Error Codes

The BinaryClassifier engine returns string error constants instead of throwing exceptions when it receives invalid input. Always check return values before using them numerically.

Pattern

use ByJG\TextClassifier\ClassificationResult;

$result = $classifier->classify($text);

if (!($result instanceof ClassificationResult)) {
    // $result is an error code string
    handleError($result);
    return;
}

if ($result->score > 0.8) { /* spam */ }

BinaryClassifier classify() error codes

Constant	String value	Trigger condition
BinaryClassifier::CLASSIFYER_TEXT_MISSING	'CLASSIFYER_TEXT_MISSING'	$text is null
StandardLexer::LEXER_TEXT_NOT_STRING	'LEXER_TEXT_NOT_STRING'	$text is not a string
StandardLexer::LEXER_TEXT_EMPTY	'LEXER_TEXT_EMPTY'	$text is an empty string

BinaryClassifier learn() / unlearn() error codes

Constant	String value	Trigger condition
BinaryClassifier::TRAINER_TEXT_MISSING	'TRAINER_TEXT_MISSING'	$text is null
BinaryClassifier::TRAINER_CATEGORY_MISSING	'TRAINER_CATEGORY_MISSING'	$category is null
BinaryClassifier::TRAINER_CATEGORY_FAIL	'TRAINER_CATEGORY_FAIL'	$category is not BinaryClassifier::SPAM or BinaryClassifier::HAM
StandardLexer::LEXER_TEXT_NOT_STRING	'LEXER_TEXT_NOT_STRING'	$text is not a string
StandardLexer::LEXER_TEXT_EMPTY	'LEXER_TEXT_EMPTY'	$text is an empty string

learn() and unlearn() return null on success.

NaiveBayes error handling

NaiveBayes::classify(), train(), and untrain() do not return error codes. They are void methods (train/untrain) or return an empty array (classify). Invalid input is handled silently:

• Non-string or empty text → lexer returns no tokens → no-op or empty result
• No trained categories → classify() returns null

Lexer error codes

These originate in StandardLexer and are propagated by BinaryClassifier:

Constant	Class	Value
LEXER_TEXT_NOT_STRING	ByJG\TextClassifier\Lexer\StandardLexer	'LEXER_TEXT_NOT_STRING'
LEXER_TEXT_EMPTY	ByJG\TextClassifier\Lexer\StandardLexer	'LEXER_TEXT_EMPTY'

If you implement a custom LexerInterface, you may return any string error code from getTokens(). BinaryClassifier will propagate it as-is.

Defensive usage example

function classifyMessage(BinaryClassifier $classifier, mixed $input): ?float
{
    if (!is_string($input) || $input === '') {
        return null;
    }

    $result = $classifier->classify($input);

    return ($result instanceof ClassificationResult) ? $result->score : null;
}