Elsa Workflows Community, 20 Mar 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Elsa Workflows Community Meeting 28

Description

Topics

- Multitenancy

- HTTP Request activities

- Workflow contexts, create sample application

A

A

Just one actually, oh yeah seems new welcome, Gloria, alright, so how's everyone doing.

B

Fine fine, you.

A

Yeah great busy I'm.

B

A

You listen to me. That's that moment, hello.

B

Where are you good to hear you.

C

My introduction or.

A

Yeah, if you want it's, it will be interesting, but.

C

You don't have to uh the microphone, not not worth sorry for why you had English, okay and.

A

No worries: it's fine.

C

In Spain, I contribute one or two poor requests at the moment to us, and I want to follow closely version. Three so I'm here to listen um and see. What's new.

A

That's great well you're welcome on the on the meeting uh good that you can join so the way this works at the start of the meeting. We come up with topics for discussion or little demos that we have prepared or just think of on the Fly and, of course, if there's any questions or if you have a request to look into some topic that fits within this one hour of the meeting that that will be good to add to the agenda.

A

So if you have anything in mind, please let me know. For the past week myself I've been busy debugging making the core engine and the design are more stable. As we've seen during the last community meeting yeah, there were a couple of issues identified during the meeting, which was fun and last week, I demonstrated the these topics here. So we talked about the flow HP request, improved Json handling for sap workflows by that I mean let's say you have an HP endpoint on your workflow. You receive some Json payloads.

A

You want to be able to treat it as actual Json, but of course, what else uh and using jint, which is a net interpreter for JavaScript syntax. You also have to deal with net types, so sometimes it could be confusing and there were a few bugs as well, so those have been fixed there.

A

We also looked at improved parent-child communication or you know, being able to pass data into sub workflows and being able to communicate it back, which is something maybe uh we can reiterate, because there were some issues when I did the demo and since you're just joined for the first time or yeah. Maybe it's interesting for you as well since you're, since you mentioned an interest in lsf3 and then and during the session, if any questions come up or topics, please let me know.

C

Okay, I see the previous video of the last meetings. I update with the points yeah perfect.

A

All right, then, we don't have to reiterate: that's yeah. Well, that's great all right! So if you guys think of something, let me know if not so I'll just start looking at the commit history, maybe something of Interest comes to mind you actually or actually this meeting I I almost forgot about it and I was in the middle of working on the following.

A

Let me just talk a little bit about that, so you know what I'm up to, and it has to do with the inputs for for these composite activities, and just to recap, like composite activity, is from the perspective of the designer is just another workflow definition, but you can configure it to be useful as activity and if it's, if this one is set to Yes, it becomes available as a regular activity from the activity picker here and of course you will, you want to be able to configure its inputs, it's its settings like what we see here.

A

So this is a custom fields on the get sign of number activity here and to configure that you go to the input output tab. So here, I'm just gonna, just on a different workflow that uses this composite activity, but we switch to the definition of the activity.

A

I can show you the input and output properties. Just let me start the engine all right that started. Let me refresh so here there's a workflow called get sign of number, and the idea is that this this workflow will accept some inputs, some number and it will return the sign of the number. So that's, either positive or negative, so positive, being 0 or greater and negative anything below zero and also as a as a test.

A

It can also return two possible outcomes, positive or negative, which allows you to connect an activity to the positive outcome and to the negative outcome and then, depending on the actual value, the sign being positive or negative to positive or the negative outcome will be executed.

A

So this is just to showcase the set workflow output activity to set the output here. So it's an activity and you can configure it by selecting the available outputs. So so this list is populated from whatever you have here and here you can provide an expression to provide the value for the output, and this value then becomes available to the workflow using this uh this this activity yeah, so any in any case.

A

So we went over this during the last meeting and while I'm implementing this for customer there's, there's a an issue with the input and the way it's implemented. So if you use it in a very simple way like I'm doing here, actually this is actually the culprit. Let me move this away for a second, let's say: use it like this, then there's no problem. All of the activities can just access the inputs by using JavaScript in this format. So if you have an input, called number request, a value using the function get number.

A

So this is made. This is a demi convention where you can just prefix with get and then the the name of the input, and it will return the value of.

B

The input it is done by chint or is it, is it implemented by you? That's the get number function.

A

Yeah so Elsa declares the functions, so declares functions in jint and also provides the implementation so.

