
After go1.16, go will use module mode by default, even when the repository is checked out under GOPATH or in a one-off directory. Add go.mod, go.sum to keep this repo buildable without opting out of the module mode. > go mod init github.com/mmcgrana/gobyexample > go mod tidy > go mod vendor In module mode, the 'vendor' directory is special and its contents will be actively maintained by the go command. pygments aren't the dependency the go will know about, so it will delete the contents from vendor directory. Move it to `third_party` directory now. And, vendor the blackfriday package. Note: the tutorial contents are not affected by the change in go1.16 because all the examples in this tutorial ask users to run the go command with the explicit list of files to be compiled (e.g. `go run hello-world.go` or `go build command-line-arguments.go`). When the source list is provided, the go command does not have to compute the build list and whether it's running in GOPATH mode or module mode becomes irrelevant.
81 lines
2.8 KiB
Python
81 lines
2.8 KiB
Python
# -*- coding: utf-8 -*-
|
|
"""
|
|
Tests for other lexers
|
|
~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
:copyright: Copyright 2006-2015 by the Pygments team, see AUTHORS.
|
|
:license: BSD, see LICENSE for details.
|
|
"""
|
|
import glob
|
|
import os
|
|
import unittest
|
|
|
|
from pygments.lexers import guess_lexer
|
|
from pygments.lexers.scripting import EasytrieveLexer, JclLexer, RexxLexer
|
|
|
|
def _exampleFilePath(filename):
|
|
return os.path.join(os.path.dirname(__file__), 'examplefiles', filename)
|
|
|
|
|
|
class AnalyseTextTest(unittest.TestCase):
|
|
def _testCanRecognizeAndGuessExampleFiles(self, lexer):
|
|
assert lexer is not None
|
|
|
|
for pattern in lexer.filenames:
|
|
exampleFilesPattern = _exampleFilePath(pattern)
|
|
for exampleFilePath in glob.glob(exampleFilesPattern):
|
|
with open(exampleFilePath, 'rb') as fp:
|
|
text = fp.read().decode('utf-8')
|
|
probability = lexer.analyse_text(text)
|
|
self.assertTrue(probability > 0,
|
|
'%s must recognize %r' % (
|
|
lexer.name, exampleFilePath))
|
|
guessedLexer = guess_lexer(text)
|
|
self.assertEqual(guessedLexer.name, lexer.name)
|
|
|
|
def testCanRecognizeAndGuessExampleFiles(self):
|
|
LEXERS_TO_TEST = [
|
|
EasytrieveLexer,
|
|
JclLexer,
|
|
RexxLexer,
|
|
]
|
|
for lexerToTest in LEXERS_TO_TEST:
|
|
self._testCanRecognizeAndGuessExampleFiles(lexerToTest)
|
|
|
|
|
|
class EasyTrieveLexerTest(unittest.TestCase):
|
|
def testCanGuessFromText(self):
|
|
self.assertLess(0, EasytrieveLexer.analyse_text('MACRO'))
|
|
self.assertLess(0, EasytrieveLexer.analyse_text('\nMACRO'))
|
|
self.assertLess(0, EasytrieveLexer.analyse_text(' \nMACRO'))
|
|
self.assertLess(0, EasytrieveLexer.analyse_text(' \n MACRO'))
|
|
self.assertLess(0, EasytrieveLexer.analyse_text('*\nMACRO'))
|
|
self.assertLess(0, EasytrieveLexer.analyse_text(
|
|
'*\n *\n\n \n*\n MACRO'))
|
|
|
|
|
|
class RexxLexerTest(unittest.TestCase):
|
|
def testCanGuessFromText(self):
|
|
self.assertAlmostEqual(0.01,
|
|
RexxLexer.analyse_text('/* */'))
|
|
self.assertAlmostEqual(1.0,
|
|
RexxLexer.analyse_text('''/* Rexx */
|
|
say "hello world"'''))
|
|
val = RexxLexer.analyse_text('/* */\n'
|
|
'hello:pRoceduRe\n'
|
|
' say "hello world"')
|
|
self.assertTrue(val > 0.5, val)
|
|
val = RexxLexer.analyse_text('''/* */
|
|
if 1 > 0 then do
|
|
say "ok"
|
|
end
|
|
else do
|
|
say "huh?"
|
|
end''')
|
|
self.assertTrue(val > 0.2, val)
|
|
val = RexxLexer.analyse_text('''/* */
|
|
greeting = "hello world!"
|
|
parse value greeting "hello" name "!"
|
|
say name''')
|
|
self.assertTrue(val > 0.2, val)
|