Improve the performance of the formatter instability check job (#14471)

We should probably get rid of this entirely and subsume it's
functionality in the normal ecosystem checks? I don't think we're using
the black comparison tests anymore, but maybe someone wants it?

There are a few major parts to this:

1. Making the formatter script idempotent, so it can be run repeatedly
and is robust to changing commits
2. Reducing the overhead of the git operations, minimizing the data
transfer
3. Parallelizing all the git operations by repository

This reduces the setup time from 80s to 16s (locally).

The initial motivation for idempotency was to include the repositories
in the GitHub Actions cache. I'm not sure it's worth it yet — they're
about 1GB and would consume our limited cache space. Regardless, it
improves correctness for local invocations.

The total runtime of the job is reduced from ~4m to ~3m.

I also made some cosmetic changes to the output paths and such.

.github/workflows/ci.yaml

@@ -563,12 +563,12 @@ jobs:
         run: rustup show
       - name: "Cache rust"
         uses: Swatinem/rust-cache@v2
-      - name: "Formatter progress"
+      - name: "Run checks"
         run: scripts/formatter_ecosystem_checks.sh
       - name: "Github step summary"
-        run: cat target/progress_projects_stats.txt > $GITHUB_STEP_SUMMARY
+        run: cat target/formatter-ecosystem/stats.txt > $GITHUB_STEP_SUMMARY
       - name: "Remove checkouts from cache"
-        run: rm -r target/progress_projects
+        run: rm -r target/formatter-ecosystem
 
   check-ruff-lsp:
     name: "test ruff-lsp"

scripts/formatter_ecosystem_checks.sh

@@ -8,7 +8,7 @@
 # errors.
 #
 # This script will first clone a diverse set of (mostly) black formatted
-# repositories with fixed revisions to target/progress_projects. Each project
+# repositories with fixed revisions to target/formatter-ecosystem. Each project
 # gets formatted (without modifying the files on disk) to check how
 # similar our style is to black. It also catches common issues such as
 # unstable formatting, internal formatter errors and printing invalid syntax.
@@ -18,72 +18,103 @@
 set -e
 
 target=$(git rev-parse --show-toplevel)/target
-dir="$target/progress_projects"
+dir="$target/formatter-ecosystem"
 mkdir -p "$dir"
 
+# Perform an idempotent clone and checkout of a commit
+clone_commit() {
+    local repo="$1"
+    local name="$2"
+    local ref="$3"
+
+    if [ -z "$repo" ] || [ -z "$name" ] || [ -z "$ref" ]; then
+        echo "Usage: clone_commit <repo> <name> <ref>"
+        return 1
+    fi
+
+    local target="$dir/projects/$name"
+
+    if [ ! -d "$target/.git" ]; then
+        echo "Cloning $repo to $name"
+        # Perform a minimal clone, we only need a single commit
+        git clone --filter=blob:none --depth=1 --no-tags --no-checkout --single-branch "$repo" "$target"
+    fi
+
+    echo "Using $repo at $ref"
+    git -C "$target" fetch --filter=blob:none --depth=1 --no-tags origin "$ref"
+    git -C "$target" checkout -q "$ref"
+}
+
 # small util library
-if [ ! -d "$dir/twine/.git" ]; then
-  git clone --filter=tree:0 https://github.com/pypa/twine "$dir/twine"
-fi
-git -C "$dir/twine" checkout -q ae71822a3cb0478d0f6a0cccb65d6f8e6275ece5
+clone_commit \
+    "https://github.com/pypa/twine" \
+    "twine" \
+    "ae71822a3cb0478d0f6a0cccb65d6f8e6275ece5" &
 
 # web framework that implements a lot of magic
-if [ ! -d "$dir/django/.git" ]; then
-  git clone --filter=tree:0 https://github.com/django/django "$dir/django"
-fi
-git -C "$dir/django" checkout -q ee5147cfd7de2add74a285537a8968ec074e70cd
+clone_commit \
+    "https://github.com/django/django" \
+    "django" \
+    "ee5147cfd7de2add74a285537a8968ec074e70cd" &
 
