tftio / pystrict3 Goto Github PK

pystrict3 is a fast plausibility code analyser for Python3.

It checks code for obvious mistakes, such as

• calling declared and imported functions with a different number of arguments than they are defined with.
• accessing attributes and methods that are never set.
• documenting the wrong number of arguments (numpy or google style docstrings).
• using variables that were never set.

All this without running your python code!

This tool complements other static analysers such as pyflakes, and can be used alongside linters and code format checkers (such as pylint and flake8).

pystrict3 assumes that no monkey patching of builtin behaviour or magic attributes (__dict__, __local__) alter classes and variables behind the scenes. Python 3.5 and above is required.

pystrict3 checks that functions are called with the right number of arguments. This catches bugs before execution, for example when call signatures change to include an additional argument:

def foo(a, b):
    return a*b
foo(1, 2)        ## OK
foo(123)         ## error: wrong number of arguments

def bar(a, b=1):
    return a*b
bar(1)           ## OK
bar(1, 2)        ## OK
bar(1, 2, 3)     ## error: wrong number of arguments

# builtin module signatures are verified too:
import os, numpy
os.mkdir("foo", "bar") ## error: wrong number of arguments (if run with --load-builtin-modules)
numpy.exp()            ## error: wrong number of arguments (if run with --load-any-modules)

pystrict3 checks that classes are instanciated with the right number of arguments, methods are called with the right number of arguments, and only attributes are accessed which have been assigned somewhere. This catches bugs before execution, for example when call signatures change to include an additional argument:

class Foo():
    def __init__(self, a):
        self.a = a
    def foo(self):
        self.b = 3
        print(self.a) ## OK
    def bar(self, c):
        print(self.b) ## OK
        return self.c ## error, because never assigned

Foo(1, 2)  ## error: wrong number of arguments for __init__
foo = Foo(1)  ## OK

# real example:
class Ellipse:
    def __init__(self,center,radius_x,radius_y,fill_color,line_color,line_width):
        self.center = center
        self.radiusx = radius_x
        self.radiusy = radius_y
        self.fill_color = fill_color
        self.line_color = line_color
        self.line_width = line_width
    def strarray(self):
        return ["  <ellipse cx=\"%d\" cy=\"%d\" rx=\"%d\" ry=\"%d\"\n" %\
            (self.center[0],self.center[1],self.radius_x,self.radius_y),
            "    style=\"fill:%s;stroke:%s;stroke-width:%d\"/>\n" % (colorstr(self.fill_color),colorstr(self.line_color),self.line_width)]
        # error above: self.radius_x vs self.radiusx

pystrict3 also checks docstrings for documented arguments and returns (numpydoc, rst and google-style is supported). It does not give an error if no docstrings are present. However, if only part of the arguments are documented, it gives an error pointing out the missing arguments to document.

For example:

def compute(num1, num2):
    """
    Combined two integer numbers.

    Parameters
    ----------
    num1 : int
        First number to add.

    Returns
    -------
    sum: int
        first number plus second number
    int
        first number minus second number
    """:
        return num1 + num2, num1 - num2, num1

This would raise two warnings:

1. parameter num2 is not documented
2. a triple is returned, but a tuple is documented.

pystrict3 (--allow-redefining disables this behaviour) can enforce that variables are only assigned once. This avoids changing the meaning of variables, and leads to cleaner, more idiomatic code with fewer side-effects.

It also prevents overwriting python builtins. Some examples:

parse = parse(foo)    ## bad
node = get_node()
node.foo()            ## ok, modification
node += 3             ## ok, modification

def format(...):      ## bad, format is a python keyword

import requests, html

html = requests.get(url)  ## bad: overwrites imported package name

Contributions are welcome.

pystrict3 may not catch all corner cases. It tries hard to avoid unintentional false positives, and has a very high code coverage with integration tests (see runtests.sh and tests/ directory).

Tested on activestate recipes, approximately half of all valid python3 programs are pystrict3 compliant, indicating that its guidelines are already adhered to.

