How to learn Julia, a new programming language 

Answered a number of the questions, some my self but most from comments. Tried getting back into Julia a few more times, didn't really click. Tried figuring out a way for using it at work, but kept running into issues. I have hope for Julia though. Will try it again in a few years

Vscode works out of the box now. Last time I tussled with Julia (a year ago?) I once again gave up on getting the language server to work correctly in emacs and just used vscode

Oh I see. Thanks for the tip! I don't think I would get along with Julia's designers 😅

@@JackofSome ​ I'm sure you would, they are very nice and often have solid reasoning behind design choices. In that case, the most basic property of addition is that its commutative, but string concatenation is not ("foo" * "bar" != "bar" * "foo"), so + is really the weird choice here. And there's really no serious downsides to using * instead of +, and no "but I'm more familiar with +" isn't a good design argument. Regardless, if you really want to use + you can just define in your startup.jl file: Base.:+(x::String, y::String) = x * y And now you have +, with no performance penalty compare to the built-in *.

It was meant to be a joke :), the not getting along bit not the disagreements with design choices. There's a number of things I don't like about Julia (e.g. 1 based indexing) but can understand where they came from (likely easier on the Matlab crowd). There's also a number of things I really appreciate. Multiplication is also commutative though, and to me concatenation is more akin to an (ordered) union which feels more like an addition than a multiplication but to each their own if course.

@@JackofSome It's not commutative in general, for example for matrices. But I'll concede it's a bit of a technicality. For unions you can use the \cup+TAB in the REPL (or in most editors) : [1,2] ∪ [3,4]. It's true that Julia tends to follow more closely mathematical notations and concepts that some other languages. Note there's also OffsetArrays.jl when different type of indexing is useful (there's some problems that are better written with 0 being the middle of the array for example). IIRC it's a zero cost abstraction.

@@STAR0SS not only zero cost but there is a [post](discourse.julialang.org/t/performance-of-offsetarrays/5769/5?u=rapus95) which suggests sometimes even being slightly faster.

I'm aware. I very much dislike it in a language like Julia since there's no reasonable way to make it obvious.

Thank you so much for the kind words. Yes I to computer vision and deep learning for a living. I'll definitely make videos about that too I just have to find the right topics and format. You can follow me on Twitter in the mean time for my ramblings 😅

@@JackofSome Thanks i'll do that ;)

Having worked at Mathworks there's a lot I love about both their tools and their philosophy. I just can't justify using it for my startup (and I still think it encourages bad programming practices).

At the time there wasn't. For fast moving projects like Julia there's almost never an up to date package in most distros :)

@@JackofSome see that's why I use archlinux ;)

I was on arch during the video 🙂

No. I mention this briefly in the Julia video. The number of iterations was the same.

@@JackofSome Thanks for the reply. I overlooked that. You don't need to create the range with [0.0:dt:T;], since this will allocate for the full array every time you run the function. You can simply use 0.0:dt:T instead, and this reduces allocations and run time. Sometimes preallocating without filling with zeros saves time: Vector{Float64}(undef, 1000). This just grabs some chunk of memory without replacing the junk inside. Here you have to make sure your function modifies every element in your array, otherwise you end up with funny values here and there. I'm not very experienced, so I asked on the official forum for further tweaks: discourse.julialang.org/t/how-to-optimize-the-following-code/33209/6 Thanks for the very informative video!

I'm too embarrassed to upload it. Lemme clean it up a bit and then I will.

@@JackofSome No problem, I am not judging. Actually I Just switched my spacemacs to develop branch and I am starting using doom modeline. So I just want to get some idea of configuration. BTW mine is not messy, it's super messy. 😜

Pluto didn't exist yet 😅 It's on my list to try out though.

How well can you manage packages?

Well then

Julia is probably my favourite language but if there's one criticism I can understand is its use of 1-indexing by default. It was strange for me too, having used python, C and Java before that. But I don't think it's a deal breaker, bearing mind all the features of Julia. And I can also understand the reasoning behind this choice: Julia is primarily for mathematical and scientific computation; and 1-indexing is more like maths speak. You don't say "the zeroth entry in this matrix" or "the zeroth row in this data frame", you say "the first entry in this matrix", "the first row in this data frame" etc. languages like C use 0-indexing because the index is an offset from an address where the array begins. That is, array[i] is really a syntactic sugar way of writing *(array+i). Fine, that makes sense if you're programming at a low level language. It's not necessary with a high level abstraction you get with Julia. In fact, it's not necessary in python either (and python doesn't even have arrays builtin! the nth item in a python list isn't offset by n units from whatever, it can be stored anywhere in memory!), it's just to uphold the tradition/convention (which is ironic for possibly the only language with meaningful whitespace/indentation but nm)

It's not __actually__ a sole deal breaker for anyone. I coded for years in Matlab, which was one of the inspirations for Julia. That's the real reason it has 1 based counting. There's nothing wrong with that honestly، just like there's nothing wrong with 0 based counting. Julia has a number of other niggling issues though and its syntax is a bit all over the place. All that combined with "oh crap I also have to completely change my mental model for indexing now" probably turns away a fair bit of people.