Storage Backends

Both the BinaryClassifier spam filter and NaiveBayes classifier use pluggable storage backends. This page compares all available options.

BinaryClassifier (spam filter) backends

Backend	Class	Persistence	External dependency
RDBMS	ByJG\TextClassifier\Storage\Rdbms	Database	byjg/micro-orm (bundled)
GDBM	ByJG\TextClassifier\Storage\Dba	File	ext-dba PHP extension

NaiveBayes backends

Backend	Class	Persistence	External dependency
Memory	ByJG\TextClassifier\NaiveBayes\Storage\Memory	Optional (JSON file)	None
RDBMS	ByJG\TextClassifier\NaiveBayes\Storage\Rdbms	Database	byjg/anydataset-db (bundled)

Feature comparison

Feature	BC Rdbms	BC Dba	NB Memory	NB Rdbms
Persistent by default	Yes	Yes	No (opt-in via save())	Yes
Multiple process safe	Yes	No	No	Yes
External server required	Optional	No	No	Optional
SQLite support	Yes	—	—	Yes
MySQL support	Yes	—	—	Yes
PostgreSQL support	Yes	—	—	Yes
createDatabase()	Yes	Yes	Not needed	Yes
Schema migrations	Yes	No	No	Yes

Choosing a backend

Use ByJG\TextClassifier\Storage\Rdbms when:

• You have an existing relational database
• Training data needs to survive process restarts
• Multiple processes or servers share the same filter
• You want SQL-level inspection of token data

Use ByJG\TextClassifier\Storage\Dba when:

• No database server is available
• You want a simple, self-contained file
• Single-process only

Use NaiveBayes\Storage\Memory when:

• Prototyping or testing
• Model is trained once and loaded per process
• Low-overhead inference without database connections

Use NaiveBayes\Storage\Rdbms when:

• Model is updated from multiple processes
• You need durable, consistent storage
• Sharing the database with BinaryClassifier

Implementing a custom storage backend

For BinaryClassifier

Implement ByJG\TextClassifier\Storage\StorageInterface. The key methods are:

public function storageOpen(): void;
public function storageClose(): void;
public function storageRetrieve(array|string $tokens): array; // returns Word[]
public function storagePut(Word $word): void;
public function storageUpdate(Word $word): void;
public function storageDel(string $token): void;

Extend ByJG\TextClassifier\Storage\Base to inherit the getInternals(), getTokens(), and processText() implementations.

For NaiveBayes

Implement ByJG\TextClassifier\NaiveBayes\Storage\StorageInterface directly:

public function getCategories(): array;
public function getDocCount(string $category): int;
public function getTotalDocCount(): int;
public function incrementDocCount(string $category): void;
public function decrementDocCount(string $category): void;
public function getTokenCount(string $token, string $category): int;
public function getTotalTokenCount(string $token): int;
public function getTokenCounts(array $tokens): array;
public function incrementToken(string $token, string $category, int $count = 1): void;
public function decrementToken(string $token, string $category, int $count = 1): void;