B

A

Dynamic configured so the gene run, the gene engine is configured dynamically with available functions. I can actually show you real quick, so this is how it works. So here we have agent engine. So this is a JavaScript interpreter, and here you can see we can just set the values on it, so that works by setting a key and a value, and that, if you put in a a Lambda or delegate I mean this is this is what gets executed when you invoke this as a function. So.

B

A

Very very extensible area.

B

Yeah ID back this week down to the chint debugger, because I I need to understand the functionality behind it uh because I added dynamically inputs and output values and need to know when to get it the literal or at the JavaScript way. And but but it's a little bit different in LS3 as in as A2. But that's fine, good yeah.

A

For sure there there's a few differences, yeah, so um yeah, then you're familiar a little bit with the engine, yeah good yeah. So so that's dynamically made available. The issue is: whenever this the workflow gets suspended and goes out of memory, then these inputs would be re-evaluated, but if, if you go one level deeper I, don't quite fully understand to be honest, the the actual issue, but you basically don't have access anymore to the outer context. And so what would happen is whatever value is set to this input.

A

So, for example, going back to the consumer here- and let's say this is not a literal but a actual expression or in fact this is an expression, because it's a JavaScript, this expression would be re-evaluated right before the inputs are evaluated of the composite activity and if, if there's too many levels, then it doesn't have all of the context, so that would then uh well basically break the inputs. It would not be able to to find the uh the requested input.

A

So it's it's very buggy at this moment and and that's what I'm fixing and the way to fix it is to Simply copy. The whatever input is provided so as soon as this activity gets scheduled for Execution. At that point, its inputs should be evaluated like with any other activity. So that's the same way. It works with with these non-composite activities, those inputs get evaluated once and then they are available to the activity.

A

The difference, of course, between a composite activity like that sign of number, is that it can potentially be long running so it has its own. It is. It is its own State machine as compared to others like ASP response request. They don't they are not long running. They just perform some functionality in C, sharp and then they're done. So that's a very short-lived context.

A

If you will- and this is long lived so that makes it a little bit complicated, but to fix that is just to evaluate the inputs right before this gets evaluated, as is with the other activities and then stored in local variables, local to this activity, and then those evaluated values are available as long as the instance of this activity lives. So so that's what that's currently my focus.

A

It's uh it's an important fix when you actually use this this feature with the sub workflows or composite activities. uh Yeah you would get. You would otherwise run into troubles uh pretty quick. So it's important to fix and another issue that I will work on right after is scoping right now. These variables that you see here they are Global in scope and that's also the case.

A

If you go into a sub workflow, like the get sign of number all of its variables, if you were to declare a variable that has the same name as the variable in the parent workflow one or the other would overwrite one or the other, depending on which one gets evaluated first, which obviously, is uh it's a a big problem? Because now you have to make sure you don't use duplicate names, but you shouldn't have to worry about that.

A

It should be just like in C sharp when you have different Scopes nested Scopes should should be able to see variables from Outer Scopes, but not the other way around, and that's how it's going to work here as well. Nice yeah, so that level of isolation is gonna, make it make it much better, and that also brings us much closer to being able to provide a uh or doing a proper, stable uh release, even though it won't be feature complete without two as of yet but at the very least, it will be useful and yeah.

A

That's a good thing, I'm, actually implementing a quite complicated workflow for a customer once that works, I can be sure that it will be. It will be useful for many other scenarios as well. So.

B

Yeah yeah, you put it the HTTP request in the workflow. Could you do it again please, and what are the dotted lines, the unmatched status code? What what does it mean? Thank.

A

You for asking me, because that's that's something I would like to demo it's new features since last week and let's create a new workflow. So there's two two HP requests, there's ASP requests and there's flow. Eight speed request. So let's talk about this one first and then I'll also explain this one so forget about the word flow for a second. Just just know that this is an 8 speed, request activity and what it allows you to do is make a HTTP call to some endpoint.

A

So let's say request here: this is uh let's imagine this is our backend to get user information so I'll copy. Just this URL.

D

A

Paste it in here, so whenever you make an HTTP call, you can usually well usually you you will test for the status code, it's being 200, okay or maybe sometimes you also want to handle the four or four or maybe even 400, in to test. If invalid input was sent.

A

So actually, let's test the ladder, that's a better, better use a better example. Let's say we do post so there's two versions of post, one that will return successful, which is 200 and one version will return. 400 unsuccessful, and that's that happens whenever you forget to provide a password, for example. So so let's say we want to make this API call so we're gonna register.

