Elsa Workflows Community, 21 Jun 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Elsa Workflows Community Meeting 14 ( 2022-06-21)

Description

Meeting 14
Topics

- Elsa 3 designer update

- "dependency properties" on activities in a flowchart activity to control things like "Wait Any" and "Wait All" when merging execution.

A

All right so um yeah, as we just discussed, uh we've been working on the designer mohammed. You mentioned as your data fabric as a tool to look to for the designer for ls3.

B

A

I thought maybe this week we can look at some of the updates I've made on the designer specifically for activity. Embedding hi, christina, hey how's, it going.

B

Yeah it's going good. I was a little bit on http request activity.

A

Oh nice, that's the lsat3 activity.

B

Right, oh, the ui looks nice. Did you change something? uh Oh.

A

We changed many things. I will actually um okay. You will see uh as I go through the update in a second that you were referring to. Is that done? That's what hp requested very cool and.

B

How are you guys doing yeah.

A

Doing pretty well having a lot of fun with the lc3 designer of course. Last week we discussed some of the changes. We specifically talked about: transforming a flowchart style workflow into a nested hierarchy structure and then mohammed came with the tip of azure data fabric, which I would like to show in a in a couple of minutes, and maybe we can talk about that. So, as you can tell already, my mind's been full with the also 3 designer and it's been.

B

A productive week yeah, that's super nice, it looked it looked nice yeah.

A

So does any one of you have any specific topics you would like to discuss during today's meeting.

B

Not from my side, no no.

C

I have no specific things to talk about all.

A

Right: cool yeah, then, let's, let's go over the lc3 updates, so you haven't seen this yet christina. Did you see the new update I found so this? This is just a splash screen and it might be temporary. It's just something nice to look at while we work on on the features.

C

Yeah, I thought briefly the previous time. I think it was already there right, yeah.

A

It's been there for a couple of weeks yeah, but this is not the interesting part, the interesting part yeah. Let me walk you through the design and changes step by step, and so we will start with a new canvas and, as you recall last week, we talked about activities that can contain other activities like the eff activity right.

D

A

Before they, this change the way it would work, you would see an if activity with two outcomes and then you can connect to the next activity, use it by connecting these these lines. So I would look like this. Of course, the effectivity has changed significantly and we'll also talk about ways where you can still have the separate line so where you have a decision activity or an effectivity with two lines. In fact, let me maybe open windows workflow foundation activities.

A

So what I wanted to explain is that, before this design exchange, what you would see is a if activity with two outcomes like then and and else and then you can connect whatever else you want to to execute right, but a defaulting change with lc3 is that if you look at the, if else, activity- or I should say the effectivity as its names, it has a 10 property and an else property. It can contain a single activity in a sequence of activity or an array of activities.

A

So if you wanted to execute multiple steps, you need a container type activity which is a different structure than having a set of connections between activities right. So that's why I was working.

D

A

A way to transpose or transform a network of activities into a container activity which itself would be a nested flowchart a little bit complicated. It worked. I actually have another branch where it does work, but yeah based on some vertical discussions I had with muhammad, I realized. Maybe maybe we should just update the designers and visually, create the workflows as it's also stored structurally. So with.

D

A

A new looking type of activity where activities that have properties of type I activity, they will show those properties visually here, and what this allows you to do is to actually drill into these these ports, as I call them for now, which is a little bit confusing frankly, because these little circles are also called ports, but they are, they represent a different kind of connection between activities.

A

So I need to sort that out before release, but for now it's also called a port and what you can do you can actually drill into it and then design a nested workflow. If you will and then on the top, you will see the path, and this is based on what we have in azure data fabric. I believe it's called. I.

D

Wasn't aware of this tool, it was actually.

A

A very very nice tip from mohammed to look into this because it seems to me it's it's heavily inspired on windows, workflow foundation, but simplified. So let me open the studio. Maybe you'll recognize it if you've used anything similar to workflow foundation, all right, so let me create a new pipeline collapse, this tool panel. So here we have an empty canvas and let me just start with the with one of these. So let's say we have a it's a condition.

A

So, as you can see, this is an activity and it has a true and a false property, and then you can go in and actually create the workflow for for this for the true activities. So what this means that, if the condition evaluates to true, then whatever workflow you create here will be executed, then you can go back and do the same for false. So it's as you can see, it's heavily inspired on that.

