There are a few I’m aware of:
- Rust-style
use(multiple imports per statement, renaming items usingas) - Java-style
import(single import per statement, no renaming (afaik)) - Zig-style
let xlib =(imports as assignments) - C-style
(no importing, just pastes the code of the named file)
Elm:
-- Import under the name "List" import List -- Import aliased under the name "JD" import Json.Decode as JD -- Import Html, and put the Html type into the unqualified namespace import Html exposing (Html) -- Import Http, and put everything from it in the unqualified namespace import Http exposing (..)In order or preference, out of the ones I’m familiar with:
- Java. Very simple. Diff friendly. Import names match the filesystem.
- Rust. Also good - you can do it diff friendly though nobody does. Slightly confusing distinction between module hierarchy and filesystem.
- Python. Pretty bad. Conflates third party packages with local files, almost nobody actually understands relative imports, even more confusing distinction between module hierarchy and filesystem. Module imports can have side effects. Unnecessarily different syntax between
import...andfrom .... import.... - C/C++. Obviously this is the worst.
Well, I have been using python for a long time and was ready to disagree with you but damn, it really do be like that.
I’m not educated enough to have an opinion that goes beyond what I like to type. I find C style to be the nicest.
Zig and Rust are good. Java is painful.
But the best for me is Clojure:
(require [my.module :refer [func1 func2]] [Some.lib :as lib])I don’t know, but my least favourite has to be Python with its
fromsyntax.e.g.,
# fine import os.path # what?? from os import pathI don’t mind the syntax as such but the whole module, file, relative hidden import rules drives me nuts.
iirc there is also this patter of putting the following inside init.py:
from .myfunction import myfunction
Kotlins import system feels a lot like Java’s, but improved, since it allows aliasing of conflicting names.
Like to import via smuggli- OH we’re talking about coding? Uhh Java does it pretty well and we’re learning C# rn so gonna have to learn how to import that soon
I think C# handles this pretty well. You include namespaces with whatever granularity you need, down to a single class (or interface or whatever) and can alias them.
using MyProject.Domian.Entities; //the whole module using Routes = MyProject.Api.Contract.Routes.Http; //an aliased single classI bet Java had this too, C# us still mostly Java: Microsoft Edition.
You also don’t have to use imports at all in C# code - you can just reference everything absolutely using the full namespace and class name.
That’s a big difference compared to the other languages. C# automatically loads the correct assembly (.NET library) on the first call to something contained within it, whereas languages like Python require an
importstatement because that’s what loads the library. I don’t think autoloading is very common - C#, PHP, Ruby and Perl use it, but I can’t think of other languages that do.Rust & Java also do this
Good catch on Java! I’ve never used rust so I wasn’t aware that it also allows this. Good to know!
Java doesnt have aliases. You can do glob imports of a particular package (IE all the classes) and you can statically glob import all the static functions in a class but uh yeah that’s it
Wow, now I have something to hold over all my Java dev friends. Thanks!
Java doesn’t have aliases, but Kotlin does! (Kotlin is just better Java)
I enjoyed playing with Kotlin on Android. Nice ergonomics.
you can also tell them java cant handle.pointers. i have worked with a lot of backend c code using c#.
not sure about the following but:
can you control how a struct is handled in memory in java? (had tto do this with a c# front end calling a credit card verifyling service/library written in c++)
do they have value types?
Java doesnt have value types or structs. You have primatives (booleans, integer types and float/double) and then everything else (even the uppercase primative types like
Boolean) are reference types, IE on the heap. For stuff that plays with pointers you generally just call into native code, get back a long (the pointer to whatever you just allocated) and you call into native code to do stuff with the pointer. Native code can make a proper java object but it’s involved and has all the same cons as allocating one in Java land.Supposedly
records (a pseudo struct) might get stronger optimization options (IE put on the stack) but fundementaly reference types are just pointers and if you get an object via constructor or parameter the language must support you putting it in global stateYou cqn’t really control the layout as that would prevent the JVM from reshaping your classes (which is a fundamental requirement for things like Dalvik (android jvm))
Conversely, I’ve written C++ code to into DLLs interop with .NET. It makes a lot of sense given what .NET was developed for.