A

So we'll use this. It will be post request and here will provide some some Json, so we'll say or actually this is a return now as a JavaScript object which will be serializing to a string. You should be able to also just provide a string, Json string directly, but I didn't test it and I'm afraid to try it right now. So let me just do it like this, because this has to be tested.

A

So let's say we want to send this payload all right, so this will return 200, we know, but we're also going to be potentially receiving a 400 response and that's where the expected status codes come in. So let me first tell you how you would deal with the potential response before we added this field. So if you forget about this for a second, the way to do it is you would look at the outputs and then you would capture the HP response into a variable.

A

Let's do that it will be of type HP response, I'm putting these stories as memory, because I know that the XP response is not serializable. The future will make this easier to automatically hint or even disable storage providers that aren't supported for given a certain data types. But for now you have to be very conscious what storage mechanism you need to use based on the type that you're gonna store in a variable, because these variables are always going to be persisted and the assistance provider depends on the the storage driver selected here Okay.

A

So Let me refresh that list so now, I'm going to capture the response and you based on this response, I can do a flow decision and in this decision I would say.

A

um Let's say: if responses that is code I believe is the name the intellisense is broken because because of a small bug, I think if I refresh it should work so I think yeah Status codes. So if it's 200, then we know it's okay. So this is what you would do before this change that I'm about to show- and here of course you would say- is hundreds okay, something like that and then based on the you could continue this way or this way.

A

But this is this extra step is now in unnecessary because you could do it easier with the expected status code. So here you could say I'm expecting 200, but also 400s, and maybe even two or one in case. You want to distinguish between these two and any number you like, based on this inputs, you will get this additional ports on on your activity. So, instead of going through this flow decision, you can now directly connect the next activity. So it's a little bit easier to to work with.

A

So that's the idea here, yeah, it's pretty good. um It also has an element status code. So if you want to handle any unexpected code, you you can do that as well. Of course, like.

C

This it's possible to disable the default output of um unexpected status codes.

A

No, it's uh well not the way it's implemented now, but if, if you think that's a useful feature to have, then we can absolutely configure it it's flexible enough to to be able to implement that functionality. So so we could have a checkbox somewhere that that either enables or disables Ahmed status code. That's uh that could be a good enhancement actually, but.

D

A

Will always be there, but if uh again, if you, if you think it's it's good to be able to hide this or you know not support it optionally, then uh then maybe we should uh consider that okay.

B

We do not connect anything to that port to that output. You nothing happens, so it doesn't matter if it's there or there or not,.

C

To have um more clean designer or okay.

B

C

Aware for the instance, Maybe, okay, yeah.

A

Yeah to be clear, if if if this is not here and the response is none of these expected status codes, the uh the activity will just end, but if there's no other additional activity to execute the entire workflow will end depending on, if you have branches, of course, but in this specific example, this workflow would just end immediately, so it would not go to any of these. Obviously, but just end yes, yes, yeah, it's! Okay, all right all right good!

A

So now that we have seen this one, we can look at the other one and that's HP request with the lmat status code and this both activities, HP requests and flow. Eight speed request they inherit from the same Base Class, but the difference is well. First of all, I don't see the setting anymore. Let me see if that's a browser gas issue, yeah okay, so that was some old script.

A

So here you can enter this the same thing, so the expected status codes for hundreds to a one and, as you can see here, instead of showing these ports here, it shows uh embedded orgs if you will so these are embedded and the difference is that what you can do with this activity with this um yeah, this flavor is you can handle 200 responses in its own little sub workflow. So here you could do whatever you want, maybe write something to the console.

A

So what will happen now is that after this one executed, it received the response it will execute. If, if the response status code is 200, it will execute this Little Flower workflow here.

A

If it's 400, it will execute this one Etc or if, if none of these mats, then you can do something here and what's nice about this uh process that you can do whatever you want before and then continue on like this, so this will be executed or this one or this one once the entire 8 speed request and and its child activities finished, the the main workflow continues on to this one.

B

Okay, um if if a sub workflow is is called, is the original response, the the body of the response available within the sub workflow yeah.

A

Yeah exactly so, it works the same way. So you you have this variable this. These.

E

A

Are accessible at any level, so yeah exactly like like this, for example,.

C

It's similar to create a workflow on Marcus, you say well as activity and use it for the outcomes.

A

um Sorry I didn't get the question. Can you repeat it.

C

um Is this juice is similar to create a workflow and Mark it as.

