vsego / deep-vars

Recursive `vars` for all objects

deep_vars

A package providing deep_vars function that serves a purpose similar to vars, but it works with any object and can recursively process attributes as well.

Content

1. Basic use
2. Configuration
   1. Handlers
   2. Further configuration specific to some handlers
   3. Extending handlers

Basic use

This is used in a way similar to vars. For example, the following code:

from dataclasses import dataclass
from typing import Optional

from deep_vars import deep_vars


@dataclass
class Foo:
    x: int
    y: Optional["Foo"] = None


foo = Foo(17, y=Foo(19))
print("vars:                     ", vars(foo))
print("deep_vars:                ", deep_vars(foo))
print("deep_vars with maxdepth=2:", deep_vars(foo, 2))

will print this:

vars:                      {'x': 17, 'y': Foo(x=19, y=None)}
deep_vars:                 {'x': 17, 'y': Foo(x=19, y=None)}
deep_vars with maxdepth=2: {'x': 17, 'y': {'x': 19, 'y': None}}

Note that, because it covers objects of all classes, deep_vars might not product a dictionary. For example, if we define foo in the above code like this:

foo = [Foo(17), "foo", 17]

the vars print would just raise an exception:

TypeError: vars() argument must have __dict__ attribute`

while the deep_var outputs will be:

deep_vars:                 [Foo(x=17, y=None), 'foo', 17]
deep_vars with maxdepth=2: [{'x': 17, 'y': None}, 'foo', 17]

Configuration

deep_vars is meant to be simple to use, like vars, so its only arguments are the object being represented and the maximum depth down to witch the conversion should go.

However, there are various possibilities for configuring it for the whole project. This is done through the attributes of DeepVars class.

Handlers

The main thing one can adjust is how various objects are handled. This is done by assigning "handler types". For example, this will cause all float and str attributes to remain hidden:

DeepVars.set_handlers(DeepVarsHandlerHide, float, str)

There are several DeepVarsHandler<something> classes that can be used:

• DeepVarsHandler: Default class, used as a parent class for all the others. It should not be used directly.
• DeepVarsHandlerShow: Show the object as it is (i.e., don't process it).
• DeepVarsHandlerHide: Skip the attribute from the output. Note that this does not affect the top level call of deep_vars.
• DeepVarsHandlerProcess: Process it in the previously described manner (similar to vars).
• DeepVarsHandlerLoop: Loop the object (usually a list or a tuple) and process each item individually.
• DeepVarsHandlerLoopDict: Loop the object as a dictionary, using its items method (can be changed in inherited classes) and process each (key, value) pair individually.
• DeepVarsHandlerCallable: Special handler for callables, converting them to signature strings.

One can also configure the default handler for those attributes that are not individually set. For example, to have them show without any processing:

DeepVars.default_handler = DeepVarsHandlerShow

Let us take another look at the above example:

DeepVars.set_handlers(DeepVarsHandlerHide, float, str)

The first argument is a handler class. The rest of them describe the objects that this class is used for. These are usually types, but they can also be strings. The following are recognised:

• "magic": A magic attribute. These are usually methods with names starting and ending in double underscores (for example, __init__), but deep_vars will consider any attribute, even non-callable ones, for the sake of output's consistency.

• "private": Any object that has a name starting with an underscore (for example, _private or __very_private).

• "callable": Any callable object (a function, a method, a lambda) except classes.

The handlers assignment is done in the same way as it is for type-based kinda of arguments:

# Hide all "callable" arguments (not including classes):
DeepVars.set_handlers(DeepVarsHandlerHide, "callable")
# Hide all `float` and all "callable" arguments (not including classes):
DeepVars.set_handlers(DeepVarsHandlerHide, float, "callable")

Further configuration specific to some handlers

Some of the handlers can be fine-tuned.

DeepVarsHandlerLoopDict normally uses object's items() method to go through its items. However, this can be replaced by changing DeepVarsHandlerLoopDict.items_method_name to something other than "items". Further, if this other method requires arguments, these can be set up as a tuple DeepVarsHandlerLoopDict.items_method_args and a dictionary DeepVarsHandlerLoopDict.items_method_kwargs.

All container handlers (DeepVarsHandlerLoopDict, DeepVarsHandlerLoop, and DeepVarsHandlerProcess) and also have the filtering of their items turned on or off. If turned off (class_name.allow_filtering = False), their items will show even if their handler is set to DeepVarsHandlerHide. This is the default for (DeepVarsHandlerLoopDict and DeepVarsHandlerLoop. If the filtering is turned on (class_name.allow_filtering = True), their items will be hidden if their handler decides so (i.e., if it is set to DeepVarsHandlerHide). This is the default for DeepVarsHandlerProcess.

Extending handlers

One can easily define their own special object types. Here is an example from the tests:

def special_detector(name: Optional[str], obj: Any) -> Optional[str]:
    return (
        "specially_detected"
        if getattr(obj, "specially_detected", False) else
        None
    )


DeepVars.special_type_detectors.append(special_detector)
DeepVars.set_handlers(DeepVarsHandlerSpecial, "specially_detected")

First, we need to define a callable (usually a function) that detects the special type of the argument. This callable takes attribute's name (either a string or None, the latter being used for the top level call of deep_vars) and the object itself, returning either a string used as the name for that kind of object or None if the object was not identified.

Second, we add this callable to the list DeepVars.special_type_detectors. The three predefined special types of arguments listed above are built-in, but special_type_detectors is checked before they are used, so its callables can override them.

Lastly, one needs to define how is this special is handled, which is done as shown above with "callable" arguments.

This allows one to extend the functionality of deep_vars. Arguably, this is not very useful in regular projects, but it can be used by other packages to offer deep_vars functionality extended to their own needs.