Re: [QUESTION] Purchase UML tool using Google security funding

[Felix Schumacher]

Am 29.08.24 um 18:29 schrieb Mark Thomas:

On 29/08/2024 15:34, Felix Schumacher wrote:

<snip/>

While I don't object to buying a license, I would love to know, which diagram you looked at and what exactly did not work out. (the activation stuff in mermaid is brittle, but I think I managed to get them all right)

I couldn't find a way to get the gap in the activation of Catalina between the call to setParentClassLoader() and start(). I see how you fixed that. Nice. There are a couple of places where the message arrows don't quite meet up with the activation bar correctly and the await note isn't quite in the right place.

I only see one gap at the message await() from Catalina to itself.

The note is a bit better placed, when we use "note right" (or even "note left") instead of "note right of Catalina". To get even more of the original look, we can use UML style, by adding "skinparam style strictuml" and to mimic the original drawing even more, we can switch off the actors at the bottom, by specifying "hide footbox". (full code at the end)

and for mermaidjs I got

I found that very hard to read but I suspect that is a fairly easy fix.

My biggest complaint with mermaidjs was that the text on messages to self is centered rather than to the right. That is probably fixable too.

---
config:
  messageAlign: left
  mirrorActors: false
---

The three dashes are important and part of the header block.

We can configure some stuff in a YAML like begin block by adding (at the top of the file) mirrorActors switch off the actors at the bottom, to mimic the original diagram (not that I like it that much).

There is a missing activation bar for the digester but that might be due to the issues you mentioned.

Yes, there was a missing "deactivate Digester" (again, full code at the end)

The label is in the right place for await() which is good.

To summarize my findings. plantuml seemed to be more predictable and feature-rich (for sequence diagrams) than mermaidjs. But I didn't see any showstoppers with both of them.

Another alternative to use would be umlet (https://www.umlet.com/), which I used way back, but haven't looked at lately.

I'll take a look.

I hope you didn't mind the inline code and thus this long message.

Not at all. This is all really useful.

I do really like the idea of the source being human readable but I think I am still leaning towards Visual Paradigm because it doesn't have any of the niggles in the output and generally, we have a lot more control over the final layout. Another factor is time. While Visual Paradigm also has its quirks, I've found I have spent far less time cajoling the tool into providing the output I want.

@startuml

hide footbox
skinparam style strictuml

activate Bootstrap
Bootstrap -> Bootstrap: initClassLoaders()

Bootstrap -->> Catalina ** : newInstance()

Bootstrap -> Catalina: setParentClassLoader()
activate Catalina
deactivate Catalina

Bootstrap -> Catalina: start()
activate Catalina
Catalina -> Catalina: load()
activate Catalina
Catalina -> Catalina: initNaming()
Catalina -> Catalina: parseServerXml()
activate Catalina
Catalina -->> Digester ** :
note right of Digester
  The digester creates all
  of the objects defined
  //server.xml// but only the
  Server is shown here for
  simplicity
end note

Catalina -> Digester: parse()
activate Digester
||||
Digester -->> Server ** :

deactivate Digester
deactivate Catalina
deactivate Catalina

Catalina -> Catalina: initStream()
Catalina -> Server: init()
activate Server
deactivate Server

Catalina -> Server: start()
activate Server

Catalina -> Catalina: await()
activate Catalina
note right
  This is where Tomcat spends
  time serving requests
end note

deactivate Catalina

Catalina -> Catalina: stop()
activate Catalina
Catalina -> Server: stop()
Catalina -> Server: destroy()
Catalina <<-- Server:
deactivate Server
deactivate Catalina

Bootstrap <<-- Catalina:
deactivate Catalina
deactivate Bootstrap
@enduml

and for mermaidjs it is

---
title: Some sequenceDiagram
config:
  messageAlign: left
  mirrorActors: false
---
sequenceDiagram
    activate Bootstrap
    Bootstrap->>Bootstrap: initClassLoader()

    create participant Catalina
    Bootstrap--)Catalina: newInstance()
    Bootstrap->>+Catalina: setParentClassLoader()
    deactivate Catalina

    Bootstrap->>+Catalina: start()
    Catalina->>+Catalina: load()
    Catalina->>Catalina: initNaming()
    Catalina->>+Catalina: parseServerXml()
    create participant Digester
    Catalina--)Digester:<br>

    Catalina->>+Digester: parse()
    create participant Server
    Digester--)Server: <br>
    deactivate Digester
    deactivate Catalina
    deactivate Catalina
    Catalina->>Catalina: initStream()
    Catalina->>+Server: init()
    deactivate Server
    Catalina->>+Server: start()
    Catalina->>+Catalina: await()

    deactivate Catalina
    Catalina->>+Catalina: stop()
    Catalina->>Server: stop()
    Catalina->>Server: destroy()
    deactivate Catalina
    Server--)Catalina: <br>
    deactivate Server
    Catalina--)Bootstrap: <br>
    deactivate Catalina
    deactivate Bootstrap

Felix

Yes, I can understand that. It seems, that we have different bars on what is acceptable visually :) As promised above in the text, the code for plantuml with the corrections is:     note right of Digester: The digester creates all<br>of the objects defined<br>in server.xml but only the<br>server is shown here for<br>simplicity     note right of Catalina: This is where Tomcat spends<br>time serving requests

Mark

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org

OpenPGP_0xEA6C3728EA91C4AF.asc
Description: OpenPGP public key

OpenPGP_signature.asc
Description: OpenPGP digital signature