diff options
Diffstat (limited to 'postgres/schema.sql')
| -rw-r--r-- | postgres/schema.sql | 18 |
1 files changed, 10 insertions, 8 deletions
diff --git a/postgres/schema.sql b/postgres/schema.sql index 548e43c..bbc670f 100644 --- a/postgres/schema.sql +++ b/postgres/schema.sql @@ -6,11 +6,13 @@ BEGIN; -- IMPORTANT(fusion): PosgreSQL doesn't have a defined `NOCASE` collation like -- SQLite and doesn't use a default case-insensitive collation like MySQL. The --- simplest alternative is to create a custom collation. --- This is only supported with PostgreSQL 10+ but it should safe to assume that --- it's widely supported, given that it's almost 10 years old and well past its --- end-of-life. --- For more information see: +-- simplest alternative is to create a custom non-deterministic ICU collation. +-- ICU as a collation provider is supported with PostgreSQL 10+, while +-- non-deterministic collations are supported with PostgreSQL 12+. Either way, +-- it should be safe to assume that it's widely supported, given that version +-- 12 is roughly 6 years old and well past its end-of-life. +-- There are also drawbacks from using these types of collations but I won't +-- go into details. For more information see: -- https://www.postgresql.org/docs/current/collation.html CREATE COLLATION NOCASE ( provider = icu, @@ -18,10 +20,10 @@ CREATE COLLATION NOCASE ( locale = 'und-u-ks-level2' ); --- IMPORTANT(fusion): Since we're already assuming PostgreSQL 10+ for collations, +-- IMPORTANT(fusion): Since we're already assuming PostgreSQL 12+ for collations, -- it's probably a good idea to use IDENTITY columns instead of SERIAL. They're --- also supported with PostgreSQL 10+ and are supposed so fix some shortcommings --- of SERIAL. +-- supported with PostgreSQL 10+ and are supposed so fix some shortcommings of +-- SERIAL. -- TODO(fusion): SQLite tables didn't use foreign key constraints so I'm also not -- using them here. It might be a good idea for consistency, but it's not a hard |