@Antoine -- on that I'm not sure. Having bidirectional read/write tests for each pair of languages feels like the most secure approach (but maybe not?), so if the runtimes grow too long, we might split up the test matrix into multiple jobs instead of a single monolithic job.
On Sun, Jun 23, 2019 at 5:41 AM Antoine Pitrou <anto...@python.org> wrote: > > > This is extremely good news. > > One question though about our integration testing strategy. It seems > right now we're testing all n² possible combinations (currently n = 4). > That won't scale very far. Perhaps at some point we should switch to a > O(n) testing strategy? For example testing all consecutive pairs along > a bidirectional ring (e.g. "C++ -> Java", "Java -> JS", "JS -> Go", "Go > -> Java", then the other way round). > > Regards > > Antoine. > > > Le 22/06/2019 à 22:06, Wes McKinney a écrit : > > I'm excited to announce that Go has become the 4th language to > > officially participate in the Arrow binary protocol integration tests, > > after Java, C++, and JavaScript: > > > > https://github.com/apache/arrow/commit/4ba2763150459c9eb4139e5954d9b5526b8ef0ee > > > > This is a huge milestone toward making Go a first-class citizen in the > > Apache Arrow world. Congrats to Sebastien, Stuart, Alexandre, and the > > rest of the Go contributors! > > > > - Wes > >
