I order my Python imports by name length. Does this make any sense?

Question

First of all, yes I'm aware of PEP8 and the alphabetical method.

I do sort my imports first by the PEP8 recommendation:

1. standard library imports
2. related third party imports
3. local application/library specific imports

But then, within each category, I like to sort by the length of the module name. Yes, it makes it look pretty, but then I find it arguably makes it easier to find a module than the alphabetical way.

Also, the shorter names tend to be the more basic, common modules (like os and sys), so there's an additional organization layer in that way. You know the length of the name you're looking for, and the visual shape of the import list tells you very quickly where to look for names that long.

An example:

import os
import sys
import time
import logging
import argparse
import subprocess

import django
import requests
import webencodings

import mymodule
import localthing
import supercustomstuff

Right now I mostly write for myself, and I don't work with a team. But I intend to in the future. So what I'm wondering is, does anyone else do this, does anyone else see the sense in it, or will everyone else think I'm nuts?

Edit: Separate subgroups in example, as suggested by Mark Ransom, in order not to distract from the core issue about the ordering.

Answer 1

First, I would say to do what works for you on your own projects. It’s not going to hurt anything and it helps you, so it’s a win!

For projects with others I would discuss it with the team. Often a team will have a style guide that they expect developers to follow. Other times they allow every developer to do their own thing. I would advise following what others on your team do.

Personally, your system seems arbitrary to me and quite odd. Others probably have systems they use for equally arbitrary reasons. I don’t feel like yours would help me in general. I have to wonder how much it helps when the module names are all close in length? At least with alphabetic (which is also not my favorite) I know to use the first 2 letters when two files start with the same letter.

Also, I’m not sure about Python but in some languages, if developers aren’t careful, import order can change the behavior of a program. (That shouldn’t be the case and it shouldn’t be relied on, but the reality is that it happens.)

Answer 2

This is a classic XY question. You ask about one issue, but I will suggest you actually have a different issue!

You say sorting by length makes it easier to find a module. So apparently you spend a significant amount of effort scanning through the list of imports to see if a certain import is present. This indicates to me you have far to many imports in a single file. Each file have too many different responsibilities and dependencies. Apply some modularization and separation of concerns to your own code, so each file will have a manageable number of dependencies. You should be able to see all imports in a section at a glance so ordering shouldn't matter that much.

As for sorting by name length, this does not make a lot of sense. The length of a name does not carry any particular meaning. Module names are selected to be simple and descriptive, but the length does not necessarily correspond to how "basic" they are, even if it happens to be the case in the examples you chose.

If you want to sort by how basic the modules are (which is reasonable) then do that, and ignore the length of names.

Answer 3

Your ordering is weird. If you intend to work with other people, don't do it. Even if you're working on your own private project, don't allow yourself to get into bad habits.

I'd also put a blank line between the groups to separate them.

Answer 4

It's indeed you own style and I see no harm in keeping it.

But, I'd say its rooted in a deeper instinct of a "natural order", or in fact a combination of various such orders:

1. Common imports that are almost in every file always go in first. This is not surprising as the imports section slowly becomes a template in your head. A quick glance and you know what's "special" imported here.

2. More specific things usually tend to have longer names, namespaces and import paths. So, the "generic first and specialized later" rule actually ends up with the length based ordering.

3. The urge to group, categorize and separately mark everything is also pretty common. Closely related imports are usually grouped together.

4. And finally, we cannot ignore the "visual aesthetics" of text, many programmers regard highly. We are arranging the information for our own quick processing.

Answer 5

I'll just add my two cents as someone who's been sorting by line length for a couple of years now.

The way I see it, the two biggest advantages of this sort are

1. Reduced cognitive load - it's natural for us to sort by length, no need to recite the alphabet in your head every time you make a change. And unlike alphabetical sorting, when things are not sorted by length, they immediately stand out, which makes it better to maintain.

2. Visuals - it is subjective, but I think the second one looks much better:

• Alphabetical + grouping:

import time
import logging
import multiprocessing as mp
import os
import time

from package1 import module1
import package1.module2
from package1.module3 import cfg

from package2 import class1
from package2.subpackage1.module5 import functionC
from package2.subpackage3 import functionB

• Length + grouping

import os
import time
import logging
import argparse
import multiprocessing as mp

import package1.module2
from package1 import module1
from package1.module3 import cfg

from package2 import class1
from package2.subpackage3 import functionB
from package2.subpackage1.module5 import functionC

As for some other comments saying it's "weird" or that "it does not make a lot of sense".

Neither does alphabetical sorting. It's a mere consensus that is totally arbitrary. There's no deeper meaning or great advantage to either one. Alphabetical sorting makes it slightly easier to find packages when they have similar lengths, but aren't you using Ctrl+F or '?' commands to search longer import lists anyway? Personally, the two advantages above far outweigh a paltry improvement in lookup time.

Answer 6

First, I'll echo @user1118321 that this rule is fine if it works for you and that in a team, just come to an agreement to avoid personal and source merge conflicts.

How does your rule handle imports that say more than import abcd? Imports frequently look like these:

from typing import Dict, Iterator, List, Set
from google.cloud.storage import Blob, Client
import numpy as np
from . import mymodule

If the module-name-length rule discourages using these import features, or make you reorder the lines when adding more types to an import line, or defeat the visual order that was the point, then the rule is probably doing more harm than good.

Personally I'd rather let PyCharm "optimize imports", and that'd also avoid some source merge conflicts, but teammates are not careful about import order, I don't want my editor to accidentally reorder their imports, and there's no reason to disagree over it.

Answer 7

Looking at your first group of imports for example, I can’t see any real reason for any ordering. Therefore, I don’t really care how you order it. Your business. It would never occur to me why the imports are ordered the way they are, so I’d add a new import at the end of the list, and you could feel free to move it elsewhere.

I find it useful to know why some import is needed, as a comment, unless it is obvious.