- Added support for xz codec via a (pull request)
- Improved the error reporting / validation for integers, calling out overflows as a separate case from non-integers.
- Add union test for integer/float case
- Add fix for case when int values are presented to a union containing a 'float' schema. Allow the int to be stored using the float schema.
- Add better handling for appending avro records when the file obejct isn't opened using the right mode.
- Fixed bug with the schema resolver where named schemas were not being handled properly
- Version bump to try and fix the docs again
- Deploy Bug: Bug with new pypi deploy led to extension code not being in the package. This release has the extension code in it.
- Pypi requires a full version to update the description, attempting to add the markdown version of the description to pypi
- Fix bug with namespace handling where names with 'dots' in them were still being concatenated with the namespace leading to bogus names
- The array data 'check' function also had a bug where it was not verifying that the datum was a list before attempting to check that all items conformed to the schema
- Fix bug with C implementation of zig zag decoder. Additional unnecessary cast was clipping during the bit shifting for larger numbers.
- Skipping 1.1.8 and 1.1.9 was missing C cythonized code and created incompatibilities with python 2.7
- Fix bug with 'bytes' type in union schemas failing to parse
- Fix bug with reference types (named references) inside unions
- Remove accidental debug loglevel logging directive
- Add more helpful exception messages (mainly for Python 3 with chained exceptions) that will describe which field in a record datum failed and when ints and strings mismatch, show the datum and the schema.
- Fix some old non-py3 incompatible utility code to be py2/py3
- Fix source distribution Cython file inclusion (pull request)
- Add more type checking in the serializer. Some fast data types were leading to spavro not rejecting bad data.
- Add tests to verify that invalid (no schema conforming data) is rejected
- Fix bug with Enum adding it to the named types that can be namespaced.
- Fix bug with 32bit systems that could potentially trucate long data at 2^31 bits
- Add code to support pickling spavro records. This allows the use of spavro in contexts like Spark that need to serialize the data to be shipped around.
- First release of spavro, speedier avro for python!