ajskateboarder wrote:

AHypnoman wrote:

Maximouse wrote:

~Snip~

My mistake (I'm not deeply familiar with Python), although given all python vars are references then why does this happen?

>>> x = "hi"
... y = "hi"
>>> id(x)
4414027184
>>> id(y)
4414027184
>>> def z(a):
...     a += "hhhh"
... z(x)
... x
'hi'

This means that strings are immutable while lists and dictionaries are mutable, but they're both passed in as refs. A string is a reference, but mostly acts as a primitive?

FYI,

a += "hhhh"
# shorthand for a = a + "hhhh"

does not mutate the variable, it just reassigns the value of a. Oh by the way, JavaScript also acts the same way

let a = "a"
function addhhhh(b) {
    b += "hhhh"
}
addhhhh(a)
console.log(a) // a

And Java while we're at it

class Main {
    public static void addhhhh(String b) {
        b += "hhhh";
    }    
    public static void main(String[] args) {
        String a = "a";
        addhhhh(a);
        System.out.println(a); // a
    }
}

AHypnoman wrote:

Also creating two separate strings assigns them the same reference if they are equivalent meaning that == and ‘is’ comparisons return the same, meaning that for almost every purpose a string acts as a js-like primitive, but isn't.

That's most likely for optimization purposes - when would you need to compare memory locations for two equivalent strings anyways?

ajskateboarder wrote:

does not mutate the variable

ajskateboarder wrote:

Oh by the way, JavaScript also acts the same way

...

ajskateboarder wrote:

And Java while we're at it

class Main {
    public static void addhhhh(String b) {
        b += "hhhh";
    }    
    public static void main(String[] args) {
        String a = "a";
        addhhhh(a);
        System.out.println(a); // a
    }
}

ajskateboarder wrote:

AHypnoman wrote:

Also creating two separate strings assigns them the same reference if they are equivalent meaning that == and ‘is’ comparisons return the same, meaning that for almost every purpose a string acts as a js-like primitive, but isn't.

That's most likely for optimization purposes - when would you need to compare memory locations for two equivalent strings anyways?

AHypnoman wrote:

ajskateboarder wrote:

Oh by the way, JavaScript also acts the same way

...

Java and JS are entirely distinct. One is used on nearly every website and the other can be found at archaeological sites.

The difference is that in JS primitives are all immutable, and, unless specified when creating one (excluding Web APIs, as they aren't a core part of JS), objects are mutable.

The reason for such behaviour is irrelevant (although I doubt this is for speed - doing this means that Python will need to utilise more memory to store a new reference which is almost certainly a slow process).

There are a few reasons you'd want to compare memory locations (such as manual memory management if you have limited resources), although my main problem with it is the inconsistency in behaviour - not lack of capability.

ajskateboarder wrote:

Yes. This is also true for Python as well. Any hashable built-in object like a str, int, float are immutable and can't be modified through a function parameter, while mutable containers (which can't be hashed) like dict and list can be

ajskateboarder wrote:

Notice I said comparing the memory locations of constant equivalent strings. What is this inconsistent with, really?

ajskateboarder wrote:

Also, meticulous memory management can be much better done through C and Cython. Python isn't quite good with controlled memory management

AHypnoman wrote:

The inconsistency I refer to is the differing mutability between types of object - strings and lists, for example, are both objects. List are mutable, while strings are immutable. Primitives are inherently immutable, so it makes sense that they are immutable in JS. Reference variables on the other hand are inherently mutable, so it makes sense that they are by default in a language (They are plenty of reasons a program may want to change their mutability, but defaults should be mutable since they are internally). In Python, everything is an object+reference and some objects are immutable. There is no reason to have this behaviour, especially since it introduces inconistencies.

This behaviour isn't uncommon in programming languages, although that certainly doesn't make it any better.