A

So what I would like to do now is give a an actual demo, that with a workflow that actually works, so let's go back to the root view, view or the root of the hierarchy, and then here we'll say we'll write something to the console called start and I'm going to remove this one and replace with 848. So we can see what it takes to loop over a set of items and then print them to the console. So here we will configure this one with little array and to access its current item as it iterates.

A

You need to capture the output, so the current value will be stored in whatever output target. We specify here right now, there's only one such type, which is workflow variables, but we might have future possibilities to capture the output or maybe even directly connected to another activity or have another activity receive the output as it executes. I don't know yet, but this is for now it's just variable. So let's create a variable called current python of type string. Let's connect these two and then in the for each body.

A

We can specify what the logic is supposed to be for each iteration. So the most simple thing, of course, to do is just write. The current item to be like this. So it's the current item. So this is the variable I declared on the workflow. So that's this guy and of course, because this is javascript, we can use string interpolation. So we could say something like this just to make it a little bit more interesting.

A

Of course you could do other things right now, it's a single activity, but you could also do things like set a delay, so it waits for let's say a second for each iteration. Alright. So as muhammad also discovered there's a bug- and in here you see it on display they're supposed to be a navigator, but there's a little bug that doesn't update it correctly.

A

When you add new levels, audio activities, so I'm just going to publish this for now refresh and now we see the the current path and it and you can go multiple levels deep. So, for example, I could add an if statement here and let's say we want to reconnect this to the.

A

If and let's say that if the current item is item 2, then we I don't know, maybe we request that the user grants permission to continue as a silly example, let's see what that looks like so, I would say if parent item equals item two. So if this evaluates to true, then we want to do something in the in the then outcome.

A

So here we would say we would instruct the user, basically by saying please approve and we'll use uh just because we don't have too many activities as of yet, but we can just say, let's request that the user approves by sending some http request to some endpoint, so we can say the proof, for example, and when it's approved we say thanks whatever right so.

D

A

This is not updating, I'm gonna refresh just so so you get a picture for what it's gonna look like visually, so this is a way to navigate back to the upper layers of the of the workflow. So here we're in the deepest layer. We can go back one level up here. We see the body of the four each. So if item equals, current item equals item two, we execute this nest activity and.

B

A

Way right now, you see flow chart because the type of activity is actually a flowchart. This thing is a flowchart. The canvas basically represents the flowchart, it's an activity and granted. This is not very descriptive. This doesn't mean anything to a user that looks at this, so we need a way to allow the user to provide some description or display name also as a future update. What I would like for the user to be able to do is instead of always defaulting to creating a workflow.

A

When you hit the edit pencil icon, you should be able to add an arbitrary activity because, as you recall, the then property is of type I activity, if there's just one activity that you want to execute, maybe it's a simple writeline or an hp request, or even a custom activity that you want to invoke.

A

You should be able to just add or assign that activity to this property, but this requires some more enhancements to the designer so that you can actually maybe drag and drop something into it or where you can click on the plus icon and it shows a drop down list with available activities like a drop down, picker or even drag and drop something into it. But those details we need to figure out still at the very least structurally that now becomes possible.

A

But for the time being it will be a flowchart, and I want to be able to highlight the flowchart so that you can edit its properties on the right hand, side and give it a custom name. That already should help a lot. But anyway, let's uh let's go back to the workflow at hand. So this is now done and then here we write the current item here we have a delay. So, let's see if this, what this does for now and let's also mark the end of the workflow all.

D

A

So let's publish this and take the definition id and then we can invoke this workflow from postman. So first let me clean the console out or the console contents, so we can see what happens so now I executed the workflow. We see that start was printed bunch of current items and then done so. This doesn't do exactly what I intended for for it to do so it could be a bug, but let's see if we can figure this out. First of all, the current item should actually display the current item right.

A

So, let's see what's going on there so for each activity I forgot to assign the its current value to the current item variable so that variable remained empty. So that's one issue: let's see what happens if we try again, but every second for the delay. We see the current item, but still it's not displaying the current text this I actually know this is a bug which has to do with the delay activity and the fact that the variable is not being persisted by default, so by default.

A

None of the stuff is persisted into the database, and this is to prevent the issues that we have with elsa too, when you, for example, use objects that aren't serializable, and then you have to do all sorts of stuff to prevent it from being serialized and in elsewhere it's the other way around. If you want something to persist, you need to be explicit about that. So here, let's use memory for now to to store the variable.

