Pular para o conteúdo principal

Storage: Memory

ByJG\TextClassifier\NaiveBayes\Storage\Memory keeps all data in PHP arrays. It is the simplest storage option — no setup, no dependencies, no files. Data is lost when the process ends unless you explicitly save it.

Setup​

use ByJG\TextClassifier\NaiveBayes\Storage\Memory;

$storage = new Memory();

No constructor arguments, no createDatabase() call required.

Persisting to disk​

Save the trained model to a JSON file:

$storage->save('/var/data/model.json');

Load it back in a subsequent request:

$storage = new Memory();
$storage->load('/var/data/model.json');

The JSON file contains two keys:

{
  "docs":   { "tech": 5, "animals": 3 },
  "tokens": { "programming": { "tech": 4, "animals": 0 } }
}

Typical lifecycle​

// --- Training phase (run once, or periodically) ---
$storage = new Memory();
$nb = new NaiveBayes($storage, $lexer);

$nb->train('PHP is a programming language', 'tech');
$nb->train('The cat sat on the mat', 'animals');
// ... more training ...

$storage->save('/var/data/model.json');

// --- Classification phase (every request) ---
$storage = new Memory();
$storage->load('/var/data/model.json');
$nb = new NaiveBayes($storage, $lexer);

$result = $nb->classify($incomingText);

Concurrent access​

Memory storage has no file locking. If multiple processes load and save the same file, use an external lock or switch to RDBMS storage.

When to use this backend​

  • Prototyping and experimentation
  • Single-request classification where the model is loaded once per process
  • Small models that fit comfortably in memory
  • Scripts, CLI tools, and batch jobs

When to use RDBMS instead​

  • Model is updated frequently from multiple processes
  • Training data is too large to hold in memory
  • Durability and ACID guarantees are required