Plans for introducing SQL-like joins?

[dxnnymurphy]

Hi,
I have a found a post from a few years ago where a user asked about client side sql-like joins into tables:
#1509
The post states:
Ad-hoc joins of already instantiated Tables are a a work-in-progress.

Are there any updates on this?

I am looking to perform LEFT joins.
Thanks in advance!
Danny

[jordanskole]

Hey Danny,

I do a lot of joining too. I don't know what platform you are on but it is possible to "left join" kind of using a shared index. Here's how to do it:

1. create an empty table using a schema, and an index (js/ts baseTable = await worker.table(schema, { index })
2. load your first data set into the table await baseTable.update(firstDataSet)
3. load your second data set into the table await baseTable.update(secondDataSet)

This works okay so long as you are fine collapsing both tables onto the index. It gets challenging however when you want to have a "tall" table and add extra information onto that.

const item_row_schema = {
  itemId: "integer", // or whatever 
  sharedId: "integer", // or whatever
  ...rest
}

const joining_row_schema = {
  sharedId: "integer", // or whatever 
  ...rest 
}

I am not sure if this part is possible.

[jordanskole]

Sorry, I should mention that this works in 2.10, I havent updated to 3+ yet

[dxnnymurphy]

Hey Jordan,
Unfortunately I am going for the sort of 'tall' table approach where there is no shared index. I will bare that in mind and maybe see if i can come up with some sort of shared index approach.
Thanks!

[jordanskole]

Yeah, I am currently stuck here too, would love to see true joins someday soon! I know that was in the 3.0+ release notes

[aszenz]

We are using duckdb-wasm for SQL like joins in the browser

The results are then piped to perspective

[jordanskole]

@aszenz I was actually considering that, how do you like it?

[aszenz]

It's good for pre data transformations before feeding the data into perspective , also helps with running totals and sums.