- timestamp-based migrations like rails 2.1:
- replace backend-specific structuredump with a PHP/JSON dump of schema and
data, including a corresponding importer. A simple YAML/JSON structure that
represents the PHP objects (TableDefinition, ColumnDefinition, IndexDefinition
- or alternately writes them out programmatically like a migration script)
would make it easiest to reconstruct identical tables between multiple
machines and multiple database types, and a simple JSON encoding of data (a
single header row, followed by one JSON array per database row per line) would
take care of most special characters and serialization issues, and allow for
fast line-by-line reading.
- add convenience methods of calling $t->string('foo'), instead of
$t->column('string', 'foo'), inside migration scripts. Will need to introspect
the nativeDatabaseTypes() list from the connection for this.
- Handle Horde_Date objects in both input/output from Horde_Db.
- what to do about '0000-00-00' with PostgreSQL?
- integer sizing for mysql like the postgres backend does it?
- port updates in schema_definitions.rb and schema_statements.rb
- allow adding multiple indexes (and columns) at once, in an aggregate
operation. at the abstract level this will just execute the changes one at a
time; for databases like mysql that can run multiple changes at once, it will,
achieving much better efficiency.
- rdo: add the ability to load a table once, then use it repeatedly in a
relationship (or to have it be a static array), instead of joining.
- composite primary key support: what does http://compositekeys.rubyforge.org/
have that we don't?
- port disable_referential_integrity?
- port query_cache.rb
- Support mysqlnd statistics: