A fragment is somewhat akin to an inline function: It makes the grammar more readable and easier to maintain.
A fragment will never be counted as a token, it only serves to simplify a grammar.
Consider:
NUMBER: DIGITS | OCTAL_DIGITS | HEX_DIGITS;
fragment DIGITS: '1'..'9' '0'..'9'*;
fragment OCTAL_DIGITS: '0' '0'..'7'+;
fragment HEX_DIGITS: '0x' ('0'..'9' | 'a'..'f' | 'A'..'F')+;
In this example, matching a NUMBER will always return a NUMBER to the lexer, regardless of if it matched "1234", "0xab12", or "0777".
See item 3
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…