Re: V2 UML diagrams

From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-01-13 08:43:16

David Abrahams wrote:
> Vladimir Prus <ghost_at_[hidden]> writes:
> 
> 
>>David Abrahams wrote:
>>
>>>Vladimir Prus <ghost_at_[hidden]> writes:
>>>
>>>
>>>
>>>>I've just uploaded a couple of UML diagrams to
>>>>
>>>> http://boost.sourceforge.net/boost-build2/diagrams/
>>>>
>>>>The first shows main classes, 
>>>
>>>
>>>I understand that, and as you can see I've transcribed it into the
>>>relevant .jam files.
>>
>>Thanks! I'm yet not on speaking terms with emacs picture-mode,
>>so hardly could do that. BTW, they look very nice!
> 
> 
> I've forgotten more than I once knew, so it was harder than it
> should've been ;-)
> 
> Basically, you just do M-x "picture mode" C-h C-b and see what you've
> got to work with. I have a few additional rules I wrote for rectangle
> manipulation in my .emacs:

I've tried that. The biggest problem is that when I have to add a method,
this is very hard.

>>>>and the second is a sequence chart showing how a project with
>>>>simple main target, created by the "make" rule, might be generated.
>>>
>>>
>>>I don't understand that one. What are v1 and v2?
>>
>>First, let's assume that jamfile contains:
>>
>> make foo : bar : make-foo-from-bar ;
>>
>>v1 and v2 are two instances of file-target class, v1 corresponds to
>>the "foo" file, and v2 --- to the "bar" file.
> 
> 
> More-descriptive names, e.g. "source" and "target", would've helped
> here.

OK, noted.

>>The anonymous instance of make-target creates both v1 and v2, as
>>well as action "a". (The "create" message means that the recipient
>>of the message is created). Later, the action for generation of v1
>>is set, and v1 is returned as result of call to generate.
> 
> 
> What's the blob on the left which is the source of the first generate
> message?

I would have asked the scanner if it could talk. Just ignore it.

>>After that, v1 is actualized, as shown by the following messages.
> 
> 
> What is the self-referential loop on V1?

It means that after getting target path from action, v1 sets it's own 
location. There's no "set-location" method, but I thought the actual
details (call to "LOCATE", etc) are not that relevant.

>>>>2. Is the quality OK? (There're scanned from hand-drawings).
>>>
>>>
>>>IMO the scan quality is OK, but the hand-drawings need to be cleaner
>>>if they're going to communicate well, which is why I converted them to
>>>ASCII.
>>
>>What should be improved, precisely. Looking at them now, I see that
>>text is probably less readable than it should. The arrows looks
>>good, however :-)
> 
> 
> It's usable, just messy. Messiness is distracting and interferes with
> understanding. Laying things out without overlapping elements
> (e.g. lines), where possible, also helps.

Hmm... the compromise between nice look (which UML editors give) and
convenient (it's very easy to use pencil) is hard to find. Problems
with adding a new method keeps me somewhat dissatisfied with picture
mode, either.

- Volodya