[prev in list] [next in list] [prev in thread] [next in thread] 

List:       graphviz-interest
Subject:    [graphviz-interest] how would I copy node/edge attributes into
From:       erg () research ! att ! com (Emden R !  Gansner)
Date:       2005-10-31 14:21:51
Message-ID: 43666ECD.5030003 () research ! att ! com
[Download RAW message or body]

Wolfgang H?felinger wrote:

> I'm using 'gvpr' to manipulate a graph. My "manipultion" is simpy to get rid
> of edges/nodes not of interest (this part works fine).
> 
> My input graph (i.e. dot file) contains also attributes. For example: 
> 
> == input.dot ==
> digraph G {
> graph [overlap = false];
> node  [color = red];
> 
> ..
> }
> 
> After having processed my input.dot file using 'gvpr', I'm getting
> 
> digraph H {
> graph [overlap = false];
> 
> ..
> }
> 

Without seeing the script you are using, I can't say precisely what to change. \
However, as an example, suppose you want to get rid of all nodes with color=blue. \
This can be done using the script 'N[color == "blue"]{delete($G,$)}' and running gvpr \
-c. The -c flag says use the input graph as the output graph.

The basic model used by gvpr is that of awk. That is, if a node or object satisfies
a certain condition, add it to the output graph. If you supply an explicit action, \
this action overrides the default one. So, if you want the node or edge to be in the \
output graph, the action needs to add it explicitly.

For example, suppose you want to get rid of all nodes with color=blue, and make all \
other nodes have color=red. This is done by the script 'N[color != \
"blue"]{color="red"; subnode($T,$)}'. This changes the non-blue nodes to red and adds \
them to the target subgraph, which is the default output graph.

Note that the target graph will have the nodes you want, but no edges. You could add \
a rule for edges but the edges you want are the ones induced by the nodes you've \
selected. This is a frequent enough occurrence that gvpr provide the -i option add \
precisely these edges on output.

Finally, the model of in-place editing of the input graph arose frequently enough \
that the -c flag was added to gvpr. With this, the subnode and subedge commands can \
be dropped. One just uses the actions to modify the input graph, knowing that the \
changed graph will be output at the end.

> Note that node attributes are gone.
> 
> So far I have been able to "rescue" graph attributes by using builtin function
> 'copyA':
> H = subg($G,"H");
> ..
> copy(H,some_current_node);
> ..
> copyA($G,H);
> induce(H);
> write(H);
> 

For many (most?) uses, you probably don't need one of the copy functions, since you \
can work with the relevant nodes directly, either by moving them to the output graph \
or using gvpr -c.

> P.S.: 'gvpr's documentation contains  
> 
> > > copy(g : graph_t, x : obj_t) : obj_t
> > > creates a copy of object x in graph g, [..]
> 
> 
> As a non-native I have some trouble to understand. Does it mean that
> a.  a copy of x (no matter to which graph it belongs) will be created in graph 'g' \
> ; or does it mean that b.  a copy of 'x in graph g' is created and returned?
> 
> After having analyzed my gvpr's output I firmly believe that a.) is correct. 

Yes, copy means make an object in graph g that looks like object x, no matter what \
graph x belongs to. Note that for edges, this is a shallow copy. This means the new \
edge will look like x, but the endpoints may not look like the endpoints of x. This \
is done using clone, which does a deep copy.

< However, in
> such a case one would expect that 'copy' has this signature:
> 
> copy(x : obj_t, g : graph_t)
> 
> to comply to the well established copy(src,dst) convention (and also to the way \
> it's written - "creates a copy of object x in ..".
> 

You mean as with C functions:

SYNOPSIS
        #include <string.h>

        char *strcpy(char *dest, const char *src);

? :)

     Emden


[prev in list] [next in list] [prev in thread] [next in thread]