Right now, we use clojure.walk/prewalk to walk the document hiccup structures and replace entries such as [:vega-lite {....}] with actual reagent component calls (e.g. [oz.core/vega-lite {...}]). The problem with this is that prewalk does not stop traversing the children of these nodes in the tree, and in particular end up iterating through all of the :data entries in the visualization specs, which is far from ideal. This potentially has an impact on performance for larger chunks of data (though probably not significant, since, we're already parsing and translating to json). More importantly though, while in general it's constructed so that it shouldn't actually edit any of the data, I don't like that I have to even think about it (and it's not impossible that there's some edge case where this could actually pose a problem).
Zippers may be the ideal solution, but are sort of a pain to write. We may be a able to get away with a custom recursion as well.
Right now, we use
clojure.walk/prewalkto walk the document hiccup structures and replace entries such as[:vega-lite {....}]with actual reagent component calls (e.g.[oz.core/vega-lite {...}]). The problem with this is thatprewalkdoes not stop traversing the children of these nodes in the tree, and in particular end up iterating through all of the:dataentries in the visualization specs, which is far from ideal. This potentially has an impact on performance for larger chunks of data (though probably not significant, since, we're already parsing and translating to json). More importantly though, while in general it's constructed so that it shouldn't actually edit any of the data, I don't like that I have to even think about it (and it's not impossible that there's some edge case where this could actually pose a problem).Zippers may be the ideal solution, but are sort of a pain to write. We may be a able to get away with a custom recursion as well.