$ pip3 install pystrict3

$ pystrict3.py --help

usage: pystrict3.py [-h] [--import-builtin] [--import-any] [--allow-redefining] [--verbose] filenames [filenames ...]

pystrict3: a Python code checker. Checks number of arguments in function, class init and method calls. Optionally also checks calls to imported modules. Checks that class attributes accessed are assigned somewhere. Checks that builtin names are
not overwritten. Checks that variables are only assigned once.

positional arguments:
  filenames           python files to parse

options:
  -h, --help          show this help message and exit
  --import-builtin    Also load builtin python modules to check function signatures. (default: False)
  --import-any        Also load any modules specified in import statements to check function signatures. Warning: can execute arbitrary module code. (default: False)
  --allow-redefining  Allow redefining variables. (default: False)
  --verbose, -v       More verbose logging output. (default: False)

Run with:

$ python3 pystrict3.py <filenames>
$ python3 pystrict3.py --import-builtin <filenames>
$ python3 pystrict3.py --import-any <filenames>

Running with multiple filenames has the benefit that all function signatures are first recorded and verified across all files.

Running with --import-builtin checks function calls to builtin modules.

Running with --import-any checks function calls to any modules, but this requires pystrict3 to import them, potentially running arbitrary module code.

Example stderr outputs:

tests/expect-fail/recipe-412717.py:32: ERROR: Variable reuse: "Test"
tests/expect-fail/recipe-425043.py:13: ERROR: Function "pow" (3..3 arguments) called with 2 arguments
tests/expect-fail/recipe-578135.py:184: ERROR: Function "encode" (3..3 arguments) called with 2 arguments
Summary:
  - checked 287 function calls.
  - checked definition of 469 new and access of 393 variables.
  - checked 4 docstrings.
pystrict3: OK

Return code is non-zero if a error was detected, or 0 otherwise.

For verbose output, pipe stdout to /dev/null.

gendocstr.py pre-generates numpy-style docstrings.

Say you have a file with this code:

def indicator(r, threshold=42):
    if r > threshold:
        return False
    else:
        return True

Rewritten by gendocstr.py, the new code is:

def indicator(r, threshold=42):
    """<summary sentence of function in imperative>.


    Parameters
    -----------
    r: <TYPE>
        <MEANING OF r>
    threshold: int
        <MEANING OF threshold>

    Returns
    ----------
    indicator: bool
        <MEANING OF indicator>
    """
    if r > threshold:
        return False
    else:
        return True

The following command creates a file myfile-new.py with suggested docstrings:

$ python3 gendocstr.py --verbose myfile.py

If you want to overwrite the source code file directly:

$ gendocstr.py --in-place myfile.py

Features:

• gendocstr.py can guess the parameter type from keywords and type annotations, if provided.
• can guess the return type if it is a input parameter or if it is True/False.
• keeps all the existing code formatting as is (thanks to RedBaron).
• can be used together with isort and black to force normalised python code.

BSD 2-clause.

It's OK to have some pystrict3 warnings and errors. Take them as guidance towards cleaner code.

How to write code that does not shadow or override variables:

• Use del to actively remove unused variables:

  answer = input("Do you want to play? (yes/no)")
  if answer == "no":
      sys.exit()
  del answer
  answer = int(input("first value"))
  print(answer * 10):
  

• Name parts of computation explicitly:

  # bad:
  magicnumber = sys.argv[1]
  magicnumber = int(magicnumber)
  # better:
  magicnumberstr = sys.argv[1]
  magicnumber = int(magicnumberstr)
  
  
  filename = 'foo.pdf'
  if condition:
     filename = 'foo.png'  # bad
  
  # better:
  if condition:
     filename = 'foo.png'
  else:
     filename = 'foo.pdf'
  
  # bad:
  path = os.path.basename(sys.argv[1])
  path = path + filename   # bad: variable changes meaning
  path = path + extension
  
  # better:
  components = []
  components.append(os.path.basename(sys.argv[1]))
  components.append(filename)
  components.append(extension)
  path = ''.join(components)
  

• Refactor into functions:

  # original: "changes" is being reused.
  USE_JYTHON = False
  try:
      # ... code detecting something, which throws an exception
      USE_JYTHON = True  ## re-assigning: not allowed
      # could use instead:
      # USE_JYTHON |= True
  except:
      pass
  # or define a function
  USE_JYTHON = check_jython()
  
  # original: a sorting construct
  changes = True
  while changes:
      changes = False
      for a in ...:
          if ...:
              changes = True
              break
      if not changes:
          break
  
  # new: function returns when no further changes are needed
  def mysort(objs):
      while True:
          changes = False
          for a in ...:
              if ...:
                  changes = True
                  break
          if not changes:
              return objs
  

• Instead of assigning to __doc__, move the docstring to the start of the file.