I am trying to auto document types with sphinx autodoc, napoleon and autodoc_typehints but I am having problems as it does not work with most of my types. I am using the deap package to do some genetic optimization algorithm, which makes that I have some very specific types I guess sphinx cannot handle.
My conf.py file looks like this:
import os
import sys
sys.path.insert(0, os.path.abspath('../python'))
extensions = [
'sphinx.ext.autodoc',
'sphinx.ext.viewcode',
'sphinx.ext.napoleon',
'sphinx_autodoc_typehints'
]
set_type_checking_flag = False
always_document_param_types = False
I have an Algo.rst file with:
.. automodule:: python.algo.algo
:members: crossover_worker,
test
and my python.algo.algo module looks like this (I've added a dummy test function to show it works whenever I have no special types specified):
# Type hinting imports
from config.config import Config
from typing import List, Set, Dict, NamedTuple, Union, Tuple
from types import ModuleType
from numpy import ndarray
from numpy import float64
from multiprocessing.pool import MapResult
from deap.tools.support import Logbook, ParetoFront
from deap.base import Toolbox
from pandas.core.frame import DataFrame
from deap import creator
...
def crossover_worker(sindices: List[creator.Individual, creator.Individual]) -> Tuple[creator.Individual, creator.Individual]:
"""
Uniform crossover using fixed threshold
Args:
sindices: list of two individuals on which we want to perform crossover
Returns:
tuple of the two individuals with crossover applied
"""
ind1, ind2 = sindices
size = len(ind1)
for i in range(size):
if random.random() < 0.4:
ind1[i], ind2[i] = ind2[i], ind1[i]
return ind1, ind2
def test(a: DataFrame, b: List[int]) -> float:
"""
test funcition
Args:
a: something
b: something
Returns:
something
"""
return b
When settings in conf.py are like above I have no error, types for my test function are correct, but types for my crossover_worker function are missing:
However, when I set the set_type_checking_flag= True to force using all types, I have a circular import error:
reading sources... [100%] index
WARNING: autodoc: failed to import module 'algo' from module 'python.algo'; the following exception was raised:
cannot import name 'ArrayLike' from partially initialized module 'pandas._typing' (most likely due to a circular import) (/usr/local/lib/python3.8/site-packages/pandas/_typing.py)
looking for now-outdated files... none found
And I never import ArrayLike so I don't get it from where it comes or how to solve it?
Or how to force to import also the creator.Individual types that appear everywhere in my code?
My sphinx versions:
sphinx==3.0.1
sphinx-autodoc-typehints==1.10.3
After some searching there were some flaws with my approach:
Firstly a "list is a homogeneous structure containing values of one type. As such, List only takes a single type, and every element of that list has to have that type." (source). Consequently, I cannot do something like List[creator.Individual, creator.Individual], but should transform it to List[creator.Individual] or if you have multiple types in the list, you should use an union operator, such as List[Union[int,float]]
Secondly, the type creator.Individual is not recognized by sphinx as a valid type. Instead I should define it using TypeVar as such:
from typing import TypeVar, List
CreatorIndividual = TypeVar("CreatorIndividual", bound=List[int])
So by transforming my crossover_worker function to this, it all worked:
def crossover_worker(sindices: List[CreatorIndividual]) -> Tuple[CreatorIndividual, CreatorIndividual]:
Note: "By contrast, a tuple is an example of a product type, a type consisting of a fixed set of types, and whose values are a collection of values, one from each type in the product type. Tuple[int,int,int], Tuple[str,int] and Tuple[int,str] are all distinct types, distinguished both by the number of types in the product and the order in which they appear."(source)
Related
I am having trouble setting up a simple dataclass using a new defaultdict(dict).
If I tell the factory to use 'dict' as below , instantiation fails with typerror collection.defaultdict object is not callable
from collections import defaultdict
from dataclasses import dataclass, field
#dataclass
class ResultSet:
changed: bool = False
mqttdata: defaultdict(dict) = field(default_factory=defaultdict(dict)) # does not work!
It does sorta work using field(default_factory=defaultdict) but then my code will fail later when it encounters missing keys - presumably because defaultdict was not set up for dict.
How do I properly set up a new defaultdict(dict) in a dataclass?
You have a few problems with the code and how you are using dataclasses currently:
Type generics in annotations need to be specified through square brackets [] generally, so default[dict] instead of defaultdict(dict) for example.
The default_factory argument to dataclasses.field() needs to be a no-arg callable which returns a new object with default values set. For example, assuming you have a nested dataclass Inner which specifies defaults for all fields, you could use default_factory=Inner to create a new Inner object each time the main dataclass is instantiated.
Note that the default_factory argument is mainly useful for mutable types
such as set, list, and dict, so that the same object isn't
shared (and potentially mutated) between dataclass instances.