A

You could also store it as part of the workflow instance, which gets persisted every time the delay activity executes or any other suspending or blocking activity. So let's try this again all.

B

A

This looks better and now we also see because item two is about to be displayed, but we we have this if activity, which included a blocking 50, which is the hp endpoint. So let's, let's see if we can resume this and I think the url was approved nope, no, that's not it or it was this thing, but there's some bug somewhere, where it doesn't currently doesn't store this correctly. So, let's let me see if I can make it work, so I'm gonna change this to something else.

A

Of course, I also have to restart the workflow, because the workflow instance doesn't take into effect the any change I make here so delete this guy here and try again nope doesn't want to work. I need to look into this issue. Sorry about that. Of course, you'll you'll believe me when I tell you that it worked before the demo.

C

Yeah, of course, no, I believe, quick question, though you have an http endpoint. Is that just listening to any request that comes in or is it tied to the workflow instance.

A

Oh, this is bound to the workflow instance, because it's not used as a starting activity, so it would only work when a workflow instance is actually blocked on this this activity and to be more specific, if we look at the workflow bookmarks, we see three of them because I started three instances of it. So if I was able to successfully resume the workflow, is this request made to approve one or a proof would execute all three bookmarks?

A

Well, not all three, because one of them is listening for or two of them are listening to a proof one, and this one is listening just for the approved path. So if this one succeeded, both of them would execute. Of course, okay in in real world world scenarios, probably what you want is to invoke one wordplaces or trigger one of them, and for that you need to rely on correlation, but we did not specify any correlation id and when you do correlate your workload instance with some identifier.

A

You also need to include that identifier, when you make an htm request, typically done through a query, string, parameter or an hp header. This is the way it works in ls2 and similarly, it will work in the same way in elsa 3., but.

C

In lsat 2, you had a receive.

A

Signal basically, and that.

C

Had the body of a workflow instance and a correlation id, and one more as I recall, but is that the workflow instance id is not longer supported that it's only being correlation id.

A

um Well, it's just not implemented yet there's no signaling. Yet you know.

D

A

Three, but probably it will be, it will work in a similar way where, if you have, you always have to work, for instance, to you at your disposal to invoke a ver. A specific instance and also correlation is supported, and these are just two bits of information you need, or you need one of them in order to be able to target a specific workflow instance. But the point of the engine is: is that you, you should be able to extend on it? You can, of course, also create custom tables and make custom associations.

A

Ultimately, we have a set of low-level apis to invoke workflows directly, including workflow instances, and then a set of higher level services that implement more commonly used scenarios, but you're not limited to those. What we do.

D

A

Which we don't have in outside too not completely anyway, is the event they uh didn't add it here, but there's um actually let me let me see if I can add it.

D

I think it's on that. Oh.

A

Yeah you're right, thank you yeah, so this one is what I was looking for. So this is an event and with the event, it allows your application to just publish an event, and then it will trigger a workflow, a new work, for instance, or resume an existing word, for instance, and it has just a simple thing called an event name. It could be custom event.

A

So let's switch this out so when this executes it will block workflow execution until it receives appreciation, event called custom event, and let's see if this works so first I'm gonna start a new instance all right. So now we have an instance blocked on the event name called custom event and to resume this work. For instance, there is an api here called trigger where you specify the event name. So in this case custom event, and then this is just an action on this resource called trigger. So this is part of the route body.

A

I think at some point will allow you to provide input, custom input that your workflow can then access and, let's see what happens. If I do, this, okay looks like it did something all right and then so there seems to be a bug in the xp endpoint. But this activity allowed me to resume the workflow instance and, as you can see, then it goes to item 2 and 3 and ultimately done awesome.

C

One quick question about the correlation id: I noticed that in elsa 2 there was a constraint that it should be a unique still the case.

A

So the uniqueness is not imposed at this point, and maybe I don't know actually if it makes sense for the correlation id to be unique per workflow instance or if it should be unique per workflow definition. So maybe you should be able to have multiple instances of a given workflow definition and then within that set only one of them can have the same correlation id. It depends on your scenario.

A

Some users want it to be unique and others have scenarios where they want to use the same correlation d across multiple instances of the same word, for definition as well as others. I think it should be. It should not be restricted. I think if you want to ensure that only one instance of a given correlation that he executes at any given time of a given word for definition, we should try and find a different way to uh to achieve that. But at this moment it's unrestricted and I- and it probably will be that way.