A

Yeah this one here it is yeah actually yep yeah. It is it's similar because when you create a sub workflow like this, so let me undo this thing here and let's say the.

C

Question is in the um in the inner: um it's create a workflow.

A

Sorry here, yep yeah, so creating a workflow. No, so instead of creating a workflow, it's actually. What happens is that right now this? This is like a property. It's a well! It's it's, a port that stores an activity. So here you can store something that implements eye activity. So what the designer does it instantiates a flowchart activity. So this entire thing is a flowchart activity and now here's another flowchart activity- and this is the designer for that flowchart activity. So in fact you see it here, flowchart one or actually that's sorry, that's the root.

A

uh I think I I hate it because it's implied uh it didn't add any value. So this is the roots flow chart. Then there's send HP request. This 200 contains this current flowchart activity and this flowchart activity then has and activities collection, and that's currently has this one activity, which is this one. So it's not an actual workflow definition. It's just uh a composite activity, a flowchart yeah.

C

Okay, okay, yeah.

A

Good question um yeah, so that's basically it it's like a hybrid model. um The the this is more of a sequential activity to you know to use the terminology a little bit from Windows workflow Foundation, where they there. They have a strict separation between the flow charts and sequence diagrams, and this designer attempts to be a hybrid, because it's easier from the perspective of migrating from Elsa to 2003, where you still want support for flowcharts.

A

But um it's too soon to have a separate, sequential diagram. Editor I mean it would be much more work and it would delay the release. So this is like a uh um intermediate version of of flowchart editor and the only disadvantage that I think at this moment is that it can be confusing to have two flavors from the Picker like here.

A

You have HP requests and Floyd speed request the way I'm going to solve it is to allow the integrator or the developer to configure the available activities, and it's up to the you know to those developers to implement Theirs to decide whether they prefer this kind of this kind of activities or these kind of activities. At the very least, we have all of these options to choose from.

C

Of doing the same I'm.

A

C

With this everybody, we have two two style or two options to doing the same. You know exactly.

A

Yeah exactly you have two options to do the same thing and in both cases you have HP requesting Floyd speed requests both do the same thing, but the the way to model your process differs so clearly here after this 200 completes.

A

There's no way to continue using this approach, just from the 200s to, for example, right line, because right line is connected to this HP request, regardless of the outcome here uh versus this sorry, this one here the same outcomes. But here you don't you you don't have these embedded ports, so you can directly connect the outcome to the next activity. So it's a matter of of your particular use case.

A

What makes more sense, because what you can do with this one is have a like a a done outcome like you have an answer too, where, where you have activities like the effectivity, it has a true Brands, a false Brands and then a done brands, and even though it's it allows you to continue with the workflow after either the true or the false brands have executed. It's also a little bit confusing because it's not visually clear, which one will execute.

A

First, and here you have at least the option to say well, if 200 executes, then that's clear or the other one, but there's no like no done outcome here.

B

All of the outcomes uh on the upload HTTP request and then you've got the same as the the normal HTTP request. That's.

A

Exactly right, yeah, you would do it like this, so then it doesn't matter which one executes you could, of course do stuff in between, like maybe this one. You want to do it like this and then join this, and then you don't care about the others. You just want to continue yeah, that's the way to do it.

C

The loops are Knight example of this, with the tone branch.

A

Yeah loop loop is uh in in L2 has is, is is a good example here we don't have a flow Loop, not yet at least maybe we'll add it, but for now there's only these looping constructs and they only support embedded ports and that fixes the issue where it could potentially go from the body and Elsa to to an arbitrary other activity. Here, that's just not possible.

A

Any everything that you have here will execute within the loop it's within the yeah within the iterations, and you won't be able to connect to, for example, this right line from this body.

C

What are the difference in in the Journal of these two options? How did how it sorry, how it.

A

Looks oh sure we can take a look at what that looks like uh do you want to look at a specific activity here that executes and then see what it looks like in the journal.

C

C

B

C

Okay, I'll, let you see the difference in the journal. Sure.

A

So, let's do it like this? Let's say we have an HP request. I would just have 200 to 400 and we'll we'll use this URL all right, so I'll copy this definition ID. So I can invoke this using the rest API.

A

So it sends wait. That's not right! It should not have returned an instance. It should have returned to it. This one, oh wait, maybe I suppose. Well, let's take a look at the workflow instances. So here we have an instance. It's finished it's not currently showing the sub status because it could be finished, but errored. Then right now you don't see that that's majorly confusing it should not display if right line, three I, don't think we are oh unless unless, let's see assume to fit, oh I see okay, so there's this sneaky.

