v0.7.20

• Support referencing resources in other projects when creating functions
• Support reading Instance result in binary form
• Bug fixes

v0.7.19

1. Add support for new window functions (cumstd, qcut, nth_value, cume_dist)
2. Upgrade cloudpickle to the latest version to support recursions and more
3. Support quantile(x) in pivot_table
4. Support visiting negative indices in List types in DataFrame
5. Bug fixes

v0.7.18

1. Add fast tunnel for simple types (bigint, double, boolean) for pandas
2. Add unique aggregation support
3. Add keep_nulls to support outer lateral view
4. Make aggregations on pandas backend consistent with ODPS SQL backend
5. Fix internal table ddl
6. Add api_proxy options
7. Add iter_xflow_sub_instances method to simplify fetching xflow sub instances
8. Add support for Python 3.7 wheels
9. Bug & doc fixes

v0.6.10

Fix job submission problem for v0.6.x

v0.5.9

Fix job submission problem for v0.5.x

v0.7.17

1. Make PyODPS compatible with Python 3.7, changes including support for the new method invocation instruction as well as changing async args into async_ to avoid conflicts with the new keyword
2. Support creating DDLs for external tables
3. Support creating table resources referencing other projects
4. Support using project arg on odps.df.func
5. Add English docs
6. Bug fixes

v0.7.16

• Add support for percentile in DataFrame
• Allow using collections in %sql parameters (#58)
• Bug fixes

v0.7.15

• Update cloudpickle. Now supports Python 3.6 and fixes WorkerCrashfile under several scenarios
• Add execution hints for XFlow
• Allow using as_type when constructing DataFrame with composite types with pandas
• Bug fixes

v0.7.14

• Add support for lateral views (see this link for more details)
• Add primitive support for arrays and maps in MaxCompute
• Bug fixes

v0.7.13

• Add method filter_parts to DataFrame, to provide a more straightforward partition representation compatible with MaxCompute representation
• Bug fixes