Version 0.25.0

Added support for sequences: see #97
Much thanks to @Stranger6667

Version 0.24.0 Feature Release

Upgrade asyncpg to 0.16.x

Version 0.23.0 Feature Release

Added a way to add a custom json parser. #84

Version 0.21.0 Feature Release

Allow user to specify their own connection class. See #77

Version 0.20.0 Feature Release

Update to use latest asyncpg. Everything else remains unchanged.

Version 0.19.2 Bug fix

Hopefully fixed bug related to #52, where some transactions could drain the pool.

Version 0.19.1 Minor Bug fix

using with with an async context now raises RuntimeError instead of SyntaxError to be consistant with other aio libs.
This is technically not backwards compatible, but since no one should be relying on this behavior, it is not being considered a minor release.

Version 0.19.0 Minor update

This is mostly a bug fix update, but it also includes bumping asyncpg to 0.14 which could potentially break things.

Other than fixing documentation, there are only two changes from 0.18.2

• Bump asyncpg to 0.14
• Check default value if its callable in sqlalchemy parsing (see #73). Thanks to @kamikaze

Version 0.18.2 Bugfix

Allow passing in a custom dialect to actually work. Thanks to @Gr1N , see #58

Version 0.18.0 Major API Change

This version will break everything. It removes the record proxy object, now returning the object returned by asyncpg, meaning that accessing columns is now dictionary bracket notation, and not dot notation.
It also removes the insert function, and adds automatic json parsing.

record.my_column becomes record['my_column']

Full list of changes:

• Dropped asyncpgsa's Record and RecordGenerator in favor of asyncpg's Records and lists, causing dot notation to be replaced with dict notation when accessing properties (row['id'] instead of row.id)
• connection cursor function now uses compile_query so it can handle the same query objects as the other functions like fetchval
• removed the insert function from the SAConnection. SA query objects will need to use query.returning(sa.text('*'))) or the like to get the values you want explicitly, and all inserts will have to move to one of the other methods like fetchval. Plain text queries will need to add ' RETURNING id ' or something similar to the query itself instead of relying it it being added by SAConnection. It should be noted that sqlalchemy does this for you as long as your table definition has a primary key.
• The postgres SA dialact is loaded into the SAConnection class now. This will cause breaking changes when using behaviors that differ based on dialact, such as using JSON column types in SA table definitions. In that case, it will actuall json dumps and loads automatically for you, which will break if you did it manually in your own code.