A

So you see this is. If you have more than one starting activity, then the first one that was added will be execute for start as a starting note, and since there's no connection, nothing else will execute. So this is something to be aware of. Let's.

B

Maybe some kind of warnings also should be implemented, yeah exactly more than one starting activity.

A

Exactly yeah and or at least say a um yeah, some helpful hints because it is a valid case where you do want to allow your Warfare to start with multiple ones, but it would be good to be aware of that if in case you have it, so let's try this again.

A

Okay, so so, apparently we did a bad request because we got a 400 results, but in any case first, let's take a look at the journal. So this is the journal here we have a view of the uh the workflow here we see the what executed so first we got flows and HP requests, which is this one here we can see what we conf, how we configured the activity at the time it executed. So imagine we were in some sort of Loop and during each iteration the the input of an activity changes.

A

We will be able to see that on its uh entry in the journal. Right now we have a very simple workflow, so we only see simple data yeah, so it went from this request to HP response to with this this one. Of course, the interesting part will be figuring out. Why did it go to this route and not to this one so.

B

How could you do that you could save the response within a variable and check? What's the output of the variable or how could you figure out why? What is the 400 executed.

A

Yeah good question: uh we should indeed inspect the response of the uh thing, so we didn't capture the response, but we can try. So let's go back to this one, and so this here we also this is the parsed object, but here's the HP response, let's capture this one. So now that we have to capture this one- and let's say we come here: let's change this to get nice speed response.

A

First, let me refresh if that intellisense bug is gone, so it has a body of type any so I, don't know what the type is here, because Annie is not very helpful, but we can try we'll see what happens so now. We got an instant study finish with an hour. So probably this voltage okay fail to convert an object of response message: oh assistant, HP, EXP, ERT yeah, that's not gonna work yeah! This is not an ATP response message, of course, but um let's do object.

A

Okay and then, if this is a response message, I don't know if it has a party to be honest, so it returns a bad requests which is expected but I'm not seeing a reason. So this was the request message: let's see what what happens if we do it in Postman hmm this. uh This is okay, so I think at this point the only way the only effective way to troubleshoot this one, because, as you see, the journal isn't too complete yet to to be able to figure this out without the proper debugging tools.

A

So so I would just dig into debugger and see what the response is from there and also the request. This is like that. Maybe the content, the content type, should be determined from the content but like we can try setting it explicitly.

A

Okay, so it didn't uh automatically include the content type which I think is supposed to happen, but I'll look into that, but now it returns this information. So it went to the to the HP response, one, which is this one.

B

um The conversion from the uh from the JavaScript code to or with to to Json is automatically done so you ended JavaScript code and uh or or the conversion.

B

um Could you maybe show it again this one, the thing yeah, you told before the yeah, the content, so is it done implicit here the return and then you added the Json code and the serialization is done automatically yeah.

A

Exactly so indeed, so here it returns, a Json object, a JavaScript object and in.net. It's it modeled as an expandable object and the expanded object is detected by this HP request activity. It sees a an object, it's not a string, so let me serialize it and when.

D

A

A expanded object: it will be properly serialized into a Json format and there's a couple of types that it it knows about some of the common ones. Okay,.

B

D

We have the Json syntax and the content. Can we test Json directly in ncert, editor yeah.

A

You should be able to so just making sure looking at the source code here. So this is the base class that handles the setting up the request, so it creates an HP client. It It prepares the request.

A

Let's see, content try get so here we get content of type object so that it could be well anything basically, but it's going to expect something based on what's determined by this HP content Factory. So here we have an innumerable of HP content factories, so we have a bunch of them. So we have support for Json text form URL, and this one will match Json. It text the application Json and text Json, and here it's going to do the test. If it's a string, it will just produce a string content.

A

Hp content object. Otherwise it's gonna create a Json content using the optic. So this object is going to be serialized here to see if that works. What we can do is just copy this and switch to liquid liquid will always return a string type. So here we can just enter the Json as we would normally. So this is adjacent string and if we publish this one and send- and then it also works and as a two I think only strings are supported for HP request or only objects. One of those two I think and.

D

Similarly, syntax to Json correctly liquid liquid can we set the syntax to Json? Oh.

