- supports `None` (which is translated into `IS NULL`) - pandas requirement is now optional - also supports an async for select thanks to infradust
0.5.0
- added support for different readers - the Pandas() reader uses pandas.read_csv which is more efficient then the previous way to build up a dataframe
0.4.3
This fixes the issue we though we fixed in 0.4.2.
When reading the response we may have empty chunks in the response. The fix was to continue reading the stream instead of stopping (break -> continue).
0.4.2
Fixes a (possible) concurrency issue. We've witnessed an issue in production which must be a concurrency issue, the fix here is the only candidate we can think of.