You should note the security model here sucks. You embed your diagram with the same unique Id that is used to edit it. However, editing isn't authenticated. So anyone who can see your diagram can figure out the edit page and alter it.
artist-mode to create maybe, but then how do you edit? Concepts such as "make this box bigger" or even "move this box over here" get kind of difficult when it's all just pixels or all just ascii chars.
Graphviz is pretty cool though. But saying that I tend to have to spend too much time tweaking it just so that all the labels don't overlap.
That is quite cool, but I found I had to give a bit too much though to plural/singular in order to get a diagram that made sense. If it could understand that "templates" and "template" are the same thing it would make a big impact on usability.
Yup, I think the world of visual programming is approaching... I know we've tried and failed before, but all the new people trying again don't know that. sssh
This is actually closer to the reverse of visual programming -- we create diagrams by providing their textual representation. I really like the idea. Every time I need to deal with some UML or diagram application I have an impression that generating diagrams from skeleton source code would be easier than the other way round. Tools like the built-in Visual Studio modeler, which updates the diagrams in real time, are also a step in good direction.
Very cool, one problem I found is that you can't have 2 word objects.
Example: Router connects to Web Server
It uses Router and Server as the objects with "connects to Web" as the connection. Putting the 2 word object (Web Server) in quotes confused it:
Example: Router connects to "Web Server"
It should be possible to have a 2 word object with some kind of escape character to create an object with more than one word without having it think you are supplying it with part of a connection.
While very neat, I'm not sure entering these with text is the easiest approach. As soon as the automation fails to pick the right flow, you're basically stuck. A combination of text input and graphical manipulation might work better.
Lucidchart is awesome. Google would do well to make an acquisition there - I really feel the pain of not having any diagramming integrated with google docs.
That's an excellent program. I downloaded it per your suggestion, but I really wanted a web based version. Thank you though, I will find many uses for this program as well.
Dot itself is also a command line tool, so it's easy to make documents and presentations with TeX that have text-format diagrams in the dot language. You can also do things like generate call graphs for programs, or parse log files into sequence diagrams by parsing input somehow and generating dot files.
i have a script that reads django models.py files and constructs OmniGraffle object models. hmm...could be neat to go the other way since i usually make the diagrams first...i wrote the script because once the main idea is figured out then i'll make lots of iterative changes in the code and not want to update the diagram so carefully..
i'm still working out the details. but i'll post here as soon as that's all good :-)
not sure if anyone else is interested, i guess.
Would be great if you could export as Graphviz dot files (which it looks like it's using behind the scenes). That way it would be useful sketching out the basics but you wouldn't have to throw all that away once you needed more features.
It doesn't understand what you are typing. It probably just splits on words and tries to match what it has seen before. I'll bet if you abuse it you can break it.
Edit:
It seems to do a split on spaces and then takes the first word and the last word as the nodes and labels the edge with what is in the middle.