[bugfix] Fix incorrect json-ld @context serialization (#3243)

vendor/github.com/superseriousbusiness/activity/streams/util.go

@@ -1,6 +1,9 @@
 package streams
 
 import (
+	"maps"
+	"slices"
+
 	"github.com/superseriousbusiness/activity/streams/vocab"
 )
 
@@ -10,47 +13,200 @@ const (
 	// rest of the payload. Important for linked-data representations, but
 	// only applicable to go-fed at code-generation time.
 	jsonLDContext = "@context"
+
+	asNS     = "https://www.w3.org/ns/activitystreams"
+	tootNS   = "http://joinmastodon.org/ns"
+	schemaNS = "http://schema.org"
 )
 
+// Map of inlines @context entries that may need to be added
+// when vocabs include "https://www.w3.org/ns/activitystreams".
+var asInlines = map[string]any{
+	"Hashtag":                   "as:Hashtag",
+	"alsoKnownAs":               "as:alsoKnownAs",
+	"manuallyApprovesFollowers": "as:manuallyApprovesFollowers",
+	"sensitive":                 "as:sensitive",
+
+	"movedTo": map[string]string{
+		"@id":   "as:movedTo",
+		"@type": "@id",
+	},
+}
+
+// Map of inlines @context entries that may need to be
+// added when vocabs include "http://joinmastodon.org/ns".
+var tootInlines = map[string]any{
+	"Emoji":        "toot:Emoji",
+	"blurhash":     "toot:blurhash",
+	"discoverable": "toot:discoverable",
+	"indexable":    "toot:indexable",
+	"memorial":     "toot:memorial",
+	"suspended":    "toot:suspended",
+	"votersCount":  "toot:votersCount",
+
+	"featured": map[string]string{
+		"@id":   "toot:featured",
+		"@type": "@id",
+	},
+
+	"featuredTags": map[string]string{
+		"@id":   "toot:featuredTags",
+		"@type": "@id",
+	},
+
+	"focalPoint": map[string]string{
+		"@container": "@list",
+		"@id":        "toot:focalPoint",
+	},
+}
+
+// Map of inlines @context entries that may need to
+// be added when vocabs include "http://schema.org".
+var schemaInlines = map[string]any{
+	"PropertyValue": "schema:PropertyValue",
+	"value":         "schema:value",
+}
+
+// getLookup returns a lookup map of all interesting field names
+// + type names on the given "in" map that may need to be inlined.
+func getLookup(in map[string]any) map[string]struct{} {
+	out := make(map[string]struct{})
+
+	for k, v := range in {
+		// Pull out keys from any nested maps.
+		if nested, ok := v.(map[string]any); ok {
+			maps.Copy(out, getLookup(nested))
+			continue
+		}
+
+		// Pull out keys from any
+		// arrays of nested maps.
+		if nestedIs, ok := v.([]any); ok {
+			for _, nestedI := range nestedIs {
+				if nested, ok := nestedI.(map[string]any); ok {
+					maps.Copy(out, getLookup(nested))
+					continue
+				}
+			}
+		}
+
+		// For types, we actually care about
+		// the *value*, ie., the name of the
+		// type, not the type key itself.
+		if k == "type" {
+			out[v.(string)] = struct{}{}
+			continue
+		}
+
+		out[k] = struct{}{}
+	}
+
+	return out
+}
+
+func copyInlines(
+	src map[string]any,
+	dst map[string]any,
+	lookup map[string]struct{},
+) {
+	for k, v := range src {
+		_, ok := lookup[k]
+		if ok {
+			dst[k] = v
+		}
+	}
+}
+
 // Serialize adds the context vocabularies contained within the type
 // into the JSON-LD @context field, and aliases them appropriately.
-func Serialize(a vocab.Type) (m map[string]interface{}, e error) {
+func Serialize(a vocab.Type) (m map[string]any, e error) {
 	m, e = a.Serialize()
 	if e != nil {
 		return
 	}
-	v := a.JSONLDContext()
-	// Transform the map of vocabulary-to-aliases into a context payload,
-	// but do so in a way that at least keeps it readable for other humans.
-	var contextValue interface{}
-	if len(v) == 1 {
-		for vocab, alias := range v {
-			if len(alias) == 0 {
-				contextValue = vocab
-			} else {
-				contextValue = map[string]string{
-					alias: vocab,
-				}
-			}
+
+	var (
+		// Slice of vocab URIs
+		// used in this vocab.Type.
+		vocabs = a.JSONLDContext()
+
+		// Slice of vocab URIs to add
+		// to the base @context slice.
+		includeVocabs []string
+
+		// Object to inline as an extra
+		// entry in the @context slice.
+		inlinedContext = make(map[string]any)
+	)
+
+	// Get a lookup of all field and
+	// type names we need to care about.
+	lookup := getLookup(m)
+
+	// Go through each used vocab and see
+	// if we need to special case it.
+	for vocab := range vocabs {
+
+		switch vocab {
+
+		case asNS:
+			// ActivityStreams vocab.
+			//
+			// The namespace URI already points to
+			// a proper @context document but we
+			// need to add some extra inlines.
+			includeVocabs = append(includeVocabs, asNS)
+			copyInlines(asInlines, inlinedContext, lookup)
+
+		case schemaNS:
+			// Schema vocab.
+			//
+			// The URI doesn't point to a @context
+			// document so we need to inline everything.
+			inlinedContext["schema"] = schemaNS + "#"
+			copyInlines(schemaInlines, inlinedContext, lookup)
+
+		case tootNS:
+			// Toot/Mastodon vocab.
+			//
+			// The URI doesn't point to a @context
+			// document so we need to inline everything.
+			inlinedContext["toot"] = tootNS + "#"
+			copyInlines(tootInlines, inlinedContext, lookup)
+
+		default:
+			// No special case.
+			includeVocabs = append(includeVocabs, vocab)
 		}
-	} else {
-		var arr []interface{}
-		aliases := make(map[string]string)
-		for vocab, alias := range v {
-			if len(alias) == 0 {
-				arr = append(arr, vocab)
-			} else {
-				aliases[alias] = vocab
-			}
-		}
-		if len(aliases) > 0 {
-			arr = append(arr, aliases)
-		}
-		contextValue = arr
 	}
-	// TODO: Update the context instead if it already exists
-	m[jsonLDContext] = contextValue
-	// TODO: Sort the context based on arbitrary order.
+
+	// Sort used vocab entries alphabetically
+	// to make their ordering predictable.
+	slices.Sort(includeVocabs)
+
+	// Create final slice of @context
+	// entries we'll need to include.
+	contextEntries := make([]any, 0, len(includeVocabs)+1)
+
+	// Append each included vocab to the slice.
+	for _, vocab := range includeVocabs {
+		contextEntries = append(contextEntries, vocab)
+	}
+
+	// Append any inlinedContext to the slice.
+	if len(inlinedContext) != 0 {
+		contextEntries = append(contextEntries, inlinedContext)
+	}
+
+	// Include @context on the final output,
+	// using an array if there's more than
+	// one entry, just a property otherwise.
+	if len(contextEntries) != 1 {
+		m[jsonLDContext] = contextEntries
+	} else {
+		m[jsonLDContext] = contextEntries[0]
+	}
+
 	// Delete any existing `@context` in child maps.
 	var cleanFnRecur func(map[string]interface{})
 	cleanFnRecur = func(r map[string]interface{}) {