A

But yeah again, we need to come up with a mechanism to also allow those other users to actually make sure that one instance executes if it's correlated.

C

Okay, yeah. I was looking into that because we are using that basically for a messaging process where there is a conversation id basically and I wanted to correlate all the mess or all the processes that handled messages within that conversation. Just by the conversation id and I set down the correlation id, but since there were multiple messages, I got tied up because id was already mentioned, and I wanted to use that just for analysis later on. So I can easily find which processes were for the same conversation.

C

But maybe I shouldn't use the correlation id for that, but something else yeah.

A

That's a good point because you should be able to associate or annotate your workflow definition, but also workflow in instances with custom data and not be limited to just using a correlation id but have different tags so a little bit similar to labels for workflow definitions. So here we can say not this one, but this one. You should be able to say and don't mind this. This ui bug here, we'll fix that. Let's say we have label one.

D

A

Label two yeah there's some ui work to be done here to be fixed on label two: okay: let's let's go with this so this these could be custom values and and similar, so this is associated with the workflow definition. But similarly, you can imagine having something like this for workflow instances. So you can associate this conversation id using a label called conversation id for example, and then you can also query by it, which also means you should be able to resume workplaces by specific conversation. If that would make sense. In your use case scenario,.

B

A

Collect all processes or workflow instances for a single conversation id and correlate it in that way. So that's that would be a nice feature to add. Actually.

C

Yeah, I think I like and the concept of labels, and I would use that, for instance, to do that conversation id. I don't see the necessity of having the ability to based on the label resume, because I think the workflow instance id for that.

B

C

And the correlation id could be the one to use for that, and it provides enough functionality based on those two and not having the labels as well. Yeah makes sense it's great if we could do this. That makes it so much easier for your process. Analysis later on yeah.

A

C

Yeah, it's looking really impressive, great progress since uh last time, yeah.

A

Thank you, yeah, I'm pretty um happy with this. This was one of the things that I was actually waiting a little bit, whether I should delve into it. It seemed very complicated to be able to. You know, create these hierarchies, but in this way it already works. Well, it's not perfect yet, but it's it's getting there and, most importantly, it's it's a very simple structure to uh to set up compared to what I was trying before transposing networks into embedded activities which is prone to error yeah.

A

I took a lot of inspiration from this, as I mentioned, and what I noticed that this engine does. If you look at the structure, so what would they allow you to do? Is they allow you to add multiple activities on a given slot? So here we have the true slot, and if you edit this, you can, let's say you can have this one and this one. So these are two activities. We can of course connect them and if we go back up, the level displays the number of activities and the way they do.

A

It is as follows. Basically, any container or any slot can have multiple activities, and this is in contrast to elsa 3, which only allows you to add to one.

B

C

B

A

Explain in a minute why this is the case, but it's interesting to see what they are doing and also the way they relate or associate one activity to the other one. They don't use connections. What they use instead is. It depends on which makes sense. So here you can just specify, for example, what I did here: the azure function or actually the custom. One activity depends on the azure function. In this way they can visualize the dependencies.

A

The connections it's very convenient, but I think it's also potentially limiting, because what else iii would allow us to do is to use different kinds of flow charts or different kinds of diagrams. I should say, for example, we could now have not just flow chart style activities, but it could also be sequential activities where it's not a flowchart, but just a series of steps. It could be a state machine or even a bpmn custom activity or anything else we come up with.

A

So it's I think it's more flexible and it's not limiting right, because you can always add a container as your activity, property.

C

I think everything should be possible with depends on actually, because if you look at any of the charts, it's always a 2b and a connection between those, and if you always look at that from the point or where it goes to, when it says, depends on you can basically create everything relating it back. That's.

A

Exactly the case, but I was pointing out that they are doing something else, that's interesting, and that is it's, for example, the true property. Here, let me pull up this guy here, so they call it the if through activities which is a property of the, if condition type, it's an array and that's another difference and it's very convenient, but I think having this as an array is also potentially limiting. Maybe not I'm not sure, but it's interesting to see how they are doing it.

A

And it's it's also in contrast to what windows workflow foundation does so maybe it's a learning that they did. Maybe they realized. Let's keep it simple, and it's good enough for this pipeline. Editor yeah.

