LightningDB 0.13.0

Performance Related Improvements

You can see the benchmarks here https://github.com/CoreyKaylor/Lightning.NET/pull/125

• MDBValue becomes more prominent in the API instead of an internal detail.
  • Exposes an AsSpan and a CopyToNewArray function.
  • AsSpan with a buffer pool can avoid any memory duplication and provides a 0-copy read
  • This is obviously a breaking change, but the changes to your application should be pretty straightforward
• MDBResultCode is now returned on all signatures where the native counterpart returns a code
  • BREAKING!!! In your code this means that if you were previously relying on LightningException for control flow, your code will now need to adapt. I think in practice this has made the API easier to work with, but every situation might be different. There are now a set of extension methods ThrowOnError and ThrowOnReadError that provide the same behavior if you need it, but now you'll have to be explicit about it.

Cursor API redesign

• LightningCursor previously took on a .NET api-centric role and there was some lossy translation due to this, in some cases some race conditions that would result in corrupted usage.
  • Now API is much closer to the native lib counterpart
  • There is now an AsEnumerable extension method as an alternative to the cursor itself being enumerable.
  • You can track the method name changes here https://github.com/CoreyKaylor/Lightning.NET/commit/3528a884bf0a421d7d65df92fdbbd89badcacac2

Less notable changes

Some improvements were made to rework some of the tests and add some test coverage that was previously lacking. More can be done, but this is a good step forward.

We'll try to be a little more sensitive to breaking changes going forward and hopefully we're through the bulk of them.