when zooming in, there should be no need to mark all nodes as hidden

[wonder-mice]

I noticed that d3-flame-graph casually traverses the entire data structure for different reasons. This makes many operations slower than it would be desired.

Specifically, when zooming in, it should be enough to mark nodes on the same level as zoomed node as hidden and when such flag is set, return null from getChildren().

[spiermar]

Possibly, feel free to submit a PR. :)

[wonder-mice]

Btw, what are the benefits of using d3? I though the idea is to handle large data sets with ease, but all d3 examples I'm looking at are super slow even on small data sets. E.g. if I just create divs for each node manually, will it be better or worse?

[spiermar]

Bit more complicated than that I believe, but sure, there are other options that might perform better in certain cases, just a lot more heavy lifting to be done.

[wonder-mice]

That's exactly what I'm curious about. What this heavy lifting is and how heavy is it. I know that from outside every project/code looks trivial and simple, so I'm trying to understand what I'm missing.

[wonder-mice]

Is there particular reason why hide and fade are injected into data items instead of d3 nodes? It'll be faster to access them in nodes and allows less conflicts with provided data json format (which can be anything if we'll go with custom getValue, getDelta, getChildren callbacks).

[spiermar]

No particular reason.