C

I'm thinking about that two activities and for me it's okay, it's in static that there is- and there are two activities being executed in that true. What I would find interesting is that it would have, for instance, two starting activities that would be more interesting to know than the amount of the whole activity flow in there, and that's one of the questions that I had in elsa 3.

C

If you don't have any depends on basically as it is used here, is it always a starting activity, then, if nothing else connects to that, it's considered as starting. That's.

A

Right yeah, so if we look at here, let's say: there's no connection. So now we have two starting ones and it's you don't know up front which one it's going to be used. Unless you look in the data structure, because the ng will now pick the first one, it finds only.

C

The first one or.

A

Execute them both good question. I think it will take the first one, but that's just an arbitrary choice. Really it could could just as well have been well, let's execute all starting activities. Maybe that makes.

C

Sense, but I like that, the concept in the elsa 2 flows, where you have multiple starts that you can use, and I use that regularly that you can say well. I want to start on a signal, but also on a con job. For instance, yeah.

A

So that's still the case in ls3, so for that you need to use triggers. So let's say we have two triggers or maybe use the same one. Let's say we have event one and event two and you can both mark them as startable. This is important by the way yeah. So now we can execute this workflow by either firing this one or this event, but you do need to mark them as startable, because that's how the engine indexes the workflow by looking at you know what triggers to create based on these settings.

A

So you have a bit more control. The downside is you actually need to be aware of this, so it can be confusing, but this would just work and you didn't market the starter ball then, and it won't work could be confusing. The reason I did it is because the imagine you have this workflow here. Let me show you, but.

C

We live without that startable because if there's no input on it or any prior one then assume it is startable because it doesn't make sense to have it like that and not startable. If there isn't an prior activity.

A

Exactly if you look at this workflow and yeah, let's forget about these these for a second. If you look at this, then you would be writing thinking. Well, it's only these two so and these are triggers. So what's the problem right, why not just take them? The thing is: is that this these two activities are part of a collection of the outer flow chart, so the root activity of the workflow, so workflow is a model that looks like this, so this is a class or actually this is yeah.

A

This is a class and it has a root. So this is the start activity basically and all workflows have a single activity with with its star, so it the workflow itself is also an activity of type composite which has a root. So it inherits from this in here and.

D

A

This workflow is a flowchart activity which is a separate. Well, it's just another kind of activity. So, of course, we could easily program it in such a way where we recognize that okay, this workflow starts with a flowchart. Let's look at its starting activities and index. Those would work, but it also seems maybe a little bit arbitrary to do it just for this specific type, because we also have sequences. So we would have to do the same for a sequence. Look at its first activity, see if it's a trigger and.

D

B

A

May very well do this, but in order to not have to do that, I'm experimenting with the startable activity and uh the downside is what I just mentioned, but the advantage, I think, is it's more explicit. So maybe you do want two triggers, but maybe you want to be able to disable one of them, and this makes it explicit action anyway. That's the thinking we may reconsider. I don't know what what's the best way to deal with this and it's it seems to me it's also a bit subjective. What's better and what's not so good.

C

Yeah, I agree. I will give it a thought, because I, like the concept of not having to force users, to use something using startable as an enabler. Disabler feels also a bit like misusing the possibilities rather having than an activity to be disabled in the workflow, to explicitly name it like that yeah we'll give it a thoughts about the pros and cons and how it is for the user to do that.

A

Yeah, please do. That would be great a big disadvantage. I think personally, is you also have to do this when you actually code a workflow, so, for example, not this one but no? Well. Let me just invent some examples, so so here, for example, we have a for simple for each workflow, it has a set of items and then the root of this workflow is a sequence, so the sequence has an activities properties.

A

This is the contents of this sequence activity. Now, let's say we want to have a trigger- let's say an http endpoint, to turn this workflow into an api endpoint, for example, with the current the way the engine currently works. You actually need to say that this is that this can start the workflow. So it's you know it's not very intuitive or convenient, because you would think that maybe you know this should just just work, but now it only works.

A

If you do this- and if you forget about this, you have some time that you're gonna spend to figure out why it's not working. So it's another disadvantage, but.

C

Was it working then, with the right line? Oh.

A

Yeah, so this example explicitly invokes this workflow, so here I have code. Let's see this is the fourth: it's that's this one. So this is work for.

