docs: allow approved third-party packages #59
@@ -3,6 +3,7 @@
|
||||
# Exit 1 if any unapproved import is found.
|
||||
#
|
||||
# The allowlist is parsed from CONVENTIONS.md to maintain a single source of truth.
|
||||
# Also enforces Scope column: "test only" packages cannot appear in non-test code.
|
||||
|
||||
set -euo pipefail
|
||||
|
||||
@@ -13,22 +14,53 @@ if [ ! -f "$CONVENTIONS_FILE" ]; then
|
||||
exit 1
|
||||
fi
|
||||
|
||||
# Parse approved packages from CONVENTIONS.md table
|
||||
# Looks for lines like: | `gopkg.in/yaml.v3` | ...
|
||||
ALLOWED=()
|
||||
while IFS= read -r line; do
|
||||
# Extract package from markdown table cell: | `package` |
|
||||
pkg=$(echo "$line" | grep -oP '\| `\K[^`]+' | head -1 || true)
|
||||
if [ -n "$pkg" ] && [[ "$pkg" != "Package" ]]; then
|
||||
ALLOWED+=("$pkg")
|
||||
fi
|
||||
done < <(grep -E '^\| `[a-zA-Z]' "$CONVENTIONS_FILE" || true)
|
||||
# Parse approved packages from CONVENTIONS.md table using awk (POSIX-compatible)
|
||||
# Format: | `package` | use case | scope |
|
||||
# Output: package:scope (e.g., "gopkg.in/yaml.v3:production")
|
||||
declare -A ALLOWED_PROD=()
|
||||
declare -A ALLOWED_TEST=()
|
||||
|
||||
if [ ${#ALLOWED[@]} -eq 0 ]; then
|
||||
while IFS= read -r line; do
|
||||
|
|
||||
# Use awk to extract package and scope from table row
|
||||
# Split on | and extract backtick-wrapped package from column 1, scope from column 3
|
||||
pkg=$(echo "$line" | awk -F'|' '{gsub(/^[[:space:]]*`|`[[:space:]]*$/, "", $2); print $2}')
|
||||
scope=$(echo "$line" | awk -F'|' '{gsub(/^[[:space:]]+|[[:space:]]+$/, "", $4); print tolower($4)}')
|
||||
|
||||
if [ -n "$pkg" ] && [ "$pkg" != "Package" ] && [[ "$pkg" =~ ^[a-zA-Z] ]]; then
|
||||
if [[ "$scope" == *"test"* ]]; then
|
||||
ALLOWED_TEST["$pkg"]=1
|
||||
else
|
||||
|
gpt-review-bot
commented
[NIT] Parsing the markdown table with **[NIT]** Parsing the markdown table with `grep`/`awk` is somewhat brittle (e.g., relies on backticks around package names and exact column positions). This is acceptable given the documented process, but a brief note in CONVENTIONS.md to preserve formatting would help avoid accidental breakage.
|
||||
ALLOWED_PROD["$pkg"]=1
|
||||
fi
|
||||
fi
|
||||
|
gpt-review-bot
commented
[MINOR] The filter **[MINOR]** The filter `[[ "$pkg" =~ ^[a-zA-Z] ]]` rejects valid import paths that begin with a digit (e.g., 9fans.net/go). Consider relaxing to `^[[:alnum:]]` or removing the check, since the header row is already excluded by the grep.
