82e1019edb
In tools/hacking.py Change-Id: I1219dcc140594be97fc680981a122157fed2e279
392 lines
13 KiB
Python
Executable File
392 lines
13 KiB
Python
Executable File
#!/usr/bin/env python
|
|
# vim: tabstop=4 shiftwidth=4 softtabstop=4
|
|
|
|
# Copyright (c) 2012, Cloudscaling
|
|
# All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License"); you may
|
|
# not use this file except in compliance with the License. You may obtain
|
|
# a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
|
|
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
|
|
# License for the specific language governing permissions and limitations
|
|
# under the License.
|
|
|
|
"""nova HACKING file compliance testing
|
|
|
|
built on top of pep8.py
|
|
"""
|
|
|
|
import inspect
|
|
import logging
|
|
import os
|
|
import re
|
|
import sys
|
|
import tokenize
|
|
import traceback
|
|
import warnings
|
|
|
|
import pep8
|
|
|
|
# Don't need this for testing
|
|
logging.disable('LOG')
|
|
|
|
#N1xx comments
|
|
#N2xx except
|
|
#N3xx imports
|
|
#N4xx docstrings
|
|
#N5xx dictionaries/lists
|
|
#N6xx Calling methods
|
|
#N7xx localization
|
|
|
|
IMPORT_EXCEPTIONS = ['sqlalchemy', 'migrate', 'nova.db.sqlalchemy.session']
|
|
DOCSTRING_TRIPLE = ['"""', "'''"]
|
|
VERBOSE_MISSING_IMPORT = False
|
|
|
|
|
|
def is_import_exception(mod):
|
|
return mod in IMPORT_EXCEPTIONS or \
|
|
any(mod.startswith(m + '.') for m in IMPORT_EXCEPTIONS)
|
|
|
|
|
|
def import_normalize(line):
|
|
# convert "from x import y" to "import x.y"
|
|
# handle "from x import y as z" to "import x.y as z"
|
|
split_line = line.split()
|
|
if (line.startswith("from ") and "," not in line and
|
|
split_line[2] == "import" and split_line[3] != "*" and
|
|
split_line[1] != "__future__" and
|
|
(len(split_line) == 4 or
|
|
(len(split_line) == 6 and split_line[4] == "as"))):
|
|
mod = split_line[3]
|
|
return "import %s.%s" % (split_line[1], split_line[3])
|
|
else:
|
|
return line
|
|
|
|
|
|
def nova_todo_format(physical_line):
|
|
"""Check for 'TODO()'.
|
|
|
|
nova HACKING guide recommendation for TODO:
|
|
Include your name with TODOs as in "#TODO(termie)"
|
|
N101
|
|
"""
|
|
pos = physical_line.find('TODO')
|
|
pos1 = physical_line.find('TODO(')
|
|
pos2 = physical_line.find('#') # make sure its a comment
|
|
if (pos != pos1 and pos2 >= 0 and pos2 < pos):
|
|
return pos, "NOVA N101: Use TODO(NAME)"
|
|
|
|
|
|
def nova_except_format(logical_line):
|
|
"""Check for 'except:'.
|
|
|
|
nova HACKING guide recommends not using except:
|
|
Do not write "except:", use "except Exception:" at the very least
|
|
N201
|
|
"""
|
|
if logical_line.startswith("except:"):
|
|
return 6, "NOVA N201: no 'except:' at least use 'except Exception:'"
|
|
|
|
|
|
def nova_except_format_assert(logical_line):
|
|
"""Check for 'assertRaises(Exception'.
|
|
|
|
nova HACKING guide recommends not using assertRaises(Exception...):
|
|
Do not use overly broad Exception type
|
|
N202
|
|
"""
|
|
if logical_line.startswith("self.assertRaises(Exception"):
|
|
return 1, "NOVA N202: assertRaises Exception too broad"
|
|
|
|
|
|
def nova_one_import_per_line(logical_line):
|
|
"""Check for import format.
|
|
|
|
nova HACKING guide recommends one import per line:
|
|
Do not import more than one module per line
|
|
|
|
Examples:
|
|
BAD: from nova.rpc.common import RemoteError, LOG
|
|
N301
|
|
"""
|
|
pos = logical_line.find(',')
|
|
parts = logical_line.split()
|
|
if pos > -1 and (parts[0] == "import" or
|
|
parts[0] == "from" and parts[2] == "import") and \
|
|
not is_import_exception(parts[1]):
|
|
return pos, "NOVA N301: one import per line"
|
|
|
|
_missingImport = set([])
|
|
|
|
|
|
def nova_import_module_only(logical_line):
|
|
"""Check for import module only.