D

A

So here I'm just saying: run workflow 11 or whatever I configure here so this. Of course this will always work, because the way this works is just execute. The workflow and the workflow itself has a root activity, so the root activity will be scheduled for execution and that's a sequence and the sequence will then schedule its child activities for execution. That will always work, but the thing is with triggers right, like hp.

A

Endpoints signal receive event that kind of thing for that the engine or the system needs to index its workflow definition and extract, or you know, interpret the triggers and and request some bookmarking information that gets stored in the database in the triggers table. So this would get populated. So that's the that's. The deal yeah.

C

Then that makes more sense. Sorry.

D

Yeah, why don't we have it in a tool as before that it can start the workflow, because it's a trigger sorry.

A

I'm not sure I get the question, can you can you repeat.

D

For example, the http endpoint can't we have the can start workflow as true by default, and then, if I want to disable it yeah.

A

That's another option: maybe it's a good one, and this would then mean for all triggers by default. They are enabled, and if you don't want that, you have to explicitly disable but.

B

Might make sense.

A

Because in most cases you want it to be startable, the tricky part, of course, is if you use a trigger as a blocking activity. That's which means it's not a starting activity, but it's somewhere on the somewhere. Here, let's say: let's say I want to move this in between so now it's no longer well, it will actually be able to start a workflow, because even though it's not a starting activity anymore, it's marked as startable.

A

So if I now trigger event one this workflow will actually start with this activity and go into the right line, and this will be ignored. So that's the other side of it. If you do what you suggest mohammed everything will be startable or can't start. Startable will be true and that's uh has a different issue, because this is not expected. I think so here you would have to uncheck and you would have to do it for every blocking activity that you do not want to start the workflow if you have to explicitly make it unstartable.

A

If that makes sense, oh yeah, it's it's a tricky problem, I think, but maybe we should just accept that there is common logic and common use cases where a workflow starts typically with a container, and we can just look at the first item of sequence activities and the first or the starting notes of a flowchart activity in most use cases that works and for those workflows that don't start with a container, but just let's say a writeline, because that's possible as well. This could be replaced with another write line, but let's say an hp.

A

Endpoint would be quite useless because now it will trigger the workflow, but nothing will happen because our hp endpoint cannot itself contain more activities, but in this case it will be found as a by the indexer as a trigger so uh yeah. Let's give that some thought see if we can come up with a more elegant approach, oh by the way. One other thing I had before a couple of or not more than a couple of changes ago, I think couple of months ago, was to allow the workflow itself to have a set of triggers.

A

I think I removed it yeah, so there was a way to. This is not a good example. By the way. Let me go into this one here, so here we have a workflow builder in this workflow builder would allow you to specify the triggers, so a trigger would be a separate notion. It would just be it would.

A

It would be analogous to azure functions where you create a function, and you can associate it with triggers, so it the trigger itself would not be an activity per se could be an activity, but it could also just be a mechanism to associate with your workflow so that when that trigger is invoked, it will evoke the associated workflow. That was another that's another way.

A

It could be done, so imagine we would say trigger, and this could be timer, let's say so now it would be stored on the workflow level itself and not on the activity structure, and this, of course works.

A

Well, I liked it because it's very explicit to associate to a workflow with triggers, but it's then also inconsistent with the ui, because how would you translate that notion into the ui one approach that I tried was adding triggers to the top, so you can add one or zero more triggers one could be a timer, not it could be hp endpoint, but then there's use cases where you actually want to perform certain logic depending on the type of trigger that was evoked.

A

For example, let's say we have these two triggers and when the hp endpoint is received, maybe I want to maybe set a variable read from the query string that was passed from this trigger, so that's different logic than when compared to the event two, when maybe it receives. Some payload requires a different bit of interpretation to get its input into some variable, so you want to maybe normalize the various types of input into a normalized variable and then continue and that implementing that logic would not make sense when you just specify these triggers.

A

So there's all sorts of considerations to keep in mind.

C

Yeah, I agree the logic and the depending on the trigger is needed. I use that now you could think about and the triggers that you put on the workflow instance as uh elements on the on the grid on the canvas, and it makes it a bit unlogical for how did they end up there, and I think this is more consistent and more intuitive. If you do it and do it like this.

A

Yeah exactly so um yeah, so for now it's it's an explicit action to mark something unstartable. If you want to to do it, but let's uh let's see if we can find a way where it's not necessary to mark something unstartable, but maybe have some rules that identify this based on common patterns that should be extensible. So let's say you also have a new kind of container activity. Then you should be able to have. Maybe some I don't know some trigger provider implementation.

