| 1 |
- Since we cannot use the createwindowevent (it is sent to the Parent, |
| 2 |
and not to us) we need to make a special call after creating a |
| 3 |
window with the create window information. The important part is |
| 4 |
the window dimension. Or! We could simply store this info in the |
| 5 |
LWindow structure. |
| 6 |
- Need an encapsulation of resource manager. Can be VERY simple |
| 7 |
subclass of LDisplay. |
| 8 |
- Need an encapsulation of fonts. I think we should ask for fonts |
| 9 |
from a third party which will cache fontname->fontID mappings. It |
| 10 |
will return fast an ID if one has been requested already. Also, it |
| 11 |
will not free any fonts until the application quits. We don't have |
| 12 |
to do this because the server will do a quick return when the font |
| 13 |
has been referenced by another application. Still, it is |
| 14 |
advantageous to eliminate any round-trip-times from the application |
| 15 |
that are easy to circumvent. |
| 16 |
- Need to think about implementing full button functionality. How |
| 17 |
does a button event get reported? It would be nice if we could |
| 18 |
specify a simple mechanism for events to propagate. Thus, we could |
| 19 |
have containers inside dialogs, but the events still go to the |
| 20 |
dialog. Or, the container could request a chance to filter |
| 21 |
notification events. |
| 22 |
- Need to think about minimum dialog template features. How to |
| 23 |
specify the dialog, and how to store this in a binary. Can we do |
| 24 |
some clever trick to implement a resource in a binary file? We |
| 25 |
could link a symbol? We could execute a simple resource compiler. |
| 26 |
Whoa. This will be very cool. |
| 27 |
It would also be nice to be able to specify both dialog units, |
| 28 |
relative to the default dialog font, and pixels for controls that |
| 29 |
need to reference exact bitmap sizes and such. |
| 30 |
- Need to focus on inter-window messaging. Look at how a child |
| 31 |
notifies the parent about a state change. There is the event ID as |
| 32 |
accepted by the parent, I think it is best to use a client message. |
| 33 |
Then there is the window ID so we can know what happened. Then, |
| 34 |
there is the event we are trying to notify ABOUT. Click, change, |
| 35 |
etc. Now, we could use a class interface for notification and |
| 36 |
circumvent the message map. This has the advantage of obviating the |
| 37 |
need for atoms and the like, and it makes it easier to cope with |
| 38 |
the structures we may want to pass. The down side is that it is a |
| 39 |
'somewhat' parallel messaging mechanism. This, the parent need to |
| 40 |
have a place to call into. I'm heasitant to start using virtual |
| 41 |
call tables since these have crummy extensibility characteristics. |
| 42 |
So instead, we could use the event dispatch table, define some |
| 43 |
high-numbered event numbers, and use our 'event' structure. Hmm. |
| 44 |
I'm pretty shy of using the C++ class structuring mechanisms. Can |
| 45 |
you tell? |
| 46 |
- In extracting from dialog templates, it would be good to have a |
| 47 |
couple of helper functions for the simpler cases, origin, extent, |
| 48 |
title, that return the arguments if the object is defined. |
| 49 |
- Need to implement dialog specific resources so we can compute the |
| 50 |
text cell size used in scaling. |
| 51 |
- Also, we need to make sure children notify the right parent on |
| 52 |
dialog creation. |
| 53 |
- Also, it would be good to allow preprocessing of dialog templates |
| 54 |
st. we can do simple math and some string concatenation. |
| 55 |
- Also, we want to fogure out how to prevent the window manager from |
| 56 |
controlling window placement of these dialogs. |
| 57 |
- Also, can we prevent a window from being resized? |
| 58 |
- Compiler bug? |
| 59 |
It is possible we have found a compiler error. If we have a method |
| 60 |
declared in a base class as well as a derived class, and if the |
| 61 |
arguments for these functions are different, a call to this method |
| 62 |
using an object pointer SHOULD find the base class method by |
| 63 |
matching the argument types. Right? |
| 64 |
- What is the goal here? |
| 65 |
It looks like the goal of XO is to provide an efficient library for |
| 66 |
interfacing to a graphical display. I'm finding that a primary |
| 67 |
motivating factor is the desire to visualize numerical data. This |
| 68 |
is turning out to be like Razor except that *I* want use the tool |
| 69 |
myself. |