It’s actually pretty similar to rust, only that rust can import multiple modules (namespaces) in one line, which is just syntax sugar, and the syntax is slightly different.
Your example would be
use my_projext::domain::entities; use my_project::api::contracts::routes::http as routes;Rust adds the curly brace syntax so you can import a specific subset from a module, but I can’t decide if I like that. In C# you just keep adding extra lines which take up more vertical space, but if you sort your using a alphabetically (which your IDE can do), I think you get a very readable list. In sum, Rust adds a feature that I don’t really actually like.
c# is delphi with some realized mfc lofty goals.
first, microsoft liked delphi and hired the guy who created delphi to create c#
second if you read an mfc book from the 90s that talks about history and goals, 6ou can see they wanted the .net framework
Yeah C# is my favorite, followed by how python does it. Both are very good.
I have so little Python experience I didn’t even mention it, but I remember enjoying that too.
Its s nothing burger so its very good.
What I don’t like is doing it 100 times to pull in a main.py but there’s ways of getting around it. Its not the languages fault but its kinda what everyone does.
I like Rust’s style. It’s simple and it does everything you need. As much as I love Elixir, which I write at my day job, it has alias, import, use, and require, which each import things different, and it can get a bit confusing.
I’m most used to c style and as a result I like it because it’s simple and uncomplicated. I recently started learning rust and quite like their way of doing it, although there’s also mod and I haven’t quite figured out when to use which. I just do what rust analyser tells me :)
useimports things (items, crates, modules), whilemoddefines a module (either in another file with a a semicolon or in the same file with curly braces)Cheers
There is a hidden detail that’s worth mentioning about C and C++ style includes: Because compilation and linking are two separate steps, the definitions (included from header files) and the compiled object code that implements those definitions can be kept strictly separate such that you could have two different implementations of the same definitions that are switched in at linker time, for example as a way to provide mock implementations of a specific compilation unit for a unit test, by modifying which object file is passed to the linker.
In some languages this is harder to do as it would require some fiddling with the classpath (in Java for example) or designing with dependency injection in mind from the beginning (inversion of control).
I’m not saying it’s a killer feature, but it’s something worth noting nonetheless.
The Java logging library “SLF4J” works basically this way. It’s a “broker” sort of library. You use the SLF4J API to log and there are different backend implementations that can tie your logging statements to an actual logging backend. So you can do things like switching backends without changing code or using one backend (say, to STDOUT) in local and a different one (Syslog, maybe) in prod.
I suspect most developers have a “favorite language” and are just going to respond with “the import style my favorite language uses”.
I would never classify Python as my favorite language, but I really like the import system.
What matters is looking at a file and being able to tell what was defined where and imported from there.
Go’s:
import "foo" // foo.X import b "bar" // b.X import _ "baz" // X // or, the same but less verbose: import ( "foo" b "bar" _ "baz" )It’s clean, clear, has renaming, allows multiple or One Big import statement, and þere’s still only one keyword.
My… no pun intended… go-to language. ;)
Perl’s system of
use ModuleName LISTis pretty nifty.The LIST may contain pretty much anything and it’s passed to the module’s
importmethod (or one that the module has imported in turn from elsewhere). It can be interpreted or parsed freely by that routine.Usually, it’s expected to contain the names of variables and subroutines to be linked into (some might say “pollute”) the current namespace, as well as special flags that can import multiple at once, but it can also be left empty with
()and then everything in the module has to be referred to by prefix in a very similar way to how C++ does it.But in theory, you could put an entirely different programming language into that LIST and have the module do magical things with it. If you thought the fact that C++'s template system being Turing complete was insane, this is basically that, but to the next order of magnitude.








