| Index: third_party/pycoverage/coverage/annotate.py
|
| diff --git a/third_party/pycoverage/coverage/annotate.py b/third_party/pycoverage/coverage/annotate.py
|
| new file mode 100644
|
| index 0000000000000000000000000000000000000000..5c396784445cabaa28be0dac7a7316d67bfd7111
|
| --- /dev/null
|
| +++ b/third_party/pycoverage/coverage/annotate.py
|
| @@ -0,0 +1,102 @@
|
| +"""Source file annotation for Coverage."""
|
| +
|
| +import os, re
|
| +
|
| +from coverage.backward import sorted # pylint: disable=W0622
|
| +from coverage.report import Reporter
|
| +
|
| +class AnnotateReporter(Reporter):
|
| + """Generate annotated source files showing line coverage.
|
| +
|
| + This reporter creates annotated copies of the measured source files. Each
|
| + .py file is copied as a .py,cover file, with a left-hand margin annotating
|
| + each line::
|
| +
|
| + > def h(x):
|
| + - if 0: #pragma: no cover
|
| + - pass
|
| + > if x == 1:
|
| + ! a = 1
|
| + > else:
|
| + > a = 2
|
| +
|
| + > h(2)
|
| +
|
| + Executed lines use '>', lines not executed use '!', lines excluded from
|
| + consideration use '-'.
|
| +
|
| + """
|
| +
|
| + def __init__(self, coverage, config):
|
| + super(AnnotateReporter, self).__init__(coverage, config)
|
| + self.directory = None
|
| +
|
| + blank_re = re.compile(r"\s*(#|$)")
|
| + else_re = re.compile(r"\s*else\s*:\s*(#|$)")
|
| +
|
| + def report(self, morfs, directory=None):
|
| + """Run the report.
|
| +
|
| + See `coverage.report()` for arguments.
|
| +
|
| + """
|
| + self.report_files(self.annotate_file, morfs, directory)
|
| +
|
| + def annotate_file(self, cu, analysis):
|
| + """Annotate a single file.
|
| +
|
| + `cu` is the CodeUnit for the file to annotate.
|
| +
|
| + """
|
| + if not cu.relative:
|
| + return
|
| +
|
| + filename = cu.filename
|
| + source = cu.source_file()
|
| + if self.directory:
|
| + dest_file = os.path.join(self.directory, cu.flat_rootname())
|
| + dest_file += ".py,cover"
|
| + else:
|
| + dest_file = filename + ",cover"
|
| + dest = open(dest_file, 'w')
|
| +
|
| + statements = sorted(analysis.statements)
|
| + missing = sorted(analysis.missing)
|
| + excluded = sorted(analysis.excluded)
|
| +
|
| + lineno = 0
|
| + i = 0
|
| + j = 0
|
| + covered = True
|
| + while True:
|
| + line = source.readline()
|
| + if line == '':
|
| + break
|
| + lineno += 1
|
| + while i < len(statements) and statements[i] < lineno:
|
| + i += 1
|
| + while j < len(missing) and missing[j] < lineno:
|
| + j += 1
|
| + if i < len(statements) and statements[i] == lineno:
|
| + covered = j >= len(missing) or missing[j] > lineno
|
| + if self.blank_re.match(line):
|
| + dest.write(' ')
|
| + elif self.else_re.match(line):
|
| + # Special logic for lines containing only 'else:'.
|
| + if i >= len(statements) and j >= len(missing):
|
| + dest.write('! ')
|
| + elif i >= len(statements) or j >= len(missing):
|
| + dest.write('> ')
|
| + elif statements[i] == missing[j]:
|
| + dest.write('! ')
|
| + else:
|
| + dest.write('> ')
|
| + elif lineno in excluded:
|
| + dest.write('- ')
|
| + elif covered:
|
| + dest.write('> ')
|
| + else:
|
| + dest.write('! ')
|
| + dest.write(line)
|
| + source.close()
|
| + dest.close()
|
|
|