A

Like here yeah, not yet, but we will add support for that as well, so you can directly say just enter Json. At least we may add it I'm, because liquid is is like Json, it doesn't really add, what's value I think uh other than maybe syntax highlighting, which is nice, but it would also be nice if you could have producer. Dynamic value, so maybe liquid is is, is even more suitable because it allows you to do uh things like you know variables that some email, if you had a variable cool, some some email.

A

So it's it's more powerful. This way.

D

Okay, can you combine both of the syntax highlighting of Json but with the liquid variable yeah.

A

Probably um it would be good interesting to look at um and maybe if we were to add Json here, maybe it should have support for liquid syntax. Just as a you know, very opinionated configuration and then you have the best of both worlds and similarly, maybe so he would have Json Plus or Liquid Plus Json Liquid Plus HTML. If you are creating a templates for send email, for example, that would be nice.

D

A

Yeah and there's and it works the same way for content from HTTP response, and it uses the same mechanism to produce these uh this. This content objects. So here too, you can provide a string or an object which will then get serialized, and it will also take into account the content type. So maybe you don't want to produce Json, maybe you want to produce XML or even an octet stream. In case your object is a byte array. Then you want to send it as a file download Maybe all right.

A

um One thing I noticed today was a response to get up about the multi-tenancy topic. So I wanted to take a look at it briefly here. So it's it's here's the roadmap, here's multi tendency. So it's an old issue. It's from 2021 by Craig and elsa2, has support for multi-tenancy at the database level. So you have basically a tenant ID column that you can configure and the API knows about tenant IDs that you can pass through the various apis that are there.

A

Elsa3 doesn't have any anything like it yet um the main reason is I. I want to really um explore the options and and come come up with a good architecture and I never got a opportunity to do it properly for Elsa 2, unfortunately yeah, but thanks to the community. We have a lot of support for multi tennis in Elsa too, but I want also want to take this opportunity with this rewrite of Elsa into Elsa.

A

3 to uh you know, to come up with a a decent implementation for multi-tenancy at the database level, so that you can use a single database with the same set of tables so not having prefixed tables, for example, but actually being able to do it at the table level.

A

And then somebody pointed out Finn, Buckle I, don't know if any of you have heard of it. I had not heard of it before, but it looks quite interesting.

A

So the what I understand is that it allows you to implement multi-tenancy in your application. So it's a library that has a bunch of abstractions and concrete options for you to implement multi-tenancy, but I've I've, yet to explore, uh if, if it supports only Entity framework core, for example, or if or if it's designed in a way that it doesn't matter what persistence mechanism you use that you can, you know maybe Implement some abstraction and uh and go from there.

A

My only hesitance of using a library like this directly um is the dependency on on such a library, but maybe we can abstract it in a way so that it, you know if somebody doesn't like this using this library for some reason they could always if they wanted to use multi-tenancy, they could Implement a different uh strategy altogether, one way or another. If we want to implement multi-tenancy at the table level, the domain objects in Elsa will have to be aware of the of multi-tenancy.

A

B

Rebus for for multi-tenancy is that what it said? No, no, but you could implement.

C

E

Like that, yeah.

B

But multi-tenancy is a big point for us, because we we we're starting right with our project and we need multi-tenancy uh for our project, uh because we we have uh a project where we, where each of the customers could design their own um workflows, and um so it shouldn't be possible to see uh workflows from another from another customer. So at some point it should be implemented, but would be great if it if you could do that, but I I know it's it's a big task for Elsa.

A

Yeah absolutely yeah and there's uh the good.

E

Thing is there's on my side. Sorry, it's okay! This is Dimitri. We also need multi-tenancy, so I just confirmed that it is important as well. Well.

A

That's great because I was just about to say, uh there's another customer. That also uh mentioned it's very important for them for also three to support. Multi-Tenancy so uh sounds like there's a there's, a good, the good there's, a good reason to uh to really explore the same in more.

B

Detail right now: it's the big biggest missing feature for us, so a BP man uh would be fine too, but um multi-tenancy is, is the biggest missing feature first right now.

A

Even bigger than.

B

A

And support yeah.

B

Yeah, even bigger yeah.

E

uh On our side, I would say basically controlling the data storage. That's most critical, because our uh cases we would physically need to have a separate databases for different antennas.

A

Right, yeah and that should already be possible using, for example, artifact where you can set up a tenant, aware service containers.

A

B

Haven't heard it before, but yeah.