|
||||
done < <(grep '| `' "$CONVENTIONS_FILE" 2>/dev/null || true)
|
||||
|
||||
ALL_ALLOWED=("${!ALLOWED_PROD[@]}" "${!ALLOWED_TEST[@]}")
|
||||
|
||||
if [ ${#ALL_ALLOWED[@]} -eq 0 ]; then
|
||||
echo "⚠️ No approved packages found in $CONVENTIONS_FILE"
|
||||
echo " (This is fine if you want stdlib-only)"
|
||||
fi
|
||||
|
||||
# Helper: check if import matches any package in an associative array (literal prefix, no glob)
|
||||
matches_allowlist() {
|
||||
local import="$1"
|
||||
shift
|
||||
local -n allowlist=$1
|
||||
|
||||
for allowed in "${!allowlist[@]}"; do
|
||||
# Exact match
|
||||
if [ "$import" = "$allowed" ]; then
|
||||
return 0
|
||||
fi
|
||||
# Literal prefix match for subpackages (no glob interpretation)
|
||||
if [ "${import#"$allowed/"}" != "$import" ]; then
|
||||
|
sonnet-review-bot
commented
[NIT] The comment says 'POSIX-compatible' for the awk parsing but the outer loop uses a Bash process substitution **[NIT]** The comment says 'POSIX-compatible' for the awk parsing but the outer loop uses a Bash process substitution `< <(...)` which is Bash-specific. The comment is mildly misleading — it means awk itself is POSIX, not the overall approach.
|
||||
return 0
|
||||
fi
|
||||
done
|
||||
return 1
|
||||
}
|
||||
|
||||
# Get DIRECT dependencies only (exclude indirect/transitive)
|
||||
# Fail closed: if go list fails, we exit non-zero
|
||||
IMPORTS=$(go list -m -f '{{if not .Indirect}}{{.Path}}{{end}}' all 2>&1) || {
|
||||
@@ -44,21 +76,13 @@ if [ -z "$IMPORTS" ]; then
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Check all direct dependencies are in the allowlist
|
||||
VIOLATIONS=""
|
||||
while IFS= read -r import; do
|
||||
[ -z "$import" ] && continue
|
||||
|
||||
# Check if import matches any allowed package (prefix match for subpackages)
|
||||
MATCHED=false
|
||||
for allowed in "${ALLOWED[@]}"; do
|
||||
if [[ "$import" == "$allowed" ]] || [[ "$import" == "$allowed/"* ]]; then
|
||||
MATCHED=true
|
||||
break
|
||||
fi
|
||||
done
|
||||
|
||||
if [ "$MATCHED" = false ]; then
|
||||
VIOLATIONS="${VIOLATIONS} - ${import}"$'\n'
|
||||
if ! matches_allowlist "$import" ALLOWED_PROD && ! matches_allowlist "$import" ALLOWED_TEST; then
|
||||
VIOLATIONS="${VIOLATIONS} - ${import} (not in allowlist)"$'\n'
|
||||
fi
|
||||
done <<< "$IMPORTS"
|
||||
|
||||
@@ -68,17 +92,31 @@ if [ -n "$VIOLATIONS" ]; then
|
||||
echo "The following imports are not in the allowlist:"
|
||||
printf "%s" "$VIOLATIONS"
|
||||
echo ""
|
||||
echo "Approved packages (from CONVENTIONS.md):"
|
||||
for pkg in "${ALLOWED[@]}"; do
|
||||
echo " - $pkg"
|
||||
done
|
||||
echo "To add a dependency, update CONVENTIONS.md (requires Aaron's approval)"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
# Enforce Scope: test-only packages must not appear in non-test code
|
||||
|
sonnet-review-bot
commented
[MINOR] The test-scope enforcement uses **[MINOR]** The test-scope enforcement uses `go list -deps -f '...' ./... | grep -v '_test'` to identify production imports, but this approach is fragile: it filters by path containing '_test', which catches test packages but the `-deps` flag on non-test packages already excludes test files. More importantly, this will fail to detect a test-only package imported via a non-test file in a sub-package whose path happens not to contain '_test'. Using `go list -f '{{if not .Standard}}{{.ImportPath}}{{end}}' $(go list ./...)` (without `-deps`) and then checking against the allowlist directly would be more precise. That said, the current approach is a reasonable heuristic.
gpt-review-bot
commented
[NIT] The **[NIT]** The `grep -v '_test'` in the production imports step is unnecessary because `go list -deps` without `-test` already excludes test-only dependencies. While harmless, it can be removed for clarity.
