Pool and re-use web workers

[oskarhane]

This introduces a web worker pool (generic so it can be used for other expensive computations as well).
Currently the pool size for cypher queries is hard coded in https://github.com/neo4j/neo4j-browser/pull/782/files#diff-046ce42896ad79b95a44f41c7faa1f94R39

This could possibly be a user config in the future, if we see the need to for it.
If no worker is free, the jobs are queued and executed in order.

This can be tested by setting the max pool size to 1 and execute a long running query (this takes about 10 secs), like

MATCH (n:Movie)<--()-->(m:Movie)<--()-->(o:Movie)<--()-->(p:Movie)<--()-->(q:Movie)
WITH n,p,count(*) as c
RETURN count(*)

on the movie data set and directly afterwards execute a few more. These will then wait with a spinner.

In the blob directory you'll only see one ww blob.
I've set the hard coded pool size to 10 for no particular reason. I've never hit that limit if I'm not hammering the keyboard with queries.

Try it

Open browser, open devtools -> Application.
Expand Other under Frames.
See all the uuid named blobs. That's ww (to verify, right click and show in network tab).

1. Reload neo4j-browser
2. Click a few meta items in sidebar (or run any queries). Wait for about 30 secs.
3. Compare results with and without the changes in this pr.

Before

After

changelog: Fix issue where neo4j-browser kept allocating more and more disk space

[reznord]

This is a huge improvement and saves a lot of disk space

[oskarhane]

For future reference: I've rewritten the original (PoC) implementation to a full blown web worker pool and focus the api on the work that are to be done rather than the web workers themselves.

const createWorker = () => /* ... your create worker function */
const maxWorkers = 10
const pool = new WorkPool(createWorker, maxWorkers)

const payload = {type: 'TEST', data: {x: 1}}

const work = pool.doWork({id: 1, payload, onmessage: (msg) => {
  // Handle response
  work.finish() // You need to manually finish the work to release the worker
}})

// additional tricks
const work = pool.getWorkById(1) // Get access to work
work.finish() // cancels if it's in queue
work.onFinish(() => { console.log('finishing now') }) // Set a function to be called once worker is done
work.execute({type: 'SECOND_MESSAGE', data: {x: 2}}) // Post additional message to the worker that's doing the `work`

[pe4cey]

FYI if we choose to surface the worker pool as :config in the future we will probably need to only allow worker pools > 0 as a pool of 0 is always in a BUSY state

[oskarhane]

Yeah, 0 would be infinite imo.
New ones won't be recreated as long as there are existing free ones anyway.