Support for other primary key names than `id`?

[metakot]

So I'm creating a new table with varchar primary key which has another name than id.

import piccolo.columns as cl
from piccolo.table import Table

class Currency(Table):
    code = cl.Varchar(primary_key=True)

The migration file created successfully, but executing it results in the following error:

File "asyncpg/protocol/protocol.pyx", line 201, in bind_execute
asyncpg.exceptions.InvalidTableDefinitionError: multiple primary keys for table "currency" are not allowed

Apparently, piccolo adds the default id serial field along with the expressly stated one.

So I removed it from the migration file and the migration executes fine. But after that trying to select this table leads to the following error:

File "asyncpg/protocol/protocol.pyx", line 168, in prepare
asyncpg.exceptions.UndefinedColumnError: column currency.id does not exist

Piccolo still tries to select the id field even if it isn't defined in the model.

So I tried to trick it with db_column_name='id' param. But this doesn't work either:

asyncpg.exceptions.DuplicateColumnError: column "id" specified more than once

[dantownsend]

Hmm, interesting.

It must be something with migrations. I think there may be a problem if you add a table with the default primary key, and then change it to a custom one.

As for selects, if I do something simple like this in the playground it works:

class MyTable(Table):
    pk = Varchar(primary_key=True)

>>> await MyTable.create_table()
>>> await MyTable.insert(MyTable(pk='abc'))
>>> await MyTable.select()
[{'pk': 'abc'}]

So the bug probably isn't with select, but with the creation of the tables.

[metakot]

No, I wasn't changing the primary key, I wanted to create new table with primary key named differently than id. Via migration. And it didn't work.

[dantownsend]

OK, thanks for the clarification. I'll have to setup a test project and try it.