Miki
Miki
Generated API is not documented at all in all languages (C++, Java, Python). Because generated API after release 1.2.0 will be stable, it would be nice to document everything properly.
The hash code of an object is not stable when an enumeration is part of it at least in different runs. Lets say you read a blob for a database...
Consider to remove pre-write actions in Python to have `write()` method without any side effect. Consider as well to extend Python runtime library by `serialize()` to call method `initializeOffsets()` automatically...
Consider to remove pre-write actions in Java to have `write()` method without any side effect. Consider as well to extend Java runtime library by `serialize()` and `deserialize()` methods similarly as...
The following schema generates uncompilable Java code: ``` const uint64 NUM_BITS = 3; bitmask bit Permission { NONE = 000b, READ = 010b, WRITE }; ``` The generated Java code...
Currently, there is no test coverage in unit tests for Java runtime library. It would be good to implement it similarly to Python and C++ runtime libraries.
Currently, Java runtime library offers generic solution for reading or writing of Zserio objects (class `zserio.runtime.io.ZserioIO`). To be consistent, it would be good to implement `serialize()` and `deserialize()` methods similarly...
The Python emitter currently uses floor division operator '//' regardless of the operand types. Example: ``` 5 // 2 = 2 (correct) 5.0 // 2 = 2.0 (wrong!) ``` It...
Currently, there is no enconding guide available for Zserio. This is because at the beginning, zserio does not have a wire format. However now, Zserio contains some advanced features like...
Some generators create special source files which are not generated from zserio types directly. Example: C++ generates `InspectorZserioNames.cpp/h` in root package directory. If user specifies zserio type `InspectorZserioNames` in root...