|
|
|
|
nova HACKING guide recommends importing only modules:
|
|
Do not import objects, only modules
|
|
N302 import only modules
|
|
N303 Invalid Import
|
|
N304 Relative Import
|
|
"""
|
|
def importModuleCheck(mod, parent=None, added=False):
|
|
"""
|
|
If can't find module on first try, recursively check for relative
|
|
imports
|
|
"""
|
|
current_path = os.path.dirname(pep8.current_file)
|
|
try:
|
|
with warnings.catch_warnings():
|
|
warnings.simplefilter('ignore', DeprecationWarning)
|
|
valid = True
|
|
if parent:
|
|
if is_import_exception(parent):
|
|
return
|
|
parent_mod = __import__(parent, globals(), locals(),
|
|
[mod], -1)
|
|
valid = inspect.ismodule(getattr(parent_mod, mod))
|
|
else:
|
|
__import__(mod, globals(), locals(), [], -1)
|
|
valid = inspect.ismodule(sys.modules[mod])
|
|
if not valid:
|
|
if added:
|
|
sys.path.pop()
|
|
added = False
|
|
return logical_line.find(mod), ("NOVA N304: No "
|
|
"relative imports. '%s' is a relative import"
|
|
% logical_line)
|
|
return logical_line.find(mod), ("NOVA N302: import only "
|
|
"modules. '%s' does not import a module"
|
|
% logical_line)
|
|
|
|
except (ImportError, NameError) as exc:
|
|
if not added:
|
|
added = True
|
|
sys.path.append(current_path)
|
|
return importModuleCheck(mod, parent, added)
|
|
else:
|
|
name = logical_line.split()[1]
|
|
if name not in _missingImport:
|
|
if VERBOSE_MISSING_IMPORT:
|
|
print >> sys.stderr, ("ERROR: import '%s' failed: %s" %
|
|
(name, exc))
|
|
_missingImport.add(name)
|
|
added = False
|
|
sys.path.pop()
|
|
return
|
|
|
|
except AttributeError:
|
|
# Invalid import
|
|
return logical_line.find(mod), ("NOVA N303: Invalid import, "
|
|
"AttributeError raised")
|
|
|
|
# convert "from x import y" to " import x.y"
|
|
# convert "from x import y as z" to " import x.y"
|
|
import_normalize(logical_line)
|
|
split_line = logical_line.split()
|
|
|
|
if (logical_line.startswith("import ") and "," not in logical_line and
|
|
(len(split_line) == 2 or
|
|
(len(split_line) == 4 and split_line[2] == "as"))):
|
|
mod = split_line[1]
|
|
return importModuleCheck(mod)
|
|
|
|
# TODO(jogo) handle "from x import *"
|
|
|
|
#TODO(jogo): import template: N305
|
|
|
|
|
|
def nova_import_alphabetical(physical_line, line_number, lines):
|
|
"""Check for imports in alphabetical order.
|
|
|
|
nova HACKING guide recommendation for imports:
|
|
imports in human alphabetical order
|
|
N306
|
|
"""
|
|
# handle import x
|
|
# use .lower since capitalization shouldn't dictate order
|
|
split_line = import_normalize(physical_line.strip()).lower().split()
|
|
split_previous = import_normalize(lines[line_number - 2]
|
|
).strip().lower().split()
|
|
# with or without "as y"
|
|
length = [2, 4]
|
|
if (len(split_line) in length and len(split_previous) in length and
|
|
split_line[0] == "import" and split_previous[0] == "import"):
|
|
if split_line[1] < split_previous[1]:
|
|
return (0, "NOVA N306: imports not in alphabetical order (%s, %s)"
|
|
% (split_previous[1], split_line[1]))
|
|
|
|
|
|
def nova_docstring_start_space(physical_line):
|
|
"""Check for docstring not start with space.
|
|
|
|
nova HACKING guide recommendation for docstring:
|
|
Docstring should not start with space
|
|
N401
|
|
"""
|
|
pos = max([physical_line.find(i) for i in DOCSTRING_TRIPLE]) # start
|
|
if (pos != -1 and len(physical_line) > pos + 1):
|
|
if (physical_line[pos + 3] == ' '):
|
|
return (pos, "NOVA N401: one line docstring should not start with"
|
|
" a space")
|
|
|
|
|
|
def nova_docstring_one_line(physical_line):
|
|
"""Check one line docstring end.
|
|
|
|
nova HACKING guide recommendation for one line docstring:
|
|
A one line docstring looks like this and ends in a period.
|
|
N402
|
|
"""
|
|
pos = max([physical_line.find(i) for i in DOCSTRING_TRIPLE]) # start
|
|
end = max([physical_line[-4:-1] == i for i in DOCSTRING_TRIPLE]) # end
|
|
if (pos != -1 and end and len(physical_line) > pos + 4):
|
|
if (physical_line[-5] != '.'):
|
|
return pos, "NOVA N402: one line docstring needs a period"
|
|
|
|
|
|
def nova_docstring_multiline_end(physical_line):
|
|
"""Check multi line docstring end.
