leerho opened a new issue, #11:
URL: https://github.com/apache/datasketches-tck/issues/11
@tisonkun, thanks for building the TCK. It has already paid off: comparing
the Java and C++ snapshots side by side recently uncovered a real hashing
incompatibility in the array-of-strings tuple sketch
(apache/datasketches-cpp#533).
I'd like to propose one structural change. The language repositories that
consume cross-language snapshots all use the same layout:
| Repository | Layout |
|---|---|
| datasketches-java | `serialization_test_data/java_generated_files/`,
`…/cpp_generated_files/`, `…/go_generated_files/`, `…/rust_generated_files/` |
| datasketches-go | `serialization_test_data/{cpp,go,java}_generated_files/`
|
| datasketches-cpp | moving to
`serialization_test_data/java_generated_files/` (apache/datasketches-cpp#535) |
| datasketches-tck | `serialization/<lang>/snapshots/` |
If the TCK used the same layout
(`serialization_test_data/<lang>_generated_files/`):
- A TCK checkout could be used directly as a language repo's test-data root,
with no renaming or copying logic.
- Java's `tools/download_serialization_test_data.sh` would no longer need to
map `serialization/<lang>/snapshots/` onto
`serialization_test_data/<lang>_generated_files/`.
- Moving files between the TCK and any language repo becomes a plain copy,
and everyone uses one set of paths.
The change would touch the language adapters in
`internal/snapshots/<lang>.go`, the reconciliation and stability code, the
docs, and Java's download script, which only needs its path pattern changed.
What do you think? If you're OK with this plan but pressed for time, let us
know and we can prepare the PRs for these changes.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
