## Where communities thrive

• Join over 1.5M+ people
• Join over 100K+ communities
• Free without limits
##### Activity
Bor González Usach
@bgusach
And, moreover... they just produce side effects... that is the opposite of function purity
Jean-Jacques Dubray
@jdubray
they are pure fonctions with respect to the model
It is only with respect to the model
Bor González Usach
@bgusach

Uhm... I don't find it very intuitive to be honest.

Your idea might be brilliant, but I really can't understand it properly. You talk about pure functions (that return nothing and can perform almost any action), then the equations that describe your model have this term M.present( A(data) ), which I can't understand since the action functions return nothing. I also find kind of confusing that the SAM acronym stands for State-Action-Model, but when I take a look at your examples, I can see there is a fourth element called view (and so far I cant get my head around about the idea that state and model are different things, but maybe because something has not clicked yet in my brain)

I don't know, I am no specialist in this domain and maybe I'm trying to bite more than I can chew, but I think you have to make your material more didactic. Just my respectful opinion.

Jean-Jacques Dubray
@jdubray
Yes, your point is valid, the difficulty comes from expressing a reactive loop from an expression. Logically the view is computed from that expression. However, the pattern is reactive so it is the action which calls the present function, which can be viewed (and probably should be) a closure on the model.
All actions as structured as:
function action(data,present) {
var data_ = purefunctionAction(data) ;
present(data_) ;
}
And for the model:
function present(initialState) {
var model =initialState  ;

return ( function (data) {

... implementation of the present method ...
} ) ;

}
Jean-Jacques Dubray
@jdubray
Again, there is only so much you can express in a single expression, hopefully it still helps to understand the pattern.

I think you have to make your material more didactic.

yes, but that only happens over time, as people ask question and I understand where the disconnects happen

The View is the "State" representation. It is totally subject to the "State", it cannot stand on its own. I believe that has been one of the root cause of all MVX patterns, it is best to think in terms of business logic only because the view has none, that's why the pattern is not called VSAM (disclosure doesn't sound as good). That being said, V, the state representation, can actually be anything, SAM is really the pattern.
Stardrive ENGG
@HighOnDrive

@bgusach Progressive rock band and then a programming rock band in the form of a software project, best of both worlds :smile:

I understand where you are coming from about not finding SAM the easiest think to understand. Myself I'm here because I have discovered a pattern myself and find them very useful in managing complex reactive flows.

For sure there is a real need for a pattern to emerge that supersedes the half-full boiler plate oriented narrow boxes that frameworks try to lure you into, until you discover that circles just will not fit into squares, much to your dismay and against all the hype.

For me the SAM acronym itself makes sense because state drives all actions which affect models. That's how it lines up with how I define and use control states anyways.

Stardrive ENGG
@HighOnDrive

I think a nice top to bottom comparison table with one column per pattern/framework would be a helpful way to see the purpose of SAM. Just a succinct breakout of how different options handle the reactive loop and interact with the model, be it on the client or the backend.

Then after that is clear then various ways of wiring can be highlighted. Then finally code examples follow from that. The SAM pattern becomes lost in the shuffle so to speak if implemented in a specific wiring option before the comparison table is acturately presented.

This whole flow of revealing what SAM is follows a model (comparison table) then state (resulting wire options) then view (wired up code samples) format as well.

