[Bioc-devel] small request on graph::graphBAM()

[Robert Castelo]

hi Valerie,

thanks a lot for implementing this!!

regarding addNode() i have no specific request or opinion about it.


best regards,
robert.


On 2/19/13 9:44 PM, Valerie Obenchain wrote:
> This change is now implemented in devel 1.37.6.
>
> > nms <- paste0("Y", 8:11)
> > df <- data.frame(from=nms, to=nms, weight=rep(1, 4))
>
> Node names are not sorted when full vector of names is provided:
> > nodes(graphBAM(df, nodes=c(nms, "foo")))
> [1] "Y8"  "Y9"  "Y10" "Y11" "foo"
>
> Names are sorted when only a zero-degree node is provided:
> > nodes(graphBAM(df, nodes="foo"))
> [1] "foo" "Y10" "Y11" "Y8"  "Y9"
>
>
> A side effect of this change is that addNode(), which calls 
> graphBAM(), was returning unsorted node names. To remain consistent 
> with previous behavior I've sorted these results. Currently addNode() 
> has no way to specify a full list of node names. Do you have an 
> opinion on addNode() behavior? OK as is?
>
>
>
> Valerie
>
>
>
>
> On 02/07/2013 08:52 AM, Robert Castelo wrote:
>> Great, thanks!
>> robert.
>>
>> On 02/07/2013 05:44 PM, Valerie Obenchain wrote:
>>> Hi Robert,
>>>
>>> I will look into this next week. Thanks for providing details and
>>> examples in the request.
>>>
>>> Valerie
>>>
>>> On 02/06/2013 04:10 PM, Robert Castelo wrote:
>>>> dear maintainers of the graph package,
>>>>
>>>> i have a small request for the constructor function graphBAM() from 
>>>> the
>>>> graph package.
>>>>
>>>> the current help page of the graphBAM() constructor says the following
>>>> about the argument called 'nodes':
>>>>
>>>> A character vector of node labels. Use this to add degree
>>>> zero nodes to the graph. If ‘NULL’, the set of nodes found
>>>> in ‘from’ and ‘to’ will be used.
>>>>
>>>> It would be great if when nodes != NULL and includes the set of nodes
>>>> found in 'from' and 'to', 'nodes' would be taken as the set of 
>>>> nodes of
>>>> the graph, without re-ordering it.
>>>>
>>>> currently this function has the following line (*):
>>>>
>>>> nodes <- sort(unique(c(as.character(df$from), as.character(df$to),
>>>> nodes)))
>>>>
>>>> this is a bit of a headache for me because the alphabetical 
>>>> ordering of
>>>> vertices does not respect integer ordering (obviously):
>>>>
>>>> library(graph)
>>>> df <- data.frame(from=paste0("Y", 1:10), to=paste0("Y", 2:11),
>>>> weight=rep(1, 10))
>>>> nodes(graphBAM(df, nodes=paste0("Y", 1:10)))
>>>> [1] "Y1" "Y10" "Y11" "Y2" "Y3" "Y4" "Y5" "Y6" "Y7" "Y8" "Y9"
>>>>
>>>> however, because i often deal with vectors of vertex labels created as
>>>>
>>>> v <- paste0("Y", 1:p)
>>>>
>>>> with p whatever positive integer, my vector of vertex labels is not
>>>> identical to the vector returned by graph::nodes() which 
>>>> complicates the
>>>> way i have to deal with this feature. re-ordering my vector of vertex
>>>> labels by alphabetical order implies doing it *each* time i do one of
>>>> those 'paste0()' calls, which is quite often, and this seems quite
>>>> common, see
>>>> http://simplystatistics.org/2013/01/31/paste0-is-statistical-computings-most-influential-contribution-of-the-21st-century 
>>>>
>>>>
>>>>
>>>> :)
>>>>
>>>> for this reason it would be great for me if the argument 'nodes' in
>>>> graphBAM() would have this additional functionality of forcing an
>>>> ordering on the vertex labels when 'nodes' include all vertices 
>>>> forming
>>>> the given edges.
>>>>
>>>> i think this would amount to replacing the line above (*) by
>>>>
>>>> stopifnot(all(!duplicated(nodes))) ## given nodes should be unique
>>>> edgeNodes <- unique(c(as.character(df$from), as.character(df$to))
>>>> if (!all(edgeNodes %in% nodes))
>>>> nodes <- sort(unique(c(nodesEdges, nodes)))
>>>>
>>>> note that with this code when one just adds extra degree zero 
>>>> nodes, the
>>>> behavior does not change with respect to what it is doing right now 
>>>> and
>>>> if one likes to stick to the alphabetical ordering, you can always 
>>>> give
>>>> 'nodes=sort(myvertexlabels)' as argument to the graphBAM() constructor
>>>> function.
>>>>
>>>> thanks!!
>>>> robert.
>>>>
>>>> _______________________________________________
>>>> Bioc-devel at r-project.org mailing list
>>>> https://stat.ethz.ch/mailman/listinfo/bioc-devel
>>>
>>>
>>
>