| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The shared UI assumed the whole market was local: it showed nothing
until a file was loaded, let the user act as anyone in the list and
offered to create events by hand. None of that holds for a client of a
shared server.
AppState and AppView now take an optional session user. When it is set:
- the tabs show from the start, since other users' events are there
whether or not we loaded a file ourselves;
- the acting user is fixed, and the user list becomes a directory of
the other users with their balance and role;
- the buttons that create events are hidden, and the Users tab gets a
Deposit Funds button with a small dialog;
- loading a file skips the two-second pause Ex2 asked for.
Without a session user everything behaves as before.
Also keep the event's Trade and Close buttons from collapsing to an
ellipsis when the description is long.
|
|
|
Exercise 3 asks for a new desktop client that reuses the Exercise 2
screens, and the course asks that clients depend on the DTO contract,
never on the engine. ui-desktop mixed both: its views and controllers
only talk to the api interfaces, but AppView and AppState each had a
convenience constructor that built a ServiceEngine, which tied every UI
class to the service module.
Move the screens, controllers, components, FXML, skins and fonts into
a new ui-fx module that depends on api alone. Packages are unchanged,
so FXML controller names and resource paths keep working as they are.
ui-desktop keeps only the Exercise 2 app: Launcher, App, which now
builds the ServiceEngine and passes its contexts in, and ServiceEngine
itself. The Exercise 3 client can then reuse ui-fx with remote
contexts and never put the engine on its classpath. The build proves
the split: ui-fx compiles with nothing but api visible.
Tests move with the code they cover. AppStateTest drives the real
engine, so it stays in ui-desktop.
|