A strict and mostly lock-free Software Transactional Memory (STM) for .NET
C#
Latest commit 5d83a07 Mar 21, 2017 @jbakic Version 4.4.4

README.md

Shielded

Available on NuGet.

Shielded is a full-featured implementation of Software Transactional Memory in .NET. It provides a system (the Shield static class) for running in-memory transactions, and data structures which are aware of transactions. It can also generate transaction-aware proxy subclasses based on a POCO class type. The implementation is strict, with strong guarantees on safety. It is mostly lock-free, using only one major lock which is held during the pre-commit check.

Here is a small example:

var n = new Shielded<int>();
int a = n;
Shield.InTransaction(() =>
    n.Value = n + 5);

Shielded fields are thread-safe. You can read them out of transaction, but changes must be done inside. While inside, the library guarantees a consistent view of all shielded fields.

Another example, the STM version of "Hello world!" - parallel addition in an array. Here, in a dictionary:

var dict = new ShieldedDict<int, int>();
ParallelEnumerable.Range(0, 100000)
    .ForAll(i => Shield.InTransaction(() =>
        dict[i % 100] = dict.ContainsKey(i % 100) ? dict[i % 100] + 1 : 1));

Shielded works with value types, and the language automatically does the needed cloning. For ref types, it only makes the reference itself transactional. The class should then be immutable, or, if you have a class you want to make transactional:

public class TestClass {
    public virtual Guid Id { get; set; }
    public virtual string Name { get; set; }
}

Then you create instances like this:

using Shielded.ProxyGen;
...
var t = Factory.NewShielded<TestClass>();

The Factory creates a proxy sub-class, using CodeDom, which will have transactional overrides for all virtual properties of the base class that are public or protected. Due to CodeDom limitations, the getter and setter must have the same accessibility! The proxy objects are thread-safe (or, at least their virtual properties are), and can only be changed inside transactions. Usage is simple:

var id = t.Id;
Shield.InTransaction(() =>
    t.Name = "Test object");

It is safe to execute any number of concurrent transactions that are reading from or writing into the same shielded fields - each transaction will complete correctly. This is accomplished by:

  • ensuring that in one transaction you read a consistent state of all shielded fields
  • buffering writes into storage which is local for each thread

Your changes are commited and made visible to other threads only if all the shielded fields you read or wrote into have not changed since you started. If any have new changes, your transaction is retried from the beginning, but this time reading the new data. Though it may seem so, this cannot create an infinite loop since for any conflict to occur at least one transaction must successfully commit. Overall, the system must make progress.

This quality would place Shielded in the lock-free class of non-blocking concurrency mechanisms, according to academic classification. However, this is not accurate since the commit check gets done under a lock. Hence the word "mostly" in the short description.

Features

  • MVCC: Each transaction reads a consistent snapshot of the state without the need for locking, since updates just create new versions.

    • Old versions are dropped soon after no one is capable of reading them any more.
  • Read-only transactions always complete without any repetitions and without entering the global lock!

  • Strictness: If a write is made anywhere, the system will insist that all touched locations, read or written, still contain the same version of data that they had when the transaction opened. This means it does not suffer from the Write Skew issue.

  • Transactional collections: Included in the library are ShieldedDict<> (dictionary), ShieldedSeq<> (singly linked list) and ShieldedTree<> (a red-black tree implementation).

  • Transaction-local storage: ShieldedLocal<> allows storing anything in the transaction context, visible only from within that transaction.

  • To perform side-effects (IO, and most other operations which are not shielded) you use the SideEffect method of the Shield class, which takes optional onCommit and onRollback lambdas, or the SyncSideEffect method which allows you to execute code during a commit, while the changed fields are still locked.

  • Conditional transactions: Method Shield.Conditional allows you to define something similar to a database AFTER trigger. It receives a test, and an action to perform, both lambdas. It runs the test, makes a note of all shielded objects that the test had accessed, and later re-executes the test when any of those objects is committed into. If test passes, the action is called.

    • Implemented transactionally, so can be called from transactions, and can be triggered by the transaction that created it.
    • Returns an IDisposable for deactivating the subscription, also transactionally. It may even deactivate itself, e.g. to guarantee one-time execution.
  • Pre-commit checks: Shield.PreCommit is very similar to Shield.Conditional, but executes the test within a transaction that changes one of the fields it is interested in, just before that transaction will commit.

    • Can be used to ensure certain invariants are held, or to implement thread prioritization by allowing only some threads which access a field to commit into it.
  • Custom commit operations: You can integrate your own code into the commit process, to execute while the shielded fields, that are being written, are held locked.

    • Already mentioned Shield.SyncSideEffect does this on the level of one transaction.
    • Using Shield.WhenCommitting, you subscribe globally for any commit, or based on the type of field being written. These subscriptions should never throw!
    • Shield.RunToCommit runs a transaction just up to commit, and allows you to commit/rollback later, or from another thread. This is useful for asynchronous programming.
  • Commutables: operations which can be performed without conflict, because they can be reordered in time and have the same net effect, i.e. they are commutable (name borrowed from Clojure). Incrementing an int is an example - if you don’t care what the int’s value is, you can increment it without conflict by simply incrementing whatever value you encounter there at commit time. Using commutes, when appropriate, reduces conflicts and improves concurrency. Incrementing an int, conflict-free:

    n.Commute((ref int a) => a++);
    • Commutes are not performed under any lock, but rather in a special commute subtransaction, which reads the latest data, and tries to commit with the same stamp as your main transaction. If only the commutes fail, then only the commutes get retried.
    • If, in the example above, your main transaction has already (or perhaps will later) read the n field or written to it (non-commutatively), the commute “degenerates” - it gets executed in place, in your transaction, and you can see it’s effect. This means consistency - if you read it, it will stay as read when you commit. But, it is now a potential conflict.
    • Shield has various commutable operations defined in it. Appending to a sequence is commutable - if you do not touch the seq, it never conflicts. Collection Count fields are comuted over, to avoid unnecessary conflicts.