Composite index

[sinisaos]

Related to #1188. This is almost identical PR as for composite unique constraints, only it doesn't use custom ddl but modified Piccolo methods create_index and drop_index. Although it serves a different purpose, feel free to remove it if it represents too much code duplication and adds complexity, as adding and dropping a composite index can easily be done with manual migrations as described in the comment.

Usage is:

from piccolo.table import Table
from piccolo.columns import Varchar, Boolean
from piccolo.composite_index import CompositeIndex


class Task(Table):
    name = Varchar()
    completed = Boolean(default=False)
    composite_name_completed = CompositeIndex(["name", "completed"])

[dantownsend]

I like the API - using CompositeIndex.

I agree that using strings is just much easier than actual column references.

class Album(Table):
    name = Varchar()
    released = Boolean(default=False)
    name_released_idx = CompositeIndex(["name", "released"])

Is there a way to specify the index type?

It looks like B-tree, GiST, GIN, and BRIN can be used with multi-column indexes:

https://www.postgresql.org/docs/current/indexes-multicolumn.html

[sinisaos]

@dantownsend Great idea. I'll try to implement it. Thanks

[sinisaos]

I added a new commit with the index type in CompositeIndex. I didn't write any test because for that we need to create an extension for each of those index types (except B-tree) either in Github workflows or in PostgresEngine. I tried it locally and it works. The test would be the same as this, except we add e.g. "index_type": IndexMethod.gin instead of the default "index_type": IndexMethod.btree and create the extension with

CREATE EXTENSION pg_trgm; 
CREATE EXTENSION btree_gin; 

for that index type.

[sinisaos]

Tests pass locally without any problems. Maybe the next update with the master branch will be better.