Is there a way to only select vias that cross a shape boundry?
I have a very large FET that has numerous Source and Drain Vias and the top level metal covering them now has to be slotted leaving a large amount of existing Vias to be manually deleted where the slots pass over them.
I can generate another layer that consists of simply the slots and then, if CDS OA 6.1.5 has the capability, then one could simply select only the Vias that touch the newly created shapes.
Is that possible to do in Skill?
Use the bBox from your slots and then this Skill command to get your vias. Finally you can use geSelectFig for each of the vias.
dbViaQuery( d_cellviewIdl_bBox[ x_startLevel ] [ x_stopLevel ][ n_filterSize ] [ l_transform ] ) => ld_via | nil
Returns a list of all the vias in a cellview whose bounding boxes overlap the region specified by l_bBox.
In reply to dmay:
Building on what Derek suggested (nice tip!), here is some more code around it. Assume that the slot shapes have been selected and then stored in a list, in my example I used testShapes and 'lcv' is the layout cellview database object, then you can identify and delete the overlapped vias like this:
vias = apply('nconc foreach(mapcar shape testShapes dbViaQuery(lcv shape~>bBox )))
The first step build a "flat" list of the list-of-list shapes returned by the foreach-mapcar; the dbViaQuery might return more than one shape. Then map the function dbDeleteObject over this list to delete each via found.
Hopefully this will answer your question?
In reply to skillUser:
A little further optimization (just to give an alternative and illustrate some of the flexibility of SKILL) - use:
vias=foreach(mapcan shape testShapes dbViaQuery(lcv shape~>bBox))mapc('dbDeleteObject vias)
The foreach mapcan will append the lists as it goes along - no need to use the apply 'nconc afterwards. And for the deletion, since there's not much likelihood that you'll want the return value of the mapcar (which will be a list of t's and nil's depending upon the success of the dbDeleteObject), use mapc which just returns the original list (and so is cheaper).