I don't mean to say JS is a perfect language, or that Python is a particularly bad one. I'm just providing reasoning for my preference.

ajskateboarder wrote:

Also, meticulous memory management can be much better done through C and Cython. Python isn't quite good with controlled memory management

Whether Python is best suited to it or not, it must be a consideration. Most AI software is written in Python now and memory management is incredibly important to that (granted, anyone writing AI in Python chose the wrong language for the job: languages like C and Rust are far faster and better for memory management so are better suited for AI).

ajskateboarder wrote:

A lot of ML libraries like TensorFlow or PyTorch (as well as numpy and scipy for statistical calculations) are primarily written with C++ and Cython, so a lot of overhead just disappears. C and Rust aren't exactly the best languages to teach data scientists/engineers due to:

- how specialized their role is anyways
- how much of the ML ecosystem already uses Python
- Python being good for rapid development, which is primarily what data science involves (source: see the amount of jupyter notebooks that people make)

That being said, Rust does have a TensorFlow binding

JavaScript also has a port of TensorFlow, but configuring web workers to load 500mb chunks of model checkpoints into the browser is a headache, and the Node API is just… weird

ajskateboarder wrote:

API design is just kind of strange with Node. It's super fragmented, and only the most updated of libraries will work, even though Node has been supposedly stable for years now. There's also like 4 competing package managers (which mostly will not cooperate with each other) and a couple of alternative runtimes (most of which thankfully aren't being pervasively adopted)

AHypnoman wrote:

ajskateboarder wrote:

A lot of ML libraries like TensorFlow or PyTorch (as well as numpy and scipy for statistical calculations) are primarily written with C++ and Cython, so a lot of overhead just disappears. C and Rust aren't exactly the best languages to teach data scientists/engineers due to:

- how specialized their role is anyways
- how much of the ML ecosystem already uses Python
- Python being good for rapid development, which is primarily what data science involves (source: see the amount of jupyter notebooks that people make)

That being said, Rust does have a TensorFlow binding

JavaScript also has a port of TensorFlow, but configuring web workers to load 500mb chunks of model checkpoints into the browser is a headache, and the Node API is just… weird

Yes - that's something I did't consider. Given the existent support for AI that Python has, it's probably best for AI at this point. I was referencing the speed benefits and requirements rather than infrastructure and syntactic suitability (which admittedly I overlooked entirely).
…

print(“Hello World”)

console.log('Hello World');

julmik6478 wrote:

Python is lot easier. Look at this python code:

print(“Hello World”)

And now look at same, but in javascript:

console.log('Hello World');

Ok, this one is easy in javascript too, but more advanced things are easier in python and in python you don't need to add “;” at the end.

console.log('Hello World')

console.log('Hello World');

BigNate469 wrote:

(#594)
Technically speaking, in newer versions of JS you don't need to add the semicolon either, just a line break.

mybearworld wrote:

BigNate469 wrote:

(#594)
Technically speaking, in newer versions of JS you don't need to add the semicolon either, just a line break.

Is that newer? I thought ASI has existed since pretty early on.

BigNate469 wrote:

mybearworld wrote:

Is that newer? I thought ASI has existed since pretty early on.

I thought it was newer (I consider “newer” versions of JS as ES5 and up, since a small fraction of people still use Internet Explorer), but I might be wrong. MDN isn't giving me an answer, nor is the internet (or even the standard itself), so you could be correct.

alwayspaytaxes wrote:

(#597)
In terms of style guides and formatting, Python undoubtedly wins. Using Black + Pylint helps format your code for basically all aspects, even how much line spacing should go between functions! ESLint is a nightmare to setup, its VSCode plugin barely works with TypeScript, and while Prettier is nice, it's just not opinionated enough for me.

mybearworld wrote:

alwayspaytaxes wrote:

(#597)
In terms of style guides and formatting, Python undoubtedly wins. Using Black + Pylint helps format your code for basically all aspects, even how much line spacing should go between functions! ESLint is a nightmare to setup, its VSCode plugin barely works with TypeScript, and while Prettier is nice, it's just not opinionated enough for me.

Can you give an example of Prettier not being opinionated enough? I haven't had that experience.