tree

[mbostock]

Introduces two new options transforms: Plot.treeNode, for laying out tree nodes, and Plot.treeLink, for laying out tree links. The new convenience mark Plot.tree combines a link, dot, and text with these options transforms to generate a tree diagram.

The Plot.treeNode options transform supports special evaluators for tree nodes. If a channel is defined as a string starting with “node:” then it is interpreted as one of the built-in node evaluators:

• node:path - the normalized, slash-delimited path to this node
• node:name - the last part of this node’s path
• node:depth - this node’s distance from its furthest ascendant
• node:height - this node’s distance from its furthest descendant
• node:internal - true if this is an internal node; false if this is a leaf node

Alternatively you can specify a channel as an {node} object where node is a function. This function is then passed the d3-hierarchy node instance.

The Plot.treeLink options transform similarly supports special evaluators for tree links. If the “node:” prefix or {node} syntax is used, it refers to child node of the link; if the “parent:” prefix syntax is used, it refers to the parent node of the link. If the {link} syntax is used, the function is passed two arguments (child, parent) and can compute the desired corresponding value.

TODO

• Use root.count or root.sum and expose a node:value somehow? Not now.
• Support different sort orders (as a treeSort option?)
• Support different separations (as a treeSeparation option?)
• Support different orientations of the tree (as a treeAnchor option?)
• Support faceting
• Support imputed nodes
• Documentation
• Fix escaped backslashes when evaluating node:name
• Have the path option default to identity
• Ignore data with undefined path values

Related #136.

[mbostock]

Okay, I figured out how to get faceting and imputed nodes to work: now the tree transforms generate one element per node or per link, respectively, and for imputed nodes the associated data is null (but you can still access node-related metadata using a node accessor). Here’s a little example of the flare dataset with small nodes (<5000 bytes) on the left and large nodes (≥5000 bytes) on the right. Fun!

[Fil]

Excellent! I've added a few suggestions (which can all be addressed later) in #843.