Minor Description
I have noticed some inconsistencies in the way default styles (things like colors, fonts, padding/margins, etc.) are set in the constructors of the Terra*Skin classes, such as:
- Color values set directly from the theme vs. set through the setter methods.
- Missing color setter methods using the theme color index.
- Derived colors being set in two different places, potentially leading to inconsistencies because of the duplication.
- Possible inconsistencies between controls in the theme colors used (such as for backgrounds, borders, etc.) – this has not been verified, because it will require a lot of research.
- Since these defaults are all set in code, even though (for instance) the theme colors themselves are set in the "TerraTheme_default.json" file, changing the overall look-and-feel is difficult at present.
So, for all these reasons I propose to use a new "terra_theme_defaults.json" file, which can be overridden by a new setting in the "TerraTheme*.json" file(s) which will set all the defaults for all the Terra*Skin classes similar to this (taken from TerraExpanderSkin):
Previously:
setBackgroundColor(theme.getColor(4));
titleBarBackgroundColor = theme.getColor(10);
titleBarBorderColor = theme.getColor(7);
titleBarColor = theme.getColor(12);
shadeButtonColor = theme.getColor(12);
disabledShadeButtonColor = theme.getColor(7);
borderColor = theme.getColor(7);
padding = new Insets(4);
New:
TerraExpanderSkin : {
backgroundColor : 4,
titleBarBackgroundColor : 10,
titleBarBorderColor : 7,
titleBarColor : 12,
shadeButtonColor : 12,
disabledShadeButtonColor : 7,
borderColor : 7,
padding : 4
},
Note: this approach also fits well with the overall Pivot philosophy of doing things in a "declarative" fashion (via BXML / JSON files) rather than in code, where possible.
I have prototyped this in a couple of classes and it seems to work well, AND it has revealed a number of the color properties that were missing the "setXXX(int index)" methods.
Note: there is potential for slowdown during control construction since we're executing a LOT more code to set these defaults (including expensive reflection operations inside BeanAdapter), but I have not yet been able to measure the speed differential to see if it will be of concern.