The morph/unmorph verbs are used to create temporary representations of the graph, such as e.g. its search tree or a subgraph. A morphed graph will accept any of the standard dplyr verbs, and changes to the data is automatically propagated to the original graph when unmorphing. Tidygraph comes with a range of morphers, but is it also possible to supply your own. See Details for the requirement for custom morphers. The crystallise verb is used to extract the temporary graph representation into a tibble containing one separate graph per row and a name and graph column holding the name of each graph and the graph itself respectively. convert() is a shorthand for performing both morph and crystallise along with extracting a single tbl_graph (defaults to the first). For morphs were you know they only create a single graph, and you want to keep it, this is an easy way.

morph(.data, .f, ...)

unmorph(.data)

crystallise(.data)

crystallize(.data)

convert(.data, .f, ..., .select = 1, .clean = FALSE)

Arguments

.data

A tbl_graph or a morphed_tbl_graph

.f

A morphing function. See morphers for a list of provided one.

...

Arguments passed on to the morpher

.select

The graph to return during convert(). Either an index or the name as created during crystallise().

.clean

Should references to the node and edge indexes in the original graph be removed when using convert

Value

A morphed_tbl_graph

Details

It is only possible to change and add to node and edge data from a morphed state. Any filtering/removal of nodes and edges will not result in removal from the main graph. However, nodes and edges not present in the morphed state will be unaffected in the main graph when unmorphing (if new columns were added during the morhped state they will be filled with NA).

Morphing an already morhped graph will unmorph prior to applying the new morph.

During a morphed state, the mapping back to the original graph is stored in .tidygraph_node_index and .tidygraph_edge_index columns. These are accesible but protected, meaning that any changes to them with e.g. mutate will be ignored. Furthermore, if the morph results in the merging of nodes and/or edges the original data is stored in a .data column. This is protected as well.

When supplying your own morphers the morphing function should accept a tbl_graph as its first input. The provided graph will already have nodes and edges mapped with a .tidygraph_node_index and .tidygraph_edge_index column. The return value must be a tbl_graph or a list of tbl_graphs and these must contain either a .tidygraph_node_index column or a .tidygraph_edge_index column (or both). Note that it is possible for the morph to have the edges mapped back to the original nodes and vice versa (e.g. as with to_linegraph). In that case the edge data in the morphed graph(s) will contain a .tidygraph_node_index column and/or the node data a .tidygraph_edge_index column. If the morphing results in the collapse of multiple columns or edges the index columns should be converted to list columns mapping the new node/edge back to all the nodes/edges it represents. Furthermore the original node/edge data should be collapsed to a list of tibbles, with the row order matching the order in the index column element.

Examples

create_notable('meredith') %>%
  mutate(group = group_infomap()) %>%
  morph(to_contracted, group) %>%
  mutate(group_centrality = centrality_pagerank()) %>%
  unmorph()
#> # A tbl_graph: 70 nodes and 140 edges
#> #
#> # An undirected simple graph with 1 component
#> #
#> # Node Data: 70 × 2 (active)
#>    group group_centrality
#>    <int>            <dbl>
#>  1     1              0.1
#>  2     1              0.1
#>  3     1              0.1
#>  4     1              0.1
#>  5     1              0.1
#>  6     1              0.1
#>  7     1              0.1
#>  8     2              0.1
#>  9     2              0.1
#> 10     2              0.1
#> # ℹ 60 more rows
#> #
#> # Edge Data: 140 × 2
#>    from    to
#>   <int> <int>
#> 1     1     5
#> 2     1     6
#> 3     1     7
#> # ℹ 137 more rows