Putting it all together, here is the working code which sets a default value for a field of type defaultdict[dict]:
from collections import defaultdict
from dataclasses import dataclass, field
#dataclass
class ResultSet:
changed: bool = False
mqttdata: defaultdict[dict] = field(default_factory=lambda: defaultdict(dict)) # works!
print(ResultSet())
In Python versions earlier than 3.9, which is when PEP 585 was introduced, you'll need to add the following import at the top so that any type annotations are lazy-evaluated:
from __future__ import annotations
Using the AST in python 3, how do you build a directory or list of code dependencies of a given function?
Consider the following code, where my_clever_function has the desired behaviour:
////// myfile2.py
import numpy as np
a = 1
a += 1
def my_other_function():
def f():
return a
return np.random.randint() + f()
////// myfile1.py
import numpy as np
from . myfile2 import my_other_function
def external(a, b):
return np.sqrt(a * b) + my_other_function
class A:
def afunc(self, a, b):
v = external(a, b)
return v
>>> my_clever_function(A.afunc)
[myfile1.A.afunc, myfile1.external, myfile2.my_other_function, myfile2.a]
with the following structure:
project/
myfile1.py
myfile2.py
I want to retrieve the dependencies of the method afunc as a list.
I'm assuming that there is no funny business about functions altering global variables.
external is a dependency because it is not defined inside A.afunc
np.sqrt is not a "dependency" (in this sense anyway) because it is not defined in my project
likewise for np.random.randint
my_other_function is a dependency because it is not defined inside A.afunc
f is not a dependency because it is inside my_other_function
f needs the global variable a.
My motivation is to see if there have been any code changes between two project versions (in git perhaps).
We could find the dependencies of function like above and store the source.
In the future, we find the dependencies again and see if the source code is different.
We only compare the bits that are required (barring any funny global variables messing inside functions).
It is possible to walk the AST with python's builtin module ast.
So my_clever_function could look like this:
import ast
import dill
class Analyzer(ast.NodeVisitor):
def __init__(self):
self.stats = {...}
...
def report(self):
pprint(self.stats)
def my_clever_function(f):
source = dill.source.getsource(f)
tree = ast.parse(source)
analyser = Analyser()
analyser.visit(tree)
But how do you walk from a given function outwards to its dependencies?
I can see how you can just list symbols (https://www.mattlayman.com/blog/2018/decipher-python-ast/) but how do only list those which depend on the start node?
I'm trying to find out for myself how I could work around the problem I recently asked here and I come across a potential solution, but I honestly don't understand why it works, and why the other doesn't.
For context, model requires variables a and b to be defined before being successfully loaded and defined in a module. Otherwise, it throws an error: NameError: name 'a' is not defined.
Starting off with model.py:
import pickle
from tensorflow import keras
# loads and returns the variables needed by model
def load_model_vars():
return pickle.load(open('./file.pkl', 'rb'))
# loads and returns the model
def load_model():
return keras.models.load_model('./model.h5')
Now to minimally reproduce and identify the problem I created a new module, foo.py:
from model import load_model_vars, load_model
# goal here is to supposedly expose only the model to other modules
a, b = load_model_vars()
globals()['model'] = load_model()
I then created another module to import foo.py into, let's name it bar.py:
import foo
# just checks if the model is defined
foo.model.summary()
Which for some reason throws the formerly mentioned NameError. Why? The variables are defined, it was executed in order(load variables first, then model), and even if I change a, b to globals()['a'], globals()['b'], import foo to from foo import * or from foo import a, b or even combinations of any of these, it always arrives into this error.
But when I introduce another module, say, baz.py, that contains these two lines:
from model import load_model_vars
a, b = load_model_vars()
Then import it to bar.py:
from baz import a, b
import foo
# just checks if the model is defined
foo.model.summary()
With foo.py unchanged, or with a, b = load_model_vars() commented out:
from model import load_model_vars, load_model
# goal here is to supposedly expose only the model to other modules
# a, b = load_model_vars()
globals()['model'] = load_model()
It successfully loads the freaking model! Why? What's this sorcery underneath the import function? What actually happens under the hood?
I am having some trouble assigning the variables in a Python 3.6 class to a particular type--a Pathlib path. Following an example from link, I tried to create a TypeVar, but mypy is still throwing errors. I want to make sure that the class variables initialized in the __init__.py only receive a particular type at compile time. So this is just a check to make sure I don't inadvertently set a string or something else to these class variables.
Can anyone suggest the correct way to do this?