 # an ML project
-if [ ! -d "$dir/transformers/.git" ]; then
-  git clone --filter=tree:0 https://github.com/huggingface/transformers "$dir/transformers"
-fi
-git -C "$dir/transformers" checkout -q ac5a0556f14dec503b064d5802da1092e0b558ea
+clone_commit \
+    "https://github.com/huggingface/transformers" \
+    "transformers" \
+    "ac5a0556f14dec503b064d5802da1092e0b558ea" &
 
 # type annotations
-if [ ! -d "$dir/typeshed/.git" ]; then
-  git clone --filter=tree:0 https://github.com/python/typeshed "$dir/typeshed"
-fi
-git -C "$dir/typeshed" checkout -q d34ef50754de993d01630883dbcd1d27ba507143
+clone_commit \
+    "https://github.com/python/typeshed" \
+    "typeshed" \
+    "d34ef50754de993d01630883dbcd1d27ba507143" &
 
 # python 3.11, typing and 100% test coverage
-if [ ! -d "$dir/warehouse/.git" ]; then
-  git clone --filter=tree:0 https://github.com/pypi/warehouse "$dir/warehouse"
-fi
-git -C "$dir/warehouse" checkout -q 5a4d2cadec641b5d6a6847d0127940e0f532f184
+clone_commit \
+    "https://github.com/pypi/warehouse" \
+    "warehouse" \
+    "5a4d2cadec641b5d6a6847d0127940e0f532f184" &
 
 # zulip, a django user
-if [ ! -d "$dir/zulip/.git" ]; then
-  git clone --filter=tree:0 https://github.com/zulip/zulip "$dir/zulip"
-fi
-git -C "$dir/zulip" checkout -q ccddbba7a3074283ccaac3bde35fd32b19faf042
+clone_commit \
+    "https://github.com/zulip/zulip" \
+    "zulip" \
+    "ccddbba7a3074283ccaac3bde35fd32b19faf042" &
 
 # home-assistant, home automation with 1ok files
-if [ ! -d "$dir/home-assistant/.git" ]; then
-  git clone --filter=tree:0 https://github.com/home-assistant/core "$dir/home-assistant"
-fi
-git -C "$dir/home-assistant" checkout -q 3601c531f400255d10b82529549e564fbe483a54
+clone_commit \
+    "https://github.com/home-assistant/core" \
+    "home-assistant" \
+    "3601c531f400255d10b82529549e564fbe483a54" &
 
 # poetry, a package manager that uses black preview style
-if [ ! -d "$dir/poetry/.git" ]; then
-  git clone --filter=tree:0 https://github.com/python-poetry/poetry "$dir/poetry"
-fi
-git -C "$dir/poetry" checkout -q 36fedb59b8e655252168055b536ead591068e1e4
+clone_commit \
+    "https://github.com/python-poetry/poetry" \
+    "poetry" \
+    "36fedb59b8e655252168055b536ead591068e1e4" &
 
 # cpython itself
-if [ ! -d "$dir/cpython/.git" ]; then
-  git clone --filter=tree:0 https://github.com/python/cpython "$dir/cpython"
-fi
-git -C "$dir/cpython" checkout -q 28aea5d07d163105b42acd81c1651397ef95ea57
+clone_commit \
+    "https://github.com/python/cpython" \
+    "cpython" \
+    "28aea5d07d163105b42acd81c1651397ef95ea57" &
+
+# wait for the concurrent clones to complete
+wait
 
 # Uncomment if you want to update the hashes
 #for i in "$dir"/*/; do git -C "$i" switch main && git -C "$i" pull; done
 #for i in "$dir"/*/; do echo "# $(basename "$i") $(git -C "$i" rev-parse HEAD)"; done
 
 time cargo run --bin ruff_dev -- format-dev --stability-check \
-  --error-file "$target/progress_projects_errors.txt" --log-file "$target/progress_projects_log.txt" --stats-file "$target/progress_projects_stats.txt" \
-  --files-with-errors 3 --multi-project "$dir" || (
+  --error-file "$dir/errors.txt" \
+  --log-file "$dir/log.txt" \
+  --stats-file "$dir/stats.txt" \
+  --files-with-errors 3 --multi-project "$dir/projects" \
+|| (
   echo "Ecosystem check failed"
-  cat "$target/progress_projects_log.txt"
+  cat "$dir/log.txt"
   exit 1
 )
-cat "$target/progress_projects_stats.txt"
+
+cat "$dir/stats.txt"