Optimized/Rebuild LevelDB Native for LevelDBStore

[cschuchardt88]

Change Log

• Optimized LevelDBStore
• Code Cleanup
• Rebuild

[cschuchardt88]

@cschuchardt88 @Jim8y @vncoelho @superboyiii @chenzhitong @AnnaShaleva
Review Please

[vncoelho]

@cschuchardt88 ,I would like to see a benchmark on this.
This is not the first time this part is being refactoring by you, right? Or is this continuation of previous commit?

[cschuchardt88]

@vncoelho This is refactoring everything. Im sure there is performance benefits. But that isn't the point of this PR. This PR fixes dotnet management of the leveldb objects from marshal objects. Also you have access to leveldb classes for easy accessibly.

Example

var options = new Options { CreateIfMissing = true };
using (var db = new DB(options, path))
{
    db.Put("NA", "Na");

    using(var batch = new WriteBatch())
    {
        batch.Delete("NA")
             .Put("Tampa", "Green")
             .Put("London", "red")
             .Put("New York", "blue");
        db.Write(batch);
    }
}

[shargon]

Reviewed 17/36

[shargon]

Move _disposed to beginning?

Suggested change

	if (disposing)
	{
	FreeManagedObjects();
	}
	if (this.Handle != IntPtr.Zero)
	{
	FreeUnManagedObjects();
	this.Handle = IntPtr.Zero;
	}
	_disposed = true;
	_disposed = true;
	if (disposing)
	{
	FreeManagedObjects();
	}
	if (this.Handle != IntPtr.Zero)
	{
	FreeUnManagedObjects();
	this.Handle = IntPtr.Zero;
	}

[cschuchardt88]

_disposed is saying have already disposed objects. So happens after it completes.

[shargon]

Prevents double dispose in multiple threads.

[cschuchardt88]

Moving above all that doesn't do anything either way. Only difference would be you set the _disposed variable early.

[cschuchardt88]

@shargon @Jim8y review please

[shargon]

@cschuchardt88 You never use this. and the code have a lot of this.. Is this code yours?

[cschuchardt88]

@shargon to answer your question I was keeping consistent with existing code in that library.

[AnnaShaleva]

This PR contains a massive set of changes, it's hard to review and some mistake may be done occasionally. There should be a way to split this PR into multiple granular PRs, each PR for its own optimisation. It also would be nice to distinguish optimisations and code refactoring.

Also, every optimisation should be confirmed by benchmarks, otherwise it's controversial. I'm sure there is a way of C# code profiling. Take a look at nspcc-dev/neo-go@cd42b8b, nspcc-dev/neo-go@9712be7, nspcc-dev/neo-go@ffeb3b8 and other optimisation-related PRs in NeoGo.

There shouldn't be TODOs in the resulting code, since every TODO is an issue.

Also, have you tested the resulting code against the current mainnet/testnet? Is the state compatible?

[cschuchardt88]

@AnnaShaleva

Also, have you tested the resulting code against the current mainnet/testnet? Is the state compatible?

Yes

This adds the full features of the leveldb library.

[shargon]

Repository moved to neo, please re-open there