# -*- coding: utf-8 -*- """ Basic JavaLexer Test ~~~~~~~~~~~~~~~~~~~~ :copyright: Copyright 2006-2020 by the Pygments team, see AUTHORS. :license: BSD, see LICENSE for details. """ import pytest from pygments.token import Text, Name, Punctuation, Keyword, Number from pygments.lexers import JavaLexer @pytest.fixture(scope='module') def lexer(): yield JavaLexer() def test_enhanced_for(lexer): fragment = 'label:\nfor(String var2: var1) {}\n' tokens = [ (Name.Label, 'label:'), (Text, '\n'), (Keyword, 'for'), (Punctuation, '('), (Name, 'String'), (Text, ' '), (Name, 'var2'), (Punctuation, ':'), (Text, ' '), (Name, 'var1'), (Punctuation, ')'), (Text, ' '), (Punctuation, '{'), (Punctuation, '}'), (Text, '\n'), ] assert list(lexer.get_tokens(fragment)) == tokens def test_numeric_literals(lexer): fragment = '0 5L 9__542_72l 0xbEEf 0X9_A 0_35 01 0b0___101_0' fragment += ' 0. .7_17F 3e-1_3d 1f 6_01.9e+3 0x.1Fp3 0XEP8D\n' tokens = [ (Number.Integer, '0'), (Text, ' '), (Number.Integer, '5L'), (Text, ' '), (Number.Integer, '9__542_72l'), (Text, ' '), (Number.Hex, '0xbEEf'), (Text, ' '), (Number.Hex, '0X9_A'), (Text, ' '), (Number.Oct, '0_35'), (Text, ' '), (Number.Oct, '01'), (Text, ' '), (Number.Bin, '0b0___101_0'), (Text, ' '), (Number.Float, '0.'), (Text, ' '), (Number.Float, '.7_17F'), (Text, ' '), (Number.Float, '3e-1_3d'), (Text, ' '), (Number.Float, '1f'), (Text, ' '), (Number.Float, '6_01.9e+3'), (Text, ' '), (Number.Float, '0x.1Fp3'), (Text, ' '), (Number.Float, '0XEP8D'), (Text, '\n') ] assert list(lexer.get_tokens(fragment)) == tokens