Friday, October 21, 2022
HomeData ScienceImporting Objects from a Python Module Can Be Dangerous | by Marcin...

Importing Objects from a Python Module Can Be Dangerous | by Marcin Kozak | Oct, 2022


Have you learnt when you work with unique objects after importing them? Possibly with their copies? Possibly some are unique and others are copies?

Which horse is unique and which is a duplicate? It’s so troublesome to say! Supply: Picture by creator.

Whether or not it’s best to import a module (import module) or import objects from a module (from module import obj1, obj2) shouldn’t be a brand new situation. On this article, I’m not going to reinvent the wheel. We all know that importing a module has extra benefits, however in some conditions direct importing objects from a module works completely effective; it may well make an app somewhat faster, and the code may be shorter. If you wish to learn extra about this subject, Mark Lutz’s ebook (Lutz 2013) gives good studying.

There’s one truth about importing that’s simple to neglect. One in all these two strategies doesn’t import objects from one other module, however creates their copies as a substitute. Whereas often this doesn’t make a lot of a distinction, typically it may well, and the distinction may be vital.

On this article, I examine copying objects from a module as a substitute of importing them. I’ll present one benefit of importing a module over importing objects from a module, a bonus that typically can prevent a lot hassle.

Let’s say we have now an utility that consists of three modules:

  • helpers, which incorporates variables and capabilities for use in the principle utility;
  • motion, which incorporates the principle perform of the applying;
  • __main__, which is liable for operating the applying.

That is the contents of helpers:

I made it so simple as doable. Now motion:

So, run_app() is the principle perform of the applying. Will probably be known as within the __main__ module, which is liable for operating the applying. In fact, in our instance, __main__ is overly simplified, too:

This behaves as anticipated. Once we import the helpers module, we are able to use its foo() perform and helpers_value world variable. Be aware that helpers_value shouldn’t be in higher case as a result of it’s not a continuing; it’s only a world variable. Right here, “world” means world for the module; however that is anticipated: that is how scopes work in Python (Lutz 2013, Ramalho 2022). Then, we alter helpers.helpers_value and helpers.foo, and this modification is mirrored within the unique location, that’s, in helpers. We are able to see this by calling motion.run_app(), a perform that calls helpers.helpers_value and helpers.foo.

Now, allow us to change the way in which we import the helpers objects in __main__:

That’s fairly a distinction! What occurred?

We modified helpers_value and foo, however this didn’t have an effect on the unique objects, positioned in helpers. It is because import from creates a duplicate of the imported object. That is essential, so let me emphasize:

import from creates a duplicate of the imported object.

Thus, whereas our modifications to helpers_value and foo() affected these copies, they didn’t have an effect on the unique values from the helpers module. The motion.run_app() perform doesn’t use these copies; it makes use of the unique objects from helpers, which is why we noticed totally different habits within the two instances.

This implies loads, and modifications loads. Happily, in most conditions none of this issues, as a result of we’d hardly ever change objects from one module in one other module. Generally, nevertheless, we do, after which we should be cautious.

To make issues much more difficult, we have to take into account two conditions: importing immutable objects and importing mutable objects from a module.

Immutable objects

Whenever you import an immutable object from a module, as in from helpers import helpers_value, a duplicate of the article is created. So, your world namespace will now comprise two objects known as helpers_value: helpers.helpers_value and __main__.helpers_value. Did you count on this to occur?

In the identical approach, the foo() perform was not modified within the unique place: There are two foo() capabilities after import: helpers.foo() and __main__.foo().

Therefore, after we import a module (import helpers), we use the unique (immutable) objects straight from the module; after they change, we use modified objects.

Nonetheless, after we import immutable objects straight from a module (from helpers import foo, helpers_value), Python makes their copies, and we use these copies — not the unique objects. Subsequently, when the unique objects change within the module, this won’t have an effect on the objects we’re utilizing — as a result of we’re utilizing their copies.

Mutable objects

Mutable objects can’t be copied, and any modifications to their unique objects or so-called copies are made to the unique objects. It is because Python doesn’t create a duplicate of a mutable object; as a substitute, a brand new title is created which references the unique object.

This function of mutable objects impacts how importing works.

Let’s simplify our utility. It should now comprise just one object, a dictionary, which is a mutable Python object. So, helpers are as follows:

The motion module incorporates the run_app() perform, which now merely returns the helpers_value dictionary:

And that is the __main__ module:

Do you see what occurred? This time, though we imported helpers_value from helpers, we obtained a special outcome: the unique object was modified in-place.

This habits is typical for mutable objects. Python doesn’t create their copies; as a substitute, it creates totally different names that time to the identical object. So, altering the worth of the article assigned to any of those names could have the identical impact: the unique object will probably be affected.

Combining the 2 kinds of import will change nothing. Immutable objects will behave like immutable objects, and mutable objects will behave like mutable objects. Let’s see the way it works. You may take into account it an train.

This time, we are going to work with solely two modules, helpers and __main__. The previous will outline two objects, one immutable and one mutable:

In __main__, we are going to import them in each methods after which change the objects. Earlier than persevering with, attempt to guess the output of the beneath snippet. I’ve added citation marks in locations the place it’s best to guess the output, after every print().

And that is the output:

I hope you bought this proper. Understanding this side of importing objects and modules will assist you keep away from unusual errors that may occur when your utility incorporates a number of modules with some dependency hierarchy (e.g., module1 imports module2 that imports module3).

As I discussed above, understanding these intricacies of Python imports ought to assist you keep away from errors in functions consisting of a number of modules. Bear in mind:

  • import module allows you to use the objects from this module as module.obj. This implies you employ the unique object, the one positioned in module. Therefore, any change to this very object, accomplished wherever within the code as module.obj = …, will probably be mirrored in your calls to module.obj.
  • Within the case of immutable objects, from module import obj creates a duplicate of the article. This implies this copy is positioned inside the module’s scope (within the module wherein you imported obj). Altering the worth of this copy (i.e., obj) won’t have an effect on the unique object (module.obj).
  • Within the case of mutable objects, from module import obj creates solely a brand new title of the article. Subsequently, altering its worth (i.e., obj = …) will have an effect on the unique object (module.obj).
  • Importing the module (import module) as a substitute of importing objects from it (from module import obj) is thus safer. It is because you all the time work this the objects positioned straight within the module, and their copies are by no means created. Thus, it’s sufficient to hint what is occurring with the unique objects (the one positioned in module), and no matter varied different modules do to this object, they are going to be doing this to the unique object. When it’s essential create a duplicate, you are able to do it explicitly, the place you want it.

The final level is probably probably the most essential. Whenever you do want a duplicate, create it, however do it explicitly the place you want this copy. Creating a duplicate by a selected approach of importing is certainly implicit, oblique and unclear. Freshmen and even some intermediate Pythonistas can neglect, or not know in any respect, that from module import obj creates a duplicate or a brand new title of module.obj, and that the __main__ module’s scope now incorporates two objects:

  • within the case of immutable objects, module.obj and obj objects;
  • within the case of mutable objects, it’s really one object with two names: module.obj and obj.

In fact, all of this issues once you change objects from one module in one other module. You should be cautious then, and that is when all these items we’ve been discussing can occur.

I believe it’s much better to keep away from the confusion of making copies of objects by importing. Because of this, at the very least on this context, it’s safer to import a module than to import objects from it.

  • Lutz M. (2013). Studying Python. fifth version. O’Reilly Media.
  • Ramalho, L. (2022). Fluent Python. Clear, Concise, and Efficient Programming. 2nd. version. O’Reilly.
RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

- Advertisment -
Google search engine

Most Popular

Recent Comments