner on Jul 26, 2022. It is now read-only.

Support index-only scans for DATETIME, TIMESTAMP, and DOUBLE #26

Closed
spetrunia opened this issue Feb 2, 2015 · 2 comments
Closed

Support index-only scans for DATETIME, TIMESTAMP, and DOUBLE #26

spetrunia opened this issue Feb 2, 2015 · 2 comments
Assignees
Milestone

Comments

@spetrunia
Copy link

Currently, index scans for DATETIME, TIMESTAMP, and DOUBLE are not supported

Testcase:

create table t31 (pk int auto_increment primary key, key1 double, key(key1)) engine=rocksdb;
insert into t31 values (),(),(),(),(),(),(),();
explain select key1 from t31 where key1=1.234;

create table t32 (pk int auto_increment primary key, key1 datetime, key(key1))engine=rocksdb;
insert into t32 values (),(),(),(),(),(),(),();
explain select key1 from t32 where key1='2015-01-01 00:11:12';

create table t33 (pk int auto_increment primary key, key1 timestamp, key(key1))engine=rocksdb;
insert into t33 values (),(),(),(),(),(),(),();
explain select key1 from t33 where key1='2015-01-01 00:11:12';

This task is about to support them.
DATETIME/TIMESTAMP use Field_temporal_with_date_and_timef::make_sort_key, which just does memcpy().
DOUBLE uses change_double_for_sort(), we will need to code a reverse function.

@spetrunia spetrunia self-assigned this Apr 1, 2015
@spetrunia spetrunia added this to the high-pri milestone Apr 1, 2015
@spetrunia
Copy link
Author

@spetrunia
Copy link
Author

The task of supporting DOUBLE is filed as #56. Other datatypes are fixed.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Projects
None yet
Development

No branches or pull requests

1 participant