Migrate to spago v1

[pete-murphy]

Description of the change

Resolves #315

This migrates the client and staging Spago projects to use latest (1.0.4), and updates compiler version and package set to latest while I was at it. Note that 0.15.15 is latest version available on Hackage at time of writing.

For client updates, I just followed these instructions in Spago repo: https://github.com/purescript/spago#migrate-from-spagodhall-to-spagoyaml.

For staging/Haskell updates I followed the instructions in RELEASE.md that I also updated in these changes. I did not change stack.yaml resolver because 0.15.15 is still on lts-20.9.

This is a large diff, but most of that is in generated files. Ignoring staging/* (effectively generated from following the RELEASE.md instructions) and client/spago.lock, the diff is much more reasonable:

$ git diff origin/master --stat ':(exclude)staging/*' ':(exclude)client/spago.lock'

 .github/workflows/ci.yml              |  4 +--
 CHANGELOG.md                          |  3 ++
 README.md                             |  9 +++---
 RELEASE.md                            | 72 ++++++++++++++++++++++++--------------------
 client/config/dev/Try.Config.purs     | 15 ---------
 client/config/prod/Try.Config.purs    | 15 ---------
 client/package.json                   | 19 ++++++------
 client/packages.dhall                 |  5 ---
 client/spago.dhall                    | 45 ---------------------------
 client/spago.yaml                     | 48 +++++++++++++++++++++++++++++
 client/src/Try/Config.js              | 19 ++++++++++++
 client/src/Try/Config.purs            |  5 +++
 client/src/Try/SharedConfig.purs      |  7 +++--
 client/updateSharedConfigVersions.mjs | 20 +++++++++---
 stack.yaml                            |  2 +-
 stack.yaml.lock                       |  8 ++---
 16 files changed, 154 insertions(+), 142 deletions(-)

That said, I am open splitting this up into separate MRs.

---

Checklist:

• Added the change to the changelog's "Unreleased" section with a reference to this PR (e.g. "- Made a change (#0 by @)")
• Linked any existing issues or proposals that this pull request should close
• Updated or added relevant documentation
• Added a test for the contribution (if applicable)

[pete-murphy]

Oh no, I tried running locally with same memory settings as production build (set -o noglob && stack exec trypurescript -- +RTS -N2 -A128m -M3G -s -RTS 8081 $(spago sources)) and ran out of memory.

❯ (set -o noglob && stack exec trypurescript -- +RTS -N2 -A128m -M3G -s -RTS 8081 $(spago sources))
Reading Spago workspace configuration...

✓ Selecting package to build: try-purescript-server


Warning: nix (Nix) 2.34.7+1 is on the PATH (at /run/current-system/sw/bin/nix)
         but Stack's Nix integration is disabled. To mute this message in
         future, set notify-if-nix-on-path: false in Stack's configuration.
         
trypurescript: Heap exhausted;
trypurescript: Current maximum heap size is 3221225472 bytes (3072 MB).
trypurescript: Use `+RTS -M<size>' to increase it.
  36,907,900,144 bytes allocated in the heap
   7,635,067,328 bytes copied during GC
   3,479,980,704 bytes maximum residency (28 sample(s))
      21,874,656 bytes maximum slop
            3739 MiB total memory in use (0 MB lost due to fragmentation)

                                     Tot time (elapsed)  Avg pause  Max pause
  Gen  0       108 colls,    42 par    2.344s   3.292s     0.0305s    0.1313s
  Gen  1        28 colls,     5 par   39.524s  111.686s     3.9888s    7.0771s

  Parallel GC work balance: 86.50% (serial 0%, perfect 100%)

  TASKS: 9 (1 bound, 8 peak workers (8 total), using -N2)

  SPARKS: 7427 (7381 converted, 0 overflowed, 0 dud, 0 GC'd, 46 fizzled)

  INIT    time    0.001s  (  0.080s elapsed)
  MUT     time   42.664s  (  9.678s elapsed)
  GC      time   41.868s  (114.978s elapsed)
  EXIT    time    0.002s  (  0.005s elapsed)
  Total   time   84.534s  (124.742s elapsed)

  Alloc rate    865,091,047 bytes per MUT second

  Productivity  50.5% of total user, 7.8% of total elapsed

Running with -M4G succeeds.

[pete-murphy]

Dumping the result of investigating this with Opus: https://gist.github.com/pete-murphy/bd776708d72a00a80ba1b7e7845f2982. Option 3 seems the most attractive if it could work.

[thomashoneyman]

Option 3 is indeed attractive and I’m perfectly happy experimenting with reworking try PureScript, it’s definitely not received a lot of love recently.

I wouldn’t dismiss option 2 offhand — Try PureScript is browser based so certainly node packages, etc. can safely be excluded

[pete-murphy]

I wouldn’t dismiss option 2 offhand — Try PureScript is browser based so certainly node packages, etc. can safely be excluded

I did end up looking into Option 2 after all since it seems much easier and I am less likely to get to the harder Option 3 in near term. If I trim out Node packages & some large packages that have no dependents, I am able to run the server with (set -o noglob && stack exec trypurescript -- +RTS -N2 -A128m -M3G -s -RTS 8081 $(spago sources)) without exceeding that max heap size. The result was this commit: 0fd79d9.

The non-Node packages (largest packages with no dependents) that were omitted:

• framer-motion
• motion
• next-purs-rsc
• react-basic-dom-beta
• react-icons
• css-frameworks
• lumi-components
• yoga-sqlite
• rito

#4269d0 are packages that were kept
#e0820e is omitted (because Node)
#c33d3d is omitted (because big package with no dependents)

Here's a notebook for exploring what packages to omit if it's useful: https://new.observablehq.com/@pete-murphy/try-pure-script-package-set-trim-explorer (mostly made by Opus 4.8)

The "no dependents" criterion is arbitrary, open to ideas if you can think of something that would be more fair? (Age of package, some other popularity metric...)