seanodell / codesqueuml Goto Github PK

The component id map is kind of an internal structure which could be made easier to use/understand, but all I really want is a build script that automatically maps methods called to code files. Either a build script, or a loader which fully parses and resolves all code and generates the map so it can be updated before compilation. Keeping it external enables other neat features such as linking to external resources.

Blank lines should become '|||' PlantUML which adds space between connections

@startuml

Alice -> Bob: message 1
Bob --> Alice: ok
Alice -> Bob: message 2
Bob --> Alice: ok
|||
Alice -> Bob: message 3
Bob --> Alice: ok

@enduml

Right now the component node was really designed with class#method calls in mind. There's no concept of namespaces or packages or anything that groups related classes together. I don't want pseudocode to become as verbose as real-world code, but perhaps if there were the concept of a collection of classes it would help with organization. I think having code files in folders which act as namespaces which can be referred to like Java packages would work.

Keep in mind that there's also a need to allow real package/namespace names to be documented as well. I think it would okay to make those simple attributes of components, but they could be inferred somehow from folders. Perhaps a folder metadata file? I would hate to force package/namespaces to be directly tied to folders. The whole idea of pseudocode documentation is that you're not replicating the actual code base, just highlighting key bits. The coupling between folders and package/namespace documentation should be loose.

Need to represent simple setting of a variable.

Code should look something like:

Code:

Box#Alice()
  Bob()
  set variable = call()

PlantUML:

@startuml

box "Box"
participant Alice
participant Bob
end box

Alice -> Bob
rnote over Bob #FFFACD
//variable// = [[/call call()]]
endrnote
Bob --> Alice

@enduml