|
||||
# Get imports used by non-test code only
|
||||
PROD_IMPORTS=$(go list -deps -f '{{if not .Standard}}{{.ImportPath}}{{end}}' ./... 2>/dev/null | grep -v '_test' || true)
|
||||
|
[MINOR] Production imports discovery allows go list to fail open (2>/dev/null | ... || true). If go list fails, PROD_IMPORTS becomes empty and the script won’t detect misuse of test-only dependencies, weakening enforcement. Failing closed here would improve robustness. **[MINOR]** Production imports discovery allows go list to fail open (2>/dev/null | ... || true). If go list fails, PROD_IMPORTS becomes empty and the script won’t detect misuse of test-only dependencies, weakening enforcement. Failing closed here would improve robustness.
|
||||
|
||||
TEST_ONLY_IN_PROD=""
|
||||
|
gpt-review-bot
commented
[MINOR] The scope enforcement uses **[MINOR]** The scope enforcement uses `grep -q "^${test_pkg}"` which can produce false positives for modules with a similar prefix (e.g., github.com/google/go-cmppro would match github.com/google/go-cmp). Use a delimiter-aware pattern like `^${test_pkg}(/|$)` to avoid accidental matches.
|
||||
for test_pkg in "${!ALLOWED_TEST[@]}"; do
|
||||
|
sonnet-review-bot
commented
[MINOR] The regex **[MINOR]** The regex `"^${test_pkg}(/|\$|$)"` has a redundant `$` — `\$` escapes the literal dollar sign for the shell but in the regex context inside `grep -qE`, the trailing `|$)` is a proper end-of-line anchor followed by the closing group, making `\$` (literal `$` character) unreachable in practice. This is harmless but the intent (match exact package or subpackage) is already handled by the `^pkg($|/)` pattern. Consider simplifying to `"^${test_pkg}(/|$)"`.
|
||||
if echo "$PROD_IMPORTS" | grep -q "^${test_pkg}"; then
|
||||
TEST_ONLY_IN_PROD="${TEST_ONLY_IN_PROD} - ${test_pkg} (marked 'test only' but used in production code)"$'\n'
|
||||
fi
|
||||
done
|
||||
|
sonnet-review-bot
commented
[MINOR] The test-only production check uses **[MINOR]** The test-only production check uses `grep -q "^${test_pkg}"` without anchoring the end of the pattern. A package path like `github.com/google/go-cmp` would inadvertently match a hypothetical `github.com/google/go-cmp-extended`. Should be `grep -qE "^${test_pkg}(/|$)"` to match only exact or subpackage imports.
|
||||
|
||||
|
gpt-review-bot
commented
[MINOR] The grep regex **[MINOR]** The grep regex `^${test_pkg}(/|\$|$)` redundantly includes both `\$` and `$`. It can be simplified to `^${test_pkg}(/|$)` without changing semantics.
|
||||
if [ -n "$TEST_ONLY_IN_PROD" ]; then
|
||||
echo "❌ TEST-ONLY DEPENDENCIES IN PRODUCTION CODE"
|
||||
echo ""
|
||||
echo "To add a dependency:"
|
||||
echo " 1. Open a PR that ONLY updates CONVENTIONS.md"
|
||||
echo " 2. Get explicit approval from Aaron"
|
||||
echo " 3. After merge, use the package in a separate PR"
|
||||
printf "%s" "$TEST_ONLY_IN_PROD"
|
||||
echo ""
|
||||
echo "These packages are marked 'test only' in CONVENTIONS.md"
|
||||
echo "and must only be imported from *_test.go files."
|
||||
exit 1
|
||||
fi
|
||||
|
||||
echo "✅ All dependencies are approved"
|
||||
echo " Direct deps: $(echo "$IMPORTS" | wc -l | tr -d ' ')"
|
||||
echo " Production: ${#ALLOWED_PROD[@]}, Test-only: ${#ALLOWED_TEST[@]}"
|
||||
|
||||
[MINOR] Uses 'grep -P' (Perl regex), which is not available on macOS/BSD grep by default. This reduces cross-platform developer usability for the precommit hook.