A

Let's say that implements a strategy to to return some information about some trigger. That would, from the user's point of view, would be even better because then they don't need to think about startable. Then this would just work because, from a user's point of view is intuitive, it doesn't require them to do anything special. So.

C

It's worthwhile considering.

A

C

Maybe a question bit of topic, but I was explaining elsa to flow and workflows by code to one colic today and was a bit struggling with the then and when and nested uh sequences in that and.

B

I was wondering, did.

C

Is there ever considered to see that the designer you can draw them and you can generate, for instance, a sharp code out of that, how it would be coded yeah.

A

So this question has been asked on github, at least once I believe, but yeah that question did come up and it's I think it would be great to have to have a designer can spit out c sharp code. That then you know, allows you to have a workflow in c sharp, but created use the designer. It's.

C

A

Not out of the question okay, I.

C

Think it it will reduce the learning curve for the coding part dramatically, and the adaptation then of the workflows, is a bit higher. I guess.

A

Yeah, and especially if you are talking about the api for ls2, it gets quite complicated and hard to work with, especially with the flowchart model, where the fluent api attempts to connect activities as you create them to others to the next one, but with elsa three. I don't know if you've taken a look at the also three structure, but.

C

I think it's way more clear and it's more intuitive, how to read it and also the indentations with it makes it uh way easier to read. I.

A

Think so too, yeah, because uh because of course it's it's nested right. So, as you say, there's indentation there's no manual connecting between activities. It's just you! You just declare what you want to happen here and then a sequence is a sequence of steps and then, if has well this, if has two properties here, so it should already be a little bit better. But I agree even code. It's. It is hard to read so having design and support to create that code. That will be helpful and especially interesting with code generators.

A

We can have workflows stored as a json file.

D

A

All you would have to do as a user is store the json file somewhere and then the code generator would generate for you one project to do.

A

I was hoping you would say that so I'm actually kind of looking forward to some cool demo, uh maybe in the near future. Oh.

C

Even in the near future, okay challenge accepted awesome.

A

That's great, and, and since we're on the topic of you know easily declaring workflows. Let me just remind you there's a lot of it here: there's a dsl module which is a different take on easily creating long-running workflows or any workflow. For that matter, and here's a nice example. I think this one.

D

Yeah this one is nice, so.

A

Of course, there's no syntax coloring yet, and this is irrelevant this is out of date. This is out of date too well, but here this is, I think, it's very readable right, it's just from top to bottom. This is actually not a c-sharp if statement, but the if activity where you configure the then outcome and the else outcome and each statement is actually an activity, and this is this is like an alias of to the right line activity and you can configure the alias you want to use. Of course you don't have to use aliases.

A

You can also just say right line yes and say: text equals whatever like that. I still think nope, not too many people like it. Actually, nobody likes it that I've spoken with, except for one, but I still think this is quite awesome because think of the ramifications, if you can write a simple script that implements a long running process, because this doesn't have to execute from start to end, it could also be like a wait for a customer to prove offer.

A

You know this could be a high level activity and it seems quite powerful to me that you can write scripts. This way now.

C

We have two that aren't enthusiastic about yourself.

A

Yeah since you're busy with it, I think it might be fun for you, too, okay, yeah, yeah, great, all right and here's another example where you just can say delay, and this will actually delay and suspend the workflow or wait some signal and most of this code actually works. The delay I think, works this one doesn't but but yeah. This works all right enough about the dso. Any other questions from anyone. Yeah.

D

I heard one question uh in insta3 designer: I know that you can connect the output of one activity to multiple activities. Oh.

A

Yeah, you mentioned this yeah.

D

In that case, the execution is not very clear to me.

A

Right so you're saying let's say we have a right line: okay, yeah! So so the way it should work is it goes from right lane to this one, then, to this one and then to this one, and then this one right.

D

Yeah- and I also had I joined them all into one activity and that one got executed twice.

A

Yeah yeah exactly so like this, then, let's add another one like this and this this is, and this is going to execute twice so line 1 3 3a the end yeah. So it goes from this to this then, to this one then to this one, but everything is scheduled at the same time, so the flowchart activity schedules all connections. So this is the first connection and it schedules this one as well, and then it's going to execute and because this one finished it's going to look for its connection, which is this one.