Jean-Jacques Dubray
@jdubray
Thank you Tom. I have actually the first table for React: http://jdubray.github.io/sam/#react
@gunar also shared an interesting video from the React Conference which started at ... V = f(M) and went on to compare React with other frameworks. https://www.youtube.com/watch?v=Pm0uwncSCo4&feature=youtu.be&t=14366
Esteban Negri
@enbits
Here's the standalone talk from Andrew Clark: https://www.youtube.com/watch?v=ZVYVtUFDf28
Jean-Jacques Dubray
@jdubray
thank you, I'd really be curious if he read the article prior to preparing his slides.
Esteban Negri
@enbits
I'm watching it for the first time and I see what you mean
Jean-Jacques Dubray
@jdubray
I say that because in everything I had read about React prior to the article, I had never seen someone saying react is v = f(d), and for me that was the ah ah moment that started everything. Once you understand that it's that simple, then you want/can close the loop.
The react component model itself does not offer a good way to close the loop that's why they ended up adding Flux/Redux, then Thunks then Sagas, then GraphQL, then Relay...
that's a bit too much, if they had started with v = f(d) (and JSX) they probably would have created something that was 99% simpler.
You also see from his presentation that stream don't really add much to the architecture:
v$= f(d$) or v[] = f( d[] ) does not change anything, just some wiring.
IMHO, react is out of control
Esteban Negri
@enbits
That's what confusing me lately. Seems like the big ones (I mean, Google, Facebook, etc) doesn't know what to do. I'm pretty new at these Javascript frameworks and I wanted to use Angular as a base for all future developments, but with all this Angular 2 ramble I'm really lost and I thought it was just me, then found your article.
And PHP frameworks keep working with MVC so that adds more confusion when people talk about using Angular + Laravel to build apps
Jean-Jacques Dubray
@jdubray
They sure don't make it easy. I say SAM gives you a good enough structure to start with, you can choose how much framework you really need on top of that.
For the sad part is all these frameworks tend to obscure the value of HTML5/CSS3 because they want you to use plain vanilla HTML
Esteban Negri
@enbits
Well I like minimal stuff and Angular is a monster, never understood V1. So what are the key js libs that I should use if I want to implement SAM?
Gunar Gessner
@gunar
You don't need absolutely any lib to implement SAM. That's the beauty of it.
That being said, I like myself some React (or any other VirtualDOM lib) for the View.
NB: There is no V in SAM, so it's really a matter of taste and doesn't change anything.
Jean-Jacques Dubray
@jdubray
+1, you can use anything for V, Stateless React Component I believe are pretty good. You can take a look at the sample list: http://sam.js.org/#rocket
Esteban Negri
@enbits
Thanks guys!
Aliaksei Kuncevič
@kuncevic
This message was deleted
dnson
@imnutz

@jdubray my another attempt using SAM to build webapp :relieved:
https://github.com/imnutz/dashboard

In this, I tried to split the webapp into folders and components.
May you take a look?

Jean-Jacques Dubray
@jdubray
sure, you are amazing! thank you for all your contributions.
Jean-Jacques Dubray
@jdubray
the actions are about validating and enriching
It looks great except for one thing, I would not put the CRUD operations (services) in the actions and call these actions from nap(). Logically this is truly what happens but IMHO, it is better when the model persists its values
There could be some race conditions that occur when you handle the persistence with nap()
dnson
@imnutz
@jdubray so from what you said, there should be some thing like this in the model:
model.fetchingData = function() {
callRestAPI({})
.then(function(response) {
- update model's data
- call State.render
})
}

model.updateContact = function() {
callRestAPI({})
.then(function(response) {
- update model's data
- call State.render
})
}

// same for create and delete
Jean-Jacques Dubray
@jdubray
Yes, but this is called from within the model, not the actions.
the rationale is that there is no point in rendering the view until the model has persisted and the persistence layer has also accepted the values of the model. If you could be guaranteed that whatever value has been accepted by the model will persist as is, then you could use actions in nap().
the other aspect to consider is concurrency. As soon as actions can be triggered and processed before the present model has completed, then you will run into concurrency issues. The SAFE middleware "blocks" the present method and serializes all action execution, in particular it would check for the "allowedActions" in the new State before processing an action waiting in the queue.
Jean-Jacques Dubray
@jdubray
I believe that's the problem with React, the reducer could technically process more than one action, then depending which one finishes first would be immediately overwritten by the last one.
You have to always ask the concurrency question when you build such a system.
dnson
@imnutz
@jdubray thank you for reviewing and explaining this.
Jean-Jacques Dubray
@jdubray
you are welcome, again happy to discuss better alternatives, this is a very important question
tomorrownull
@tomorrownull
m
Bang-Equal
@Bang-Equal
I am trying to run the crud-blog but none of the button clicks work because "actions is not defined". I dopn'
I don't see actions in server.js but I do see it in blog.js
Jean-Jacques Dubray
@jdubray
@ReketeBravo could try that repo? https://github.com/jdubray/sam-safe
Did you do npm install?
Bang-Equal
@Bang-Equal
yes