[awaythrowact]

I second most of this.

A lot of problems are fixable with time and money. Maybe the Series A will help!

But some problems might be related to Julia's design choices. One thing I really missed in Julia is Object Oriented Programming as a first class paradigm. (Yes I know you can cobble together an approximation with structs.)

OOP gets a lot of hate these days. Mostly deserved. But in some large complex projects it's absolutely the right abstraction. I've used OOP in several Python projects. Most of the big Python DS/ML packages use OOP.

Maybe you think PyTorch, SciKit, etc are all wrong, and would be better off with a more functional style. I know it's fashionable in some circles to make fun of the "model.fit(X,Y); model.predict(new_data)" style of programming but it works well for us busy professionals just trying to ship features.

I don't think Julia is wrong for enforcing a more functional style. It probably makes it easier for the compiler to generate great performance.

But Python has achieved its success because of its philosophy of being the "second best tool for every job" and that requires a more pragmatic, multiparadigm approach.

[amkkma]

How is "model.fit(X,Y)" better than "fit!(model,X,Y)"?

Julia is object oriented in a broad sense, it just uses multiple dispatch which is strictly more expressive than single dispatch, so doesn't make sense to have dot notation for calling methods because types don't own methods.

For giving up some facility in function discover, you get speed, composability, generic code...and a net gain in usability because you can have one array abstraction for GPUs, CPUs etc etc, which is just an instance of having common verbs across the ecosystem (enabled by MD). Instead of everyone having their own table type or stats package, you have Tables.jl or Statsbase.jl that packages can plug into and extend without the issues inherent in subclassing, monkeypatching etc.

This is a much better, more powerful and pleasant experience

Closing the gap in Method discovery will simply require a language feature with tooling integration, where you write the types and then tab to get the functions. There's already an open issue/PR for this

[east2west]

I don't miss OOP in Julia but I do feel there need to be more ways to abstract things than multiple dispatch and structs. One thing I do miss is interfaces, which can group functions for a common purpose. I understand it may be not feasible in a dynamic language, but hopefully there will be something above functions as an abstraction mechanism.

[freemint]

Interfaces are not currently defined in the language but socially if you are lucky the package whose abstract type you subtype has a generic test you can call with your new implementation but that is not the standard.

[awaythrowact]

I'll give you my two cents, recognizing that I very well might just be ignorant about Julia and multiple dispatch, and if so please continue to educate me.

Consider if we want to run many different types of models. Logistic regression, gradient boosting, NNs, etc. We want the ability to easily plug in any type of model into our existing code base. That's why model.fit(X,Y) is attractive. I just need to change "model = LogisticRegressionModel" to "model = GradientBoostingModel" and the rest of the code should still Just Work. This is a big part of SciKit's appeal.

But all these different models have very different training loops. So with "fit!(model,X,Y)" I need to make sure I am calling the compatible "fit" function that corresponds to my model type.

You might now say "Ah! Multiple dispatch handles this for you. The 'fit' function can detect the type of its 'model' argument and dispatch execution to the right training loop sub function." And I suppose that's theoretically correct. But in practice I think it's worse.

It should be the responsibility of the developer of "model" to select the "fit" algorithm appropriate for "model." (They don't have to implement it, but they do have to import the right one.) The developer of "fit" should not be responsible for handling every possible "model" type. You could have the developer of "model" override / extend the definition of "fit" but that opens up its own can of worms.

So is it possible to do the same thing with "fit!(model,X,Y)"? Yes of course it is. It's possible to do anything with any turing complete language. The point is, which system provides the best developer ergonomics via the right abstractions? I would argue, in many cases, including this one, it's useful to be able to bundle functions and state, even if that is in theory "less flexible" than pure functions, because sometimes programming is easier with less flexibility.

[amkkma]

Thanks, I see where you are coming from.

>It should be the responsibility of the developer of "model" to select the "fit" algorithm appropriate for "model." (They don't have to implement it, but they do have to import the right one.) The developer of "fit" should not be responsible for handling every possible "model" type. You could have the developer of "model" override / extend the definition of "fit" but that opens up its own can of worms.

It's really the same thing as python, just better...I don't see the distinction you are drawing.

In python you have a base class with default behavior. You can subclass that and inherit or override.

Julia has abstract types with interfaces...instead of relying on implementation details like fields, you provide functions so that more types of models can work even if they don't have that one specific field. Otherwise everything is the same where it counts,- you can compose, inherit and override. Even better, you can work with multiple models and types of data, inheriting where you see fit.

I don't see any benefit to python's restrictions here, either in ease of use or in expressiveness.

For all intents and purposes it's a strict superset.

Even better, you can use macros and traits to group different type trees together.

https://www.stochasticlifestyle.com/type-dispatch-design-pos...

These seem to be in contradiction:

>It should be the responsibility of the developer of "model" to select the "fit" algorithm appropriate for "model.

>You could have the developer of "model" override / extend the definition of "fit" but that opens up its own can of worms.

