# -*- coding: utf-8 -*- # Copyright (C) 2016-2018 RhodeCode GmbH # # This program is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License, version 3 # (only), as published by the Free Software Foundation. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with this program. If not, see <http://www.gnu.org/licenses/>. # # This program is dual-licensed. If you wish to learn more about the # RhodeCode Enterprise Edition, including its added features, Support services, # and proprietary license terms, please see https://rhodecode.com/licenses/ import pytest from pygments.lexers import get_lexer_by_name from rhodecode.tests import no_newline_id_generator from rhodecode.lib.codeblocks import ( tokenize_string, split_token_stream, rollup_tokenstream, render_tokenstream) class TestTokenizeString(object): python_code = ''' import this var = 6 print "this" ''' def test_tokenize_as_python(self): lexer = get_lexer_by_name('python') tokens = list(tokenize_string(self.python_code, lexer)) assert tokens == [ ('', u'\n'), ('', u' '), ('kn', u'import'), ('', u' '), ('nn', u'this'), ('', u'\n'), ('', u'\n'), ('', u' '), ('n', u'var'), ('', u' '), ('o', u'='), ('', u' '), ('mi', u'6'), ('', u'\n'), ('', u' '), ('k', u'print'), ('', u' '), ('s2', u'"'), ('s2', u'this'), ('s2', u'"'), ('', u'\n'), ('', u'\n'), ('', u' ') ] def test_tokenize_as_text(self): lexer = get_lexer_by_name('text') tokens = list(tokenize_string(self.python_code, lexer)) assert tokens == [ ('', u'\n import this\n\n var = 6\n print "this"\n\n ') ] class TestSplitTokenStream(object): def test_split_token_stream(self): lines = list(split_token_stream( [('type1', 'some\ntext'), ('type2', 'more\n')])) assert lines == [ [('type1', u'some')], [('type1', u'text'), ('type2', u'more')], [('type2', u'')], ] def test_split_token_stream_single(self): lines = list(split_token_stream( [('type1', '\n')])) assert lines == [ [('type1', '')], [('type1', '')], ] def test_split_token_stream_single_repeat(self): lines = list(split_token_stream( [('type1', '\n\n\n')])) assert lines == [ [('type1', '')], [('type1', '')], [('type1', '')], [('type1', '')], ] def test_split_token_stream_multiple_repeat(self): lines = list(split_token_stream( [('type1', '\n\n'), ('type2', '\n\n')])) assert lines == [ [('type1', '')], [('type1', '')], [('type1', ''), ('type2', '')], [('type2', '')], [('type2', '')], ] class TestRollupTokens(object): @pytest.mark.parametrize('tokenstream,output', [ ([], []), ([('A', 'hell'), ('A', 'o')], [ ('A', [ ('', 'hello')]), ]), ([('A', 'hell'), ('B', 'o')], [ ('A', [ ('', 'hell')]), ('B', [ ('', 'o')]), ]), ([('A', 'hel'), ('A', 'lo'), ('B', ' '), ('A', 'there')], [ ('A', [ ('', 'hello')]), ('B', [ ('', ' ')]), ('A', [ ('', 'there')]), ]), ]) def test_rollup_tokenstream_without_ops(self, tokenstream, output): assert list(rollup_tokenstream(tokenstream)) == output @pytest.mark.parametrize('tokenstream,output', [ ([], []), ([('A', '', 'hell'), ('A', '', 'o')], [ ('A', [ ('', 'hello')]), ]), ([('A', '', 'hell'), ('B', '', 'o')], [ ('A', [ ('', 'hell')]), ('B', [ ('', 'o')]), ]), ([('A', '', 'h'), ('B', '', 'e'), ('C', '', 'y')], [ ('A', [ ('', 'h')]), ('B', [ ('', 'e')]), ('C', [ ('', 'y')]), ]), ([('A', '', 'h'), ('A', '', 'e'), ('C', '', 'y')], [ ('A', [ ('', 'he')]), ('C', [ ('', 'y')]), ]), ([('A', 'ins', 'h'), ('A', 'ins', 'e')], [ ('A', [ ('ins', 'he') ]), ]), ([('A', 'ins', 'h'), ('A', 'del', 'e')], [ ('A', [ ('ins', 'h'), ('del', 'e') ]), ]), ([('A', 'ins', 'h'), ('B', 'del', 'e'), ('B', 'del', 'y')], [ ('A', [ ('ins', 'h'), ]), ('B', [ ('del', 'ey'), ]), ]), ([('A', 'ins', 'h'), ('A', 'del', 'e'), ('B', 'del', 'y')], [ ('A', [ ('ins', 'h'), ('del', 'e'), ]), ('B', [ ('del', 'y'), ]), ]), ([('A', '', 'some'), ('A', 'ins', 'new'), ('A', '', 'name')], [ ('A', [ ('', 'some'), ('ins', 'new'), ('', 'name'), ]), ]), ]) def test_rollup_tokenstream_with_ops(self, tokenstream, output): assert list(rollup_tokenstream(tokenstream)) == output class TestRenderTokenStream(object): @pytest.mark.parametrize('tokenstream,output', [ ( [], '', ), ( [('', '', u'')], '<span></span>', ), ( [('', '', u'text')], '<span>text</span>', ), ( [('A', '', u'')], '<span class="A"></span>', ), ( [('A', '', u'hello')], '<span class="A">hello</span>', ), ( [('A', '', u'hel'), ('A', '', u'lo')], '<span class="A">hello</span>', ), ( [('A', '', u'two\n'), ('A', '', u'lines')], '<span class="A">two\nlines</span>', ), ( [('A', '', u'\nthree\n'), ('A', '', u'lines')], '<span class="A">\nthree\nlines</span>', ), ( [('', '', u'\n'), ('A', '', u'line')], '<span>\n</span><span class="A">line</span>', ), ( [('', 'ins', u'\n'), ('A', '', u'line')], '<span><ins>\n</ins></span><span class="A">line</span>', ), ( [('A', '', u'hel'), ('A', 'ins', u'lo')], '<span class="A">hel<ins>lo</ins></span>', ), ( [('A', '', u'hel'), ('A', 'ins', u'l'), ('A', 'ins', u'o')], '<span class="A">hel<ins>lo</ins></span>', ), ( [('A', '', u'hel'), ('A', 'ins', u'l'), ('A', 'del', u'o')], '<span class="A">hel<ins>l</ins><del>o</del></span>', ), ( [('A', '', u'hel'), ('B', '', u'lo')], '<span class="A">hel</span><span class="B">lo</span>', ), ( [('A', '', u'hel'), ('B', 'ins', u'lo')], '<span class="A">hel</span><span class="B"><ins>lo</ins></span>', ), ], ids=no_newline_id_generator) def test_render_tokenstream_with_ops(self, tokenstream, output): html = render_tokenstream(tokenstream) assert html == output @pytest.mark.parametrize('tokenstream,output', [ ( [('A', u'hel'), ('A', u'lo')], '<span class="A">hello</span>', ), ( [('A', u'hel'), ('A', u'l'), ('A', u'o')], '<span class="A">hello</span>', ), ( [('A', u'hel'), ('A', u'l'), ('A', u'o')], '<span class="A">hello</span>', ), ( [('A', u'hel'), ('B', u'lo')], '<span class="A">hel</span><span class="B">lo</span>', ), ( [('A', u'hel'), ('B', u'lo')], '<span class="A">hel</span><span class="B">lo</span>', ), ]) def test_render_tokenstream_without_ops(self, tokenstream, output): html = render_tokenstream(tokenstream) assert html == output