Skip to content
Permalink
Browse files

add weekly meeting notes

  • Loading branch information...
ahdinosaur committed Feb 5, 2019
1 parent a26334e commit 9df7f4b1a2fda541f1ede3577976501b1e1e5989
Showing with 82 additions and 0 deletions.
  1. +82 −0 meetings/2019-02-05-weekly.md
@@ -0,0 +1,82 @@
# Sunrise choir - 05-02-19

## Participants

- Pie(t)
- Mat(t)
- Mik(ey)

## Agenda

- [x] chickens
- [x] piet update
- ssb-friends as sql query
- rust is fast
- [x] January newsletter
- wrote by Piet and Mikey [here](https://hackmd.io/3WVs6nuRQjuc7p3B9_QG7w)
- made an OpenCollective, waiting on OpenCollective approval
- [x] Patchwork: what to do about it
- [x] chookouts

## piet update

sql query to get friends and friends-of-friends

```sql
SELECT
author as id
FROM
authors
WHERE authors.id IN (
SELECT
contact_author_id
FROM contacts_raw
WHERE author_id == 1 AND state == 1
UNION
SELECT
friend_contacts_raw.contact_author_id
FROM contacts_raw
join contacts_raw AS friend_contacts_raw ON friend_contacts_raw.author_id == contacts_raw.contact_author_id
WHERE contacts_raw.author_id == 1
AND contacts_raw.state == 1
AND friend_contacts_raw.state == 1
EXCEPT
SELECT
contact_author_id
FROM contacts_raw
WHERE author_id == 1
AND state == -1)
```

- pie: using console.time() and time.end()
- backlinks would take 0.5 seconds (!)
- ended up writing a benchmark
- in Rust, took 25 microseconds
- was misled because JavaScript main thread was happening, so query wouldn't run immediately
- pie: had a go at using JSON-RPC server & client in Rust
- to answer the question: what's the latest flume sequence?
- real fast if requests were batched up
- on my mind to copy xray type architecture, where we have Rust servers that are called from JavaScript clients

## Patchwork: what to do about it?

- mat: since latest meeting, Mix built pre-releases of new Patchwork
- Mix wants to run Patchwork alongside Patchbay without clobbering indexes
- taken pressure off a release
- mat: but the latest version of Patchwork is okay, but not great
- shonky, click a button and nothing happens, then something happens later
- has some regressions, using a better architecture between Patchwork and sbot, but now Patchwork is bound to sbot for updates, but sbot is unreliable and has memory leaks
- pie: as we do pull out more bits from sbot, it should get better
- mat: yeah, as i do Sunrise Choir work, it will solve the problems, but i don't know if i can be bothered to hack around it until that happens
- mat: apparently better memory, but only the case until sbot craps out
- mat: my intuition is to ignore Patchwork for a while until the deeper issues are solved
- pie: could be a good target for us to pair on next week
- i have to write a table for the votes and reduce them, when that's done we should have all the entities modelled, and then with any luck we could start porting over things to Patchwork
- mat: i think most of the work i did with the new Patchwork is transferrable to our new Rust stuff, i think it will make our life easier, i don't think it was all for nothing
- what i think would be a great win is if the next release of Patchwork had this new Sunrise Choir stuff inside
- pie: one thing that feels like a sticky issue is getting the native module building
- mat: have you tried using travis to cross build?
- that's what i'm using for Patchwork's native modules
- https://github.com/ssbc/paulcbetts-cld-prebuilt
- mik: new prebuild is prebuildify: https://github.com/mafintosh/prebuildify
- build on travis, publish to GitHub releases, then use `prebuildify-ci` to download releases before `npm publish`

0 comments on commit 9df7f4b

Please sign in to comment.
You can’t perform that action at this time.