A

When this one finishes it's executing this one. So that makes sense and then it gets this one, because it's a stack based scheduling system and then that's what we see exactly here. Then we get two to a and again you're saying. Maybe you don't expect that if you look at this from a user's point of view, maybe you would expect something else to happen. So so maybe you can talk us through what you the way it should execute.

D

The way it looks to me it's like a fork and joint scenario, so I would expect that it will execute both branches.

A

In parallel, so so this one and this one then this one and this one okay.

D

And then it is joining to the last activity, so I would expect it will execute only once so.

A

Okay, so then this will execute and this would execute and then somehow the engines should know that. Well, I just want this one, even though it's yeah, so with the alpha 2, we have a fork activity that does basically what this thing is doing here and then we use a join and the join activity has the logic to well. It receives both inbound executions, but it blocks further execution depending on the setting.

A

So it's either gonna wait for all inbound branches to complete or any one of them before it continues, would be interesting to have this notion implemented without needing an explicit, join, maybe, but I'm struggling a little bit to figure or to come up with a whole way that that makes sense. Maybe somebody else has any any thoughts on that.

C

Yeah you can detect it. It was interesting uh exercise, though, when you first showed it. I was yeah. This is the expected behavior, but I can also go with the fact that it isn't implicit for can join yeah.

A

C

Having that and knowing the elsa 2 implementation, especially for the join, where you can say, weight all or await any, that is not possible with that, and you have to have at least some properties on the join part to be able to to set that to be able to say well wait annual, wait! Oh yeah.

A

If, if you want that behavior or that feature, indeed, you need to, you need to have a way to configure that, and it cannot be setting on this flowchart that contains this diagram itself, because you could have many more branches where sometimes you want to wait all, and sometimes you want to weigh any right like here so here we have another fork and then another mercs or join.

A

So maybe here we want to wait for the first one to contin to complete, and here we want to wait for both of them, because maybe there's two triggers two, eight four or uh some two tasks this. These would be two tasks that execute a background. You want to wait for both of them.

B

A

Here we have two triggers a timeout and an event, a user action, and then you want to wait for one of them and there's, as you say, there's no way to configure that without an.

C

A

Activity like join unless we could.

C

We could do it as if there are multiple inputs on an activity that you get additional properties on that activity. For instance like this, where you could say, oh there are two inputs.

B

C

Do you want you want to execute all yes yeah, you want to wait uh any or and so on. Yeah.

A

Persist so so maybe this would even be these settings would not physically be stored on the activity itself, unless it's some metadata, that's like dependency properties that we have with xaml, where you have dependency properties, so that will be a property exposed by the flowchart, because only those settings are only relevant to the flowcharts model or activity or maybe these settings will be stored on the on the connection, because this connection could receive properties, we could make it selectable and then you can configure stuff on the connection.

C

Yeah, but it doesn't make sense to put it on the connection, because that connection doesn't have any knowledge about the other connection that I accident goes to the same activity. Yeah.

A

Exactly this, this will be identified by the flowchart activity itself, which knows about all connections, and it knows about this node having two inbound connections, so it could inspect these things. But I think your intermission is right that it should be on this activity here, because you don't want to configure it on this one and this one, because it applies to both inbound. You want to control at this level, whether to block wait for all or wait for any, but.

B

A

Would be pretty cool to pull off because then we don't need to use an explicit, join and a fork.

D

Then, when it comes to looping again, what's.

A

D

If you are looping, if you have a loop, then like.

A

C

Here that shouldn't be a problem now yeah.

A

Should work fine right, because this one has two inbound connections. If you configure this one to you say, wait any or wait all doesn't matter. So if it's let's say wait any this one comes in goes into this one, and then straight to this one and.

B

That's the nice.

A

Thing about a flat flow chart model where you don't have artificial nested activities.

D

A

We have an elsa 2 with iterate outcome and then or true and false outcomes of the effectivity which model what we have here explicitly so here they are explicitly nested. You know lower level parts of the art with this flat model.

D

A

Have that issue, so it's okay to just make our with arbitrary connections as you should be able to do in a flowchart hey. I like that idea. Let me make a note of this all right, awesome, anything else, guys. If not, then thank you all for joining and we'll see you next week.

C

Okay, I'll save my questions for the next time.

A

Great thanks bye-bye.

B