Database Schema

Both the BinaryClassifier spam filter and NaiveBayes RDBMS backends use the same database, managed by byjg/migration. Migrations are located in db/migrations/.

Migration files

File	Version	Description
db/migrations/up/00001.sql	1	Creates tc_wordlist and seeds internal variables
db/migrations/up/00002.sql	2	Creates nb_internals and nb_wordlist
db/migrations/down/00001.sql	—	Drops tc_wordlist
db/migrations/down/00002.sql	—	Drops nb_wordlist and nb_internals

Tables

tc_wordlist — BinaryClassifier spam filter tokens

CREATE TABLE tc_wordlist (
    token      VARCHAR(255) NOT NULL,
    count_ham  INTEGER DEFAULT NULL,
    count_spam INTEGER DEFAULT NULL,
    PRIMARY KEY (token)
);

Column	Type	Description
token	VARCHAR(255)	The word or URI fragment
count_ham	INTEGER	Times this token appeared in ham training texts
count_spam	INTEGER	Times this token appeared in spam training texts

Internal rows (seeded by migration):

token	count_ham	count_spam	Purpose
tc*dbversion	3	NULL	Schema version check
tc*texts	0	0	Total ham / spam text counts

nb_internals — NaiveBayes document counts

CREATE TABLE nb_internals (
    category  VARCHAR(255) NOT NULL,
    doc_count INTEGER NOT NULL DEFAULT 0,
    PRIMARY KEY (category)
);

Column	Type	Description
category	VARCHAR(255)	Category name (user-defined)
doc_count	INTEGER	Number of texts trained under this category

Rows are created on first train() for a category and deleted when all samples are untrained.

nb_wordlist — NaiveBayes token counts

CREATE TABLE nb_wordlist (
    token     VARCHAR(255) NOT NULL,
    category  VARCHAR(255) NOT NULL,
    count     INTEGER NOT NULL DEFAULT 0,
    PRIMARY KEY (token, category)
);

Column	Type	Description
token	VARCHAR(255)	The word token
category	VARCHAR(255)	Category name
count	INTEGER	Cumulative occurrence count of this token in this category

Running migrations

Migrations are applied automatically by createDatabase():

// BinaryClassifier spam filter
$storage = new \ByJG\TextClassifier\Storage\Rdbms($uri, $degenerator);
$storage->createDatabase();

// NaiveBayes
$storage = new \ByJG\TextClassifier\NaiveBayes\Storage\Rdbms($uri);
$storage->createDatabase();

Both methods call Migration::reset(), which:

1. Drops the migration version table (if it exists)
2. Runs all up scripts in numerical order

Warning: createDatabase() is destructive. Do not call it on a database that already contains training data.

Shared database

Both engines can coexist in the same database. Their tables are independent. Run createDatabase() once from either implementation — the migration applies all scripts regardless of which class calls it.