Here is some simple code.
import pathlib
from typing import Union, Dict, TypeVar, Type
Pathtype = TypeVar('Pathtype', bound=pathlib.Path)
class Request:
def __init__(self, argsdict):
self._dir_file1: Type[Pathtype] = argsdict['dir_file1']
self._dir_file2: Type[Pathtype] = argsdict['dir_file2']
The error that I am getting is:
Request.py:13: error: Invalid type "Request.Pathtype"
Request.py:14: error: Invalid type "Request.Pathtype"
Neither Type, TypeVar nor NewType are correct to use here. What you simply want to do is use Path itself:
from pathlib import Path
class Request:
def __init__(self, argsdict):
self._dir_file1: Path = argsdict['dir_file1']
self._dir_file2: Path = argsdict['dir_file2']
If you annotate your argsdict as being of type Dict[str, Path], you can skip having to annotate your fields entirely: mypy will infer the correct type:
from typing import Dict
from pathlib import Path
class Request:
def __init__(self, argsdict: Dict[str, Path]):
self._dir_file1 = argsdict['dir_file1']
self._dir_file2 = argsdict['dir_file2']
Here's a brief explanation of what the various type constructs you were attempting to use/was suggested to you actually do:
TypeVar is used when you are trying to create a generic data structure or function. For example, take List[int], which represents a list containing ints. List[...] is an example of a generic data structure: it can be parameterized by any arbitrary type.
You use TypeVar as a way of adding "parameterizable holes" if you decide you want to create your own generic data structure.
It's also possible to use TypeVars when writing generic functions. For example, suppose you want to declare that you have some function that can accept a value of any type -- but that function is guaranteed to return a value of the exact same type. You can express ideas like these using TypeVars.
The Type[...] annotation is used to indicate that some expression must be the type of a type. For example, to declare that some variable must hold an int, we would write my_var: int = 4. But what if we want to write something like my_var = int? What sort of type hint could we give that variable? In this case, we could do my_var: Type[int] = int.
NewType basically lets you "pretend" that you're taking some type and making a subclass of it -- but without requiring you to actually subclass anything at runtime. If you're careful, you can take advantage of this feature to help catch bugs where you mix different "kinds" of strings or ints or whatever -- e.g. passing in a string representing HTML into a function expecting a string representing SQL.
Replace TypeVar with NewType and remove the Type[] modifier.
I'm writing a library named Foo for an example.
The __init__.py file:
from .foo_exceptions import *
from .foo_loop import FooLoop()
main_loop = FooLoop()
from .foo_functions import *
__all__ = ['main_loop'] + foo_exceptions.__all__ + foo_functions.__all__
When installed, it can be used like this:
# example A
from Foo import foo_create, main_loop
foo_obj = foo_create()
main_loop().register(foo_obj)
or like this:
# example B
import Foo
foo_obj = Foo.foo_create()
Foo.main_loop().register(foo_obj)
I clearly prefer the example B approach. No name conflicts and the source of each external object is explicitely stated.
So much for introduction, now my question. Inside this library I need to import something from a different file. Again, I have several ways to do it. And the question is which style to prefer - C, D or E? Read below.
# example C
from . import foo_exceptions
raise foo_exceptions.FooError("fail")
or
# example D
from .foo_exceptions import FooError
raise FooError("fail")
or
# example E
from . import FooError
raise FooError("fail")
Approach C has the disadvantage, that importing a whole module instead of importing just a few required objects increases the chance of a cyclical import problem. Also consider this line:
from . import foo_exceptions, main_loop
It looks like an import of 2 symbols from one source, but it isn't. The former (foo_exceptions) is a module (.py file) in the current directory and the latter is an object defined in __init__.py.
That's why I'm not using style C and the question in its final form is: D or E (and why)?
(Thank you for reading this long question. All code fragments are examples only and may contain typos)
After the answer from alexanderlukanin:
EDIT1: corrected errors in init.py
NOTE1: foo_ prefixes are only to emphasize the relationship between objects
EDIT2: When importing an object which is not part of the library interface, style E is not usable. I think we have a winner: It's the from .module import symbol form.
Don't use old-style relative imports:
# Import from foo/foo_loop.py
# This DOES NOT WORK in Python 3
# and MAY NOT WORK AS EXPECTED in Python 2
from foo_loop import FooLoop
# This is reliable and unambiguous
from .foo_loop import FooLoop
Don't use asterisk import unless you really have to.
# Namespace pollution! Name clashes!
from .submodule import *
Don't use prefixes - you've got namespaces exactly for that purpose.
# Unpythonic
from foo import foo_something_create
foo_something_create()
# Pythonic
import foo.something
foo.something.create()
Your package's API must be well-defined. Your implementation must not be too tangled. The rest is a matter of taste.
# [C] This is good.
# Import order: __init__.py, exceptions.py
from . import exceptions
raise exceptions.FooError
# [D] This is also fine.
# Import order is the same as above,
# only name binding inside the current module is different.
from .exceptions import FooError
raise FooError
# [E] This is not as good because it adds one unnecessary level of indirection
# submodule.py -> __init__.py -> exceptions.py
from . import FooError
raise FooError
See also: Circular (or cyclic) imports in Python