A

Yeah so there's a perfect multi tenants, so uh here they go explain what is multi-tenancy in their View, some general principles and essentially the way I understood it is that you will have a container per tenant. You will have probably a root container and then child containers where a container like the difference have between Singleton and scoped. You would also have another flavor of let's say container, which is then tight to uh to a tenant.

B

And you see we could use it already, this uh uh this type of containers to to store data um for Elsa.

A

I think so, because uh I and to be clear, I haven't tried it. So I could be wrong, but my understanding is yes, uh because you don't have to use the.net built-in iservice collection default implementation. You could replace it with the artifact implementation of that interface and then also through that through through this package, you can then take advantage of multi-tenants.

E

Correctly and again, this is just purely so. It's checking you know we can use uh artifact, but rather than use a default registration of services that provided right now, as extension, we'll just register ourselves using this container uh tenant specific about attack exactly.

A

Yeah and uh I think what the way it will, or at least one approach that it takes, is that I've seen is that you configure your service collection. So the concrete object, service collection and then you construct a artifact container and then basically feed it with your service collection and uh object, and then it will internally set up all of the dependencies and from that point on you'll produce an isource provider that is implemented by artifact.

B

A

Yeah so here it talks about asp.net integration, see yeah, so here you can see you have your. This is a container build. This is artifact specific I believe.

E

It's very like I, really like this approach, again, theoretically, of course, uh because even for distributed environment where we need to have a quart, so we need to have a lot of other services being able to Notre invent the wheel but to use on a level of eye service container. That sounds very good to us. Yeah.

A

It's great because it allows you to separate configuration, so you can have a connected string for tenants, for example, but also have different implementations of services. So maybe one customer should use more fancy, email service, let's say and another one, a more crappy one because they pay less.

A

You could configure it at that level, so that's very flexible, but what you, but it will always require you to have separate connection strings, even though they could point to the same database. But then you need to make sure that you don't conflict between one tenant and another in the same database in terms of the tables being created, or maybe you can create a a schema so that it's scoped per tenant per schema.

A

So that's the that's what it doesn't give you, okay, cool, all right, so uh so yeah so, ideally, uh I I will be able to produce the reference application using artifact uh and I will probably start with that, because it's the it's it's it's very flexible and and very suitable for most multi-tenant scenarios and um after that Look Into, Thin Buckle, uh even if just to get some inspirational in terms of Designing multi-tenancy. At the database table level, so yeah, that's that's what I want to briefly mention: okay,.

B

I'm splitting the uh the information uh to different databases will databases will other be possible in the future Maybe? So maybe workflow definitions could be loaded from different databases.

A

That's that's possible today, so it's completely separated and in fact um we have a implementation for workflow instance store to be stored in leucine or not losing, but elasticsearchs, so that you can have work for definitions could come from your database, uh one one database that you have for even from blob storage or local disk stores or from from Json files as embedded resources, and that that stores mechanism is separate from the workflow instances. That could be completely different and it's even optional. You don't have to store workflow instances.

A

If you don't even care about that. Okay.

C

Yeah I can ask about the current status of the um workflow context. Providers.

A

Yes, let's see so there is a multiple workflow context.

A

It is I, started the implementation and had some success with it, but it's not fully tested out, so it could be lacking some features so just to be aware of it, but there's something there: there's no documentation yet I need to provide a proper, at least a proper sample application to showcase the usage. So it's it's there, but it's not finished yet.

C

Okay: okay, thanks.

A

But worse, it's it's also been a while so I I can't say how far off it would be from completion and what I can say. It is a little bit different from LSAT, 2 and also 2u, which you would do there is configure your workflow definition with a context type.

A

So that would be a net type name, that you would provide there fully assembly qualified and then based on that type, it would find a Associated workflow context provider that you implement yourself and then it allows you to establish some some ambient context, which is like a global variable to your workflow, but with information loaded automatically by your provider.

A

Every time the workflow executes and that's very convenient in case, you work with certain domain entities that you need to make sure they are always loaded, firstly from your own database, instead of having to load that entity or refresh the entity every step of the way in your workflow explicitly. This would be done automatically for you in the background. That's the way it works in LSAT, 2. and also three that will be the same, but it had will have the option to provide multiple contexts and also it wouldn't be Associated on the context type.

A

Let's say you have a customer type or a document type. It would be associated associated directly to your provider. So you configure workflow implementation with a set of workflow context providers that then they can then all provide the context or parts of the context to your workflow. So it's more flexible this way, that's so nice, yeah, huh I, uh yeah I, like that too, it's it was one of my uh uh Heath pets with with Elsa too. Let's see just looking at the extensions here save context.

