ソフトウェアトランザクショナルメモリ（STM）

Imagine a bank transfer: you don't want money to disappear or appear out of thin air. You want the whole operation to succeed or fail completely.

Software Transactional Memory (STM) in Clojure provides a similar guarantee for managing shared, mutable state in concurrent programs.

• It's like a mini-database transaction system for your application's memory.
• Ensures that operations on shared data are "all or nothing."

In concurrent programming, multiple parts of your code might try to change the same piece of data at the same time. This can lead to:

• Race conditions: The outcome depends on the unpredictable timing of operations.
• Inconsistent state: Data gets corrupted or partially updated.

STM helps prevent these issues by coordinating access to shared data, ensuring data integrity.

Clojure's STM manages special data containers called Refs. Unlike regular variables, Refs are designed for transactional updates.

• A Ref holds a value that can be changed, but only within a transaction.
• You create a Ref with an initial value using (ref initial-value).

Think of a Ref as a secure vault for your data that only opens for transactions.

To perform operations on Refs, you must wrap them inside a dosync macro.

• dosync defines a transaction block.
• All changes to Refs inside dosync are treated as a single, atomic unit.
• If any part of the transaction fails, all changes are rolled back.

This ensures your data remains consistent, even with concurrent access.

Inside a dosync block, you use alter to change the value of a Ref.

(alter a-ref update-fn & args)

• update-fn is a function applied to the current value of the Ref.
• & args are additional arguments passed to update-fn.
• alter will retry the transaction if a conflict (another transaction modifying the same Ref) is detected.

This example demonstrates how `dosync` and `alter` work together. We'll update a `Ref` multiple times within a transaction.

Sometimes, the order of operations doesn't matter, like adding to a list or counting. For such operations, use commute instead of alter.

(commute a-ref update-fn & args)

• commute marks an update as "commutative."
• This can reduce the chance of transaction retries, improving performance.
• It works best for operations where `(f (f x a) b)` is the same as `(f (f x b) a)`.

Here's an example where multiple threads concurrently increment a counter using `commute`. This highlights how STM manages shared state safely and efficiently.

Clojure's STM provides strong guarantees, often summarized by the ACID acronym:

• Atomicity: All or nothing. A transaction either completes entirely or fails entirely.
• Consistency: Transactions bring data from one valid state to another valid state.
• Isolation: Concurrent transactions appear to execute sequentially, preventing interference.
• Durability: (Less applicable to in-memory STM directly, but implied by successful commit) Once a transaction commits, its changes are permanent.

Consider the following Clojure code snippet:

(def data (ref []))

(defn add-item [item]
  (dosync
    (alter data conj item)))

(add-item 10)
(add-item 20)

(println @data)

What is the final output of (println @data)?

In this lesson, we explored Clojure's powerful Software Transactional Memory (STM) system.

• We learned about Refs for managing shared state.
• The dosync macro ensures atomic transactions.
• alter updates Refs with conflict detection and retries.
• commute optimizes updates for commutative operations.
• STM provides ACID guarantees for robust concurrency.

Next, we'll look at other concurrency primitives like Promises and Futures for asynchronous operations!