It's the same in python, either you inherit Fit or you can override it. What's the difference with Julia?

Except in julia all types and functions have a multiple dispatch, parametric type and possible trait lattice of things you can override, customize and compose, so that even if the model author has to override fit, they can do it using small composable building blocks.

[awaythrowact]

I agree you can achieve same benefits with Macros. Indeed, I see that MLJ, Julia's attempt at a SciKit type project, makes extensive use of Macros. But I personally think macros are an antipattern. In large projects, they can introduce subtle bugs. Especially if you're using multiple modules that are each editing your code before compile time and that don't know about each other. I know others in Julia community agree that Macros are dangerous.

I think abstract types are a brittle solution. The "can of worms" I alluded to is something like this: Library TensorFlow implements model "nn" and Library PyTorch also implements model "nn" and they both want to override "fit" to handle the new type "nn"... Good luck combining them in the same codebase. This problem is less pronounced in OOP where each development team controls their own method. Julia devs can solve this by having every developer of every "fit" function and every developer of every "model" struct agree beforehand on a common abstraction, but that's an expensive, brittle solution that hurts innovation velocity.

I think the closest I can do in Julia via pure structs is for the developer to define and expose their preferred fit function as a variable in the struct, something like "fit = model['fit_function']; fit(model,X,Y)" but that introduces a boilerplate tax with every method I want to call (fit, predict, score, cross validate, hyperpameter search, etc). (EDIT: indeed, I think this is pretty much what MLJ is doing, having each model developer expose a struct with a "fit" and "predict" function, and using the @load macro to automate the above boilerplate to put the right version of "fit" into global state when you @load each model... but as described above, I don't like macro magic like this.)

[randomswede]

None of these require macros.

If (and only if, I have not looked at our hypothetical model and fit GF, but for the sake of argument, I will assume that it does) "fit" specialises on "model" will a "mode = AnOtherModel" cause "fit(model, x, y)" to be exquivalet to Python's "model.fit(x, y)". If you need to provide a custom "fit" method, you do so by providing a method specialised on AnOtherModel to the "fit" GF.

At no point is there a macro involved.

As for the "module a, model nn" and "module b, model nn", I would naively assume that they actually are different models, and therefore something specialising on "a.nn" will not get dispatched to when you pass a "b.nn".

Disclaimer: I don't actually know Julia, at all. But I have written substantial amounts of CLOS code (and Python, but I like CL and CLOS better).

[awaythrowact]

I know nothing about Lisp, so at the risk of talking past eachother....

I never said macros were required. I said implementing this type of code without OOP required more boilerplate, and MLJ uses macros to reduce that boilerplate.

As I understand module imports in Julia: Each module developer exports a list of publicly facing objects. Obviously "fit" and "model" would be among them. If you import two modules that both export a new "nn" subtype of shared parent type "model", and both extend "fit" and "predict" and etc to accept their own subtype "nn", then you have to manually specify which module you are referring to every time you call nn, or fit, or predict, or whatever. Is that wrong? If I just import PyTorch, import TensorFlow, and then call "mymodel = TensorFlow.nn; fit(mymodel, mydata)" then Julia doesn't know that the "fit" I am calling is the TensorFlow implementation and not the PyTorch implementation; what if I had WANTED to use module A's "fit" on module B's model, and they intentionally adopted the same abstract type system to enable this interoperability? So instead I have to write "mymodel = TensorFlow.nn; TensorFlow.fit(mymodel, mydata); TensorFlow.predict(mymodel, mynewdata)". Obviously the extra typing is mildly annoying but the bigger problem is potentially introducing bugs by mismatching modules, and the developer's cognitive overload of having to keep track of modules. Python style OOP is a more elegant solution to the namespace problem and results in more readable, maintainble code, at least in my opinion. Anyways, maybe Julia has a more elegant solution I'm not aware of, if so I'd love to hear it.

[freemint]

Your explanation is stop on for Julia.

[gbrown]

Have you checked out MLJ? I think their interface does a pretty good job of what you're discussing.

[xiaodai]

Hmm, reading your comments. I think you have understood how `fit!` works.

Each develop will develop their model in their own package. But they just define this fix function. E.g.

``` # in PkgA function fit!(model::ModelTypeA, X, y) # some code ... end ```

``` # in PkgB function fit!(model::ModelTypeB, X, y) # some code ... end ```

Each `fit!` function is distinct. It's the same as `model.fit` where the `model` is controlled by the individual developers.

[chrispeel]

> One thing I really missed in Julia is Object Oriented Programming as a first class paradigm.

Julia peeps would tell you that the multiple dispatch used by Julia is a generalization of OOP. And that they like multiple dispatch

[cbkeller]

Yes indeed. While Julia does generally eschew the "model.fit(X,Y); model.predict(new_data)" style, it's not because it's functional, it's because it's dispatch-oriented, which is arguably a superset of class-based OO, and even perhaps arguably closer to Alan Kay's claimed original vision for OO than modern class-based OO is.