Massimiliano Mirra said: (by the date of Mon, 08 Sep 2003 01:38:36 +0200) > One more chance I considered is that `geometry' files don't > necessarily need be flat text files containing geometric data, they > can be executable files generating geometric data wow, really cool [snip example] I like this idea very much. > Yet another kind of attribute is the `transformations' file, that > covers non-rigid transformations (already covered by axis) which might > look like: > > stretch: 0.5 10 > squeeze: 11 > rotate: 10 and then we will need some kind of engine that reads (and executes where necessary) whole this stuff.. and - like phrostie just said - display it (through .dxf) on the screen :)) > Only when the user is sure about the transformations, he `commits' yeah. this toDxf parsing tool should use another tool that `commits' > I like clean, minimalist interfaces, too. A canvas widget might just > help you build prototypes to quickly try out ideas, because so much > work has already been done for you, and when you have found the > combination of features and functionality you like the best, you can > get back to xlib and code an efficient interface. Consider that. ;-) good point :) > > like this: > > > > +----------------------------------------------------+ > > | layer | color | linetype | name | > > +----------------------------------------------------+ > > | road | red | - - - | stripe | > > . > > . > > . > > > > here is displayed a single filename. Like in midnight commander, but > > the real filename is: > > > > <layer> <color> <name> <unique number> > > <linetype> <contents type> > > l#road-c#red-l#=.=.=-stripe-1045317918-1278-225423.line > > Yes, somewhere the object metadata has to be displayed to the user, > and that's a good way to display it. But still I'm very wary of > putting too much into a file name. Eric Wilhelm said why to not do this. In the post named: Subject: [cad-linux] Re: CAD format, virtual filesystem, editing he convinced my, that this is not a good way. But we have many more possibilites to store attributes. One of them is a file named attributes in each xspace directory object. Another one is a tag 'Attributes:' in Bruno's filesystem. Bruno asked what if we don't want to carry attributes around (the attribute will be stored inside a file)? Bruno (hope you will read this) the attributes will be stored in separate file. Imagine a following line inside a file: Attribute: Layer road ok. so this object belongs to so-called road layer. We copied this object somwhere. Did we choose to copy a definition of road layer (another file that defines this layer - anything you would like to define: color, linetype, width) along with it? If yes, then the layer will exist in new drawing. If no - the object will belong to "layer 0". Bruno - am I missing something or this is a solution for 'tables' ? but then - we will have to develop a good way to store attributes definitions. One sparks to my mind: if a directory contains directory called Attributes/ then all attributes defined there are descendant to all directories that fall in the same directory tree that Attributes/ is placed in. Inside Attributes/ directory you can have subdirectories for each attribute. So for example following file: ./Attributes/Layers/road will contain all information about a road layer (linetype, color, width, maybe even material? any definition you want) note that attributes can be redundant, because you will also have a file: ./Attributes/Linetypes/dotline (with definition of linetype) and ./Attributes/Color/red (with contents like this Color: #aa0000 ) but you can also have such a file: ./Attributes/Linetypes/reddotline with contents: Linetype: dotline Color: red Anyone could produce any useful combination of attributes. and if a file: ./Attributes/Materials/stucco could be an executable, to generate stucco information... man.. that variety of possibilities astounds me -- # Janek Kozicki