|
|
|
|
nova HACKING guide recommendation for docstring:
|
|
Docstring should end on a new line
|
|
N403
|
|
"""
|
|
pos = max([physical_line.find(i) for i in DOCSTRING_TRIPLE]) # start
|
|
if (pos != -1 and len(physical_line) == pos):
|
|
print physical_line
|
|
if (physical_line[pos + 3] == ' '):
|
|
return (pos, "NOVA N403: multi line docstring end on new line")
|
|
|
|
|
|
FORMAT_RE = re.compile("%(?:"
|
|
"%|" # Ignore plain percents
|
|
"(\(\w+\))?" # mapping key
|
|
"([#0 +-]?" # flag
|
|
"(?:\d+|\*)?" # width
|
|
"(?:\.\d+)?" # precision
|
|
"[hlL]?" # length mod
|
|
"\w))") # type
|
|
|
|
|
|
class LocalizationError(Exception):
|
|
pass
|
|
|
|
|
|
def check_l18n():
|
|
"""Generator that checks token stream for localization errors.
|
|
|
|
Expects tokens to be ``send``ed one by one.
|
|
Raises LocalizationError if some error is found.
|
|
"""
|
|
while True:
|
|
try:
|
|
token_type, text, _, _, _ = yield
|
|
except GeneratorExit:
|
|
return
|
|
if token_type == tokenize.NAME and text == "_":
|
|
while True:
|
|
token_type, text, start, _, _ = yield
|
|
if token_type != tokenize.NL:
|
|
break
|
|
if token_type != tokenize.OP or text != "(":
|
|
continue # not a localization call
|
|
|
|
format_string = ''
|
|
while True:
|
|
token_type, text, start, _, _ = yield
|
|
if token_type == tokenize.STRING:
|
|
format_string += eval(text)
|
|
elif token_type == tokenize.NL:
|
|
pass
|
|
else:
|
|
break
|
|
|
|
if not format_string:
|
|
raise LocalizationError(start,
|
|
"NOVA N701: Empty localization string")
|
|
if token_type != tokenize.OP:
|
|
raise LocalizationError(start,
|
|
"NOVA N701: Invalid localization call")
|
|
if text != ")":
|
|
if text == "%":
|
|
raise LocalizationError(start,
|
|
"NOVA N702: Formatting operation should be outside"
|
|
" of localization method call")
|
|
elif text == "+":
|
|
raise LocalizationError(start,
|
|
"NOVA N702: Use bare string concatenation instead"
|
|
" of +")
|
|
else:
|
|
raise LocalizationError(start,
|
|
"NOVA N702: Argument to _ must be just a string")
|
|
|
|
format_specs = FORMAT_RE.findall(format_string)
|
|
positional_specs = [(key, spec) for key, spec in format_specs
|
|
if not key and spec]
|
|
# not spec means %%, key means %(smth)s
|
|
if len(positional_specs) > 1:
|
|
raise LocalizationError(start,
|
|
"NOVA N703: Multiple positional placeholders")
|
|
|
|
|
|
def nova_localization_strings(logical_line, tokens):
|
|
"""Check localization in line.
|
|
|
|
N701: bad localization call
|
|
N702: complex expression instead of string as argument to _()
|
|
N703: multiple positional placeholders
|
|
"""
|
|
|
|
gen = check_l18n()
|
|
next(gen)
|
|
try:
|
|
map(gen.send, tokens)
|
|
gen.close()
|
|
except LocalizationError as e:
|
|
return e.args
|
|
|
|
#TODO(jogo) Dict and list objects
|
|
|
|
current_file = ""
|
|
|
|
|
|
def readlines(filename):
|
|
"""Record the current file being tested."""
|
|
pep8.current_file = filename
|
|
return open(filename).readlines()
|
|
|
|
|
|
def add_nova():
|
|
"""Monkey patch in nova guidelines.
|
|
|
|
Look for functions that start with nova_ and have arguments
|
|
and add them to pep8 module
|
|
Assumes you know how to write pep8.py checks
|
|
"""
|
|
for name, function in globals().items():
|
|
if not inspect.isfunction(function):
|
|
continue
|
|
args = inspect.getargspec(function)[0]
|
|
if args and name.startswith("nova"):
|
|
exec("pep8.%s = %s" % (name, name))
|
|
|
|
if __name__ == "__main__":
|
|
#include nova path
|
|
sys.path.append(os.getcwd())
|
|
#NOVA error codes start with an N
|
|
pep8.ERRORCODE_REGEX = re.compile(r'[EWN]\d{3}')
|
|
add_nova()
|
|
pep8.current_file = current_file
|
|
pep8.readlines = readlines
|
|
try:
|
|
pep8._main()
|
|
finally:
|
|
if len(_missingImport) > 0:
|
|
print >> sys.stderr, ("%i imports missing in this test environment"
|
|
% len(_missingImport))
|