A

So this is this is what you would Implement and I work for a context provider you have loads and it's safe and I I did have a sample at some point. I think I deleted it because it was completely but uh yeah. Thank you for reminding me. It's similar.

C

To confi the configuration of the provider to elsa2 or.

A

It's similar the way it would work. Maybe I can figure it out on the Fly. Let's see, let's say we have a project called hello. No, not this one. So here's an Excel workflow, let's see! Oh, wait! Sorry! This is just a static class I'm, looking I'm looking for an axle implementation of workflow base, so that's I workflow. Let's see what workflow Builder has actually I'm looking at it from the wrong perspective, because workflow context is an optional opt-in module. So the core doesn't know anything about context.

A

It's an unobtrusive implementation of sorts, so I should look at it from this way. So, let's see we have workflow execution, Builder extensions Okay. So that's that's pretty cool! Actually, here we have a middleware, that's being installed, not an a spin at core middleware, but a workflow, a workflow execution pipeline, middleware components and what that does it allows you to anytime a workflow is to be executed. It goes through a pipeline of steps and by default it will just be one or maybe two steps, the one being just responsible for running the workflow.

A

But you can configure this pipeline to do stuff before the workflow executes and, of course, after it executed and that's features what workflow context takes advantage of so it installs a middleware code, workflow context, workflow execution middleware, so it installs workflow context and it takes advantage of the workflow objects, custom properties, property bag and it's going to try and see if it yeah. If it contains any workflow context, then if not, it just continues to the next middleware, which is probably just executing the workflow.

A

Otherwise, it's gonna for each workflow context, it's gonna invoke its provider and the provider would be responsible, would be implemented by you and it would be responsible for retrieving the data or constructing some context, using whatever data is, it makes sense in your scenario, but for each context it's gonna get the provider, invoke your load, async method and then install it into the workflow context like this. So internally that looks like okay, so it's gonna set it into a transient property of your workflow execution context.

A

So basically, this model module provides extensions to conveniently and easily install workflow context and and read reads your context. Back but I will provide a proper example so that you can see with with methods to actually use nice mm-hmm context.

A

Oh, and we even have a I was just explained about workflow execution, middleware, there's also an activity, execution middleware, which looks like it's doing the same thing. So it's gonna, just it's workflow context. It's going to be loaded and then be provided, but I I need to remind myself why it's on both levels, it might seem a little bit extraneous to have two of them, but maybe there was a good reason: I need to try and remind myself, but I will come when I work on the example all right so that that went by fast.

A

Thank you for the questions. If you have any question before we leave, let me know at.

E

Least, still planning have three zero Round Table, sometimes.

A

Let's look at the calendar, so it's 16 March, so April at this point seems a little bit early given what's what's left and that's what I mentioned at the start of the meeting, although that should be solved sometime next week at the latest. But let me remind myself quickly on the outstanding issues: I mean it's possible it, but it's time goes a little bit faster than I I expected.

A

So here we have a little task overview. I want to finish a workflow context, because that's people use it in Alpha 2. So we do need it before the well I wanted to be there for the release candidate. At least this is a refactoring job that I need to do and very important I need a sufficient amount of documentation, so people can actually try.

A

Otherwise, the release candidate would be hard to to use for people without guidance on the most common scenarios to use it for so I think that's the the bark of the work honestly I, don't think beginning of April, but end of April is realistic and if not, it should be made, but I don't expect it to be much later. Even more so because I know there's a few customers that are actively working on implementing elsa3, so they are reporting issues, we're fixing them very fast.

A

So so things are moving on very quickly this this week, so it is possible, but we will see I.

E

Appreciate your transparency, thank you. Yeah.

A

No worries and uh just to add to that, if you have anything that you're missing, that you think is crucial to have as part of your RC release, please feel free to. Let me know, because that also helps Drive priorities, because, as you can imagine, also looking at this list, there's there's a whole bunch of features and functionalities that are today in Elsa too, but not yet in Elsa 3 and it's I.

E

Would say probably user task is the only one that kind of jumps into my eyes everything else. We can wait for later.

A

Yeah makes sense, I was I, wasn't sure if to put it in the release, candidate or not, but maybe I will move this one back for ours, but thank you for that. The inputs all right well great. Thank you guys for attending and uh have a good one and we'll see each other next week.

B

Thank you for the presentation.

A

My pleasure, thank you. Thank you, see.

B

C

B

C

C