Jesse Wilson
Jesse Wilson
JSON and HTTP/1 encoding has a lot of binary to string conversions. We should take inspiration from this: https://johnnylee-sde.github.io/Summary-of-fast-conversion-algorithms/
There should be a break before ‘(native)’: ![Screen Shot 2020-03-04 at 8 46 36 AM](https://user-images.githubusercontent.com/133019/75885551-b6e3b080-5df4-11ea-9f74-5618e6495b70.png) https://square.github.io/okio/2.x/okio/okio/-byte-string/to-byte-string/
Some operations are implemented by creating a conventional ByteString. We should cache this to limit the cost of using a SegmentedByteString.
Lets make a Filesystem implementation that does all the right `fsync` things when you perform a write operation. It should share most of its code with the regular Filesystem implementations....
Because these are implemented segment-at-a-time we have the opportunity for a syscall to succeed on one access and fail on the next. We should write tests to exercise this situation,...
We’ve designed `Path` to support this, but we don’t have tests that exercise it, and I don’t have confidence it’ll actually work. In particular: * Native: Kotlin’s built-in C bindings...
https://blog.jetbrains.com/kotlin/2018/08/kotlin-1-3-m2/
Not sure if this is worthwhile ’cause I don’t have any use case in mind. But we could implement an overload of `Pipe.fold()`that takes a `Source`. This would be called...
Failed on Travis here: https://api.travis-ci.org/v3/job/376414465/log.txt
Formal methods found a bug in TimSort: http://envisage-project.eu/proving-android-java-and-python-sorting-algorithm-is-broken-and-how-to-fix-it/ We could try out this software to check that tricky methods like Buffer.write satisfy their promises, specifically memory use.