Merge pull request 'checks/words: a retired word's plural is the word (hq ADR 0244)' (#112) from fix/words-match-plurals into main

checks/words/main.go

@@ -4,7 +4,8 @@
 // own code that is not a test (a tool's description, the notes and errors it answers with) and every
 // description in its manifest. Comments are not read: an agent never sees them.
 //
-// A word is matched whole and in any case, a space in it matching any run of white space. A vendor word is
+// A word is matched whole and in any case, a space in it matching any run of white space, and its plural
+// (`s` or `es` on its last part) is matched as the word. A vendor word is
 // never on the list (the glossary does not retire one for the tools), so a wrapped program's own objects —
 // an identity provider's users, a media manager's releases — are never findings.
 //
@@ -123,7 +124,9 @@ func pattern(word string) *regexp.Regexp {
 	for i, p := range parts {
 		parts[i] = regexp.QuoteMeta(p)
 	}
-	return regexp.MustCompile(`(?i)(?:^|[^\w-])(` + strings.Join(parts, `\s+`) + `)(?:$|[^\w-])`)
+	// The plural is the word too: `s` or `es` on its last part, so "control planes" is found as "control
+	// plane" is. Anything else joined on is another word.
+	return regexp.MustCompile(`(?i)(?:^|[^\w-])(` + strings.Join(parts, `\s+`) + `(?:e?s)?)(?:$|[^\w-])`)
 }
 
 func check(words []string, texts []text) []string {

checks/words/main_test.go

@@ -0,0 +1,26 @@
+package main
+
+import "testing"
+
+// A retired word's plural is the word; anything else joined on is another word.
+func TestPatternMatchesTheWordAndItsPlural(t *testing.T) {
+	cases := []struct {
+		word, text string
+		want       bool
+	}{
+		{"control plane", "ask the control plane", true},
+		{"control plane", "two control planes", true},
+		{"control plane", "two control\n\tplanes", true},
+		{"substrate", "the substrates", true},
+		{"flavor", "flavors of a module", true},
+		{"flavor", "flavored", false},
+		{"node tools", "node toolsets", false},
+		{"mesh-console", "the mesh-consoles", true},
+		{"mesh-console", "mesh-console-x", false},
+	}
+	for _, c := range cases {
+		if got := pattern(c.word).MatchString(c.text); got != c.want {
+			t.Errorf("pattern(%q) on %q = %v, want %v", c.word, c.text, got, c.want)
+		}
+	}
+}

merge-check.sh

@@ -18,10 +18,10 @@
 #   3. the words (novox/hq ADR 0244): no word the glossary retired for the tools, as copied in
 #      retired-words, in any text a module's tools can show an agent — every string in its own code that is
 #      not a test, and every description in its manifest. Run over every module, not only the touched ones,
-#      so a word newly retired is found everywhere it stands.
+#      so a word newly retired is found everywhere it stands. The check's own test runs first.
 set -eu
 
-(cd checks/words && go run . ../..)
+(cd checks/words && go test -count=1 . && go run . ../..)
 
 if [ -n "${MESH_GATE:-}" ]; then
 	checked=$("$MESH_GATE" module check modules/*/module.json) || { printf '%s\n' "$checked"; exit 1; }