Manpages - pack.3

The function ccomps takes a graph g and returns an array of pointers to subgraphs of g which are its connected components. cnt is set to the number of components. If pfx is non-NULL, it is used as a prefix for the names of the subgraphs; otherwise, the string ``_cc_’’ is used. Note that the subgraphs only contain the relevant nodes, not any corresponding edges. Depending on the use, this allows the caller to retrieve edge information from the root graph.

The array returned is obtained from malloc and must be freed by the caller. The function relies on the mark field in Agnodeinfo_t.

pinned) This is identical to ccomps except that is puts all pinned nodes in the first component returned. In addition, if pinned is non-NULL, it is set to true if pinned nodes are found and false otherwise.

This function takes a subgraph g and finds all edges in its root graph both of whose endpoints are in g. It returns the number of such edges and, if this edge is not already in the subgraph, it is added.

This function returns non-zero if the graph g is connected.

ip) putGraphs packs together a collection of laid out graphs into a single layout which avoids any overlap. It takes as input ng graphs gs. For each graph, it is assumed that all the nodes have been positioned using pos, and that the xsize and ysize fields have been set.

If root is non-NULL, it is taken as the root graph of the subgraphs gs and is used to find the edges. Otherwise, putGraphs uses the edges found in each graph gs[i].

For the modes l_node, l_clust, and l_graph, the packing is done using the polyomino-based algorithm of Freivalds et al. This allows for a fairly tight packing, in which a convex part of one graph might be inserted into the concave part of another. The granularity of the polyominoes used depends on the value of ip->mode. If this is l_node, a polyomino is constructed to approximate the nodes and edges. If this is l_clust, the polyomino treats top-level clusters as single rectangles, unioned with the polyominoes for the remaining nodes and edges. If the value is l_graph, the polyomino for a graph is a single rectangle corresponding to the bounding box of the graph.

The mode l_node specifies that the graphs should be packed as an array.

If ip->doSplines is true, the function uses the spline information in the spl field of an edge, if it exists. Otherwise, the algorithm represents an edge as a straight line segment connecting node centers.

The parameter ip->margin specifies a boundary of margin points to be allowed around each node. It must be non-negative.

The parameter ip->fixed, if non-null, should point to an array of ng booleans. If ip->fixed[i] is true, graph gs[i] should be left at its original position. The packing will first first place all of the fixed graphs, then fill in the with the remaining graphs.

The function returns an array of points which can be used as the origin of the bounding box of each graph. If the graphs are translated to these positions, none of the graph components will overlap. The array returned is obtained from malloc and must be freed by the caller. If any problem occurs, putGraphs returns NULL. As a side-effect, at its start, putGraphs sets the bb of each graph to reflect its initial layout. Note that putGraphs does not do any translation or change the input graphs in any other way than setting the bb.

This function uses the bb field in Agraphinfo_t, the pos, xsize and ysize fields in nodehinfo_t and the spl field in Aedgeinfo_t.

ip) This function takes ng subgraphs gs of a root graph root and calls putGraphs with the given arguments to generate a packing of the subgraphs. If successful, it then invokes shifts the subgraphs to their new positions. It returns 0 on success.

ip) This function simply calls packGraphs with the given arguments, and then recomputes the bounding box of the root graph.

fixed) uses packSubgraphs to place the individual subgraphs into a single layout with the parameters obtained from getPackInfo. If successful, dotneato_postprocess is called on the root graph.

putRects packs together a collection of rectangles into a single layout which avoids any overlap. It takes as input ng rectangles bbs.

Its behavior and return value are analogous to those of putGraphs. However, the modes l_node and l_clust are illegal. The fields fixed and doSplines of ip are unused.

packRects is analogous to packGraphs: it calls putRects and, if this is successful, it translates the rectangles in bbs appropriately.

The library provides several functions which can be used to tailor the packing based on graph attributes.

pinfo) analyzes p as a string representation of pack mode, storing the information in pinfo. If p is “cluster”, it returns l_clust; for “graph”, it returns l_graph; for “node”, it returns l_node; for “array”, it returns l_array; for “aspect”, it returns l_aspect; otherwise, it returns dflt. Related data is also stored in pinfo.

pinfo) This function processes the graph’s “packmode” attribute, storing the information in pinfo. It returns pinfo->mode. The attribute is processed using parsePackModeInfo with dflt passed as the default argument.

This function returns a pack_mode associated with g.

This function queries the graph attribute “pack”. If this is defined as a non-negative integer, the integer is returned; if it is defined as “true”, the value dflt is returned; otherwise, the value not_def is returned.

pack_info* pinfo) This function calls both getPackModeInfo and getPack, storing the information in pinfo. dfltMargin is used for both integer arguments of getPack, with the result saved as pinfo->margin. It returns pinfo->mode.