d3.join? about d3-array

[Firescar96]

https://github.com/Firescar96/d3-array/tree/master/src/join
https://github.com/Firescar96/d3-array/blob/master/test/join-test.js

I haven't written docs for it yet, but the tests should be helpful for how it works til I do. I went with the cartesian product via nested loop join as the default. There's also a sort merge join option, or a person can write their own.

[Fil]

My feeling after reading this several times over the last month is that it appears too complex (for me) to memorize the arguments and the way this would work. I'd fear not really seeing what's happening, and I'd personally rather create an index first for the key then join "manually". Detailed examples using ES6 and d3.group, d3.rollup or d3.index (#136) would suit me better than the "magic" of d3.join. Also a risk of (mental) collision with selection.join.

[timelyportfolio]

Just in case it helps think through, here is lodash-joins. From the R side, dplyr joins.

[Firescar96]

sure I'll give it a try, going to use sorted merge joins for functionality like

d3.join()
    .key(d => d.id)
    .reduce((a, b) => a.properties.rate = +b.rate)
    (dataset1, dataset2)

I'm not sure what you meant @mbostock by the case join.key with no arguments.

EDIT: Ah, after looking at other d3 modules as a guide for how to implement this I see what you mean, I'll use this syntax instead:

let joinOp = d3.join(dataset1, dataset2)
    .key(d => d.id)
    .reduce((a, b) => a.properties.rate = +b.rate)

console.log(joinOp.apply())

[Firescar96]

Benchmarks against lodash-joins, using their tests

[curran]

Would we ever want async join? Each record, for example, making an API request?