Drastically reduced NioPool default cache size and saved some heap.

cdstar-backend-nio/src/main/java/de/gwdg/cdstar/pool/nio/NioPool.java

@@ -64,7 +64,7 @@ public class NioPool implements StoragePool {
 	public static final int SHARD_DEPTH = 2;
 	private static JsonFormat objectLoader = new JsonFormat();
 
-	private int cacheSize = 1024;
+	private int cacheSize = 16;
 
 	// Running transactions
 	ConcurrentHashMap<String, NioSession> scopes = new ConcurrentHashMap<>();

cdstar-backend-nio/src/main/java/de/gwdg/cdstar/pool/nio/json/JsonFormat.java

@@ -120,16 +120,15 @@ public class JsonFormat {
 			case ("resources"):
 				assertNextToken(p, JsonToken.START_ARRAY);
 				doc.resources = new ArrayList<>();
-				while (p.nextToken() == JsonToken.START_OBJECT) {
+				while (p.nextToken() == JsonToken.START_OBJECT)
 					doc.resources.add(readResourceEntry(p));
-				}
 				assertCurrentToken(p, JsonToken.END_ARRAY);
 				break;
 			default:
 				if (fieldName.startsWith("x-")) {
 					if (doc.attr == null)
 						doc.attr = new HashMap<>(4);
-					doc.attr.put(fieldName.substring(2).intern(), readText(p));
+					doc.attr.put(fieldName.substring(2), readText(p));
 				} else {
 					throw new FormatError("Unknown property: " + fieldName, p);
 				}
@@ -150,9 +149,47 @@ public class JsonFormat {
 
 		assertCurrentToken(p, JsonToken.END_OBJECT);
 
+		compress(doc);
+
 		return doc;
 	}
 
+	/**
+	 * Try to reduce document size by deduplicating certain strings.
+	 */
+	private void compress(JsonIndex doc) {
+		// Resources often share the same 'type' and 'enc' values. Frozen objects likely
+		// have the same 'src' value for all resources. To speed things up and not
+		// produce more garbage than we save, we only compare consecutive resource here.
+
+		String lastType = null;
+		String lastSrc = null;
+		String lastEnc = null;
+		for (final JsonResource res : doc.resources) {
+
+			if (res.type != null) {
+				if (res.type.equals(lastType))
+					res.type = lastType;
+				else
+					lastType = res.type;
+			}
+
+			if (res.src != null) {
+				if (res.src.equals(lastSrc))
+					res.src = lastSrc;
+				else
+					lastSrc = res.src;
+			}
+
+			if (res.enc != null) {
+				if (res.enc.equals(lastSrc))
+					res.enc = lastEnc;
+				else
+					lastEnc = res.enc;
+			}
+		}
+	}
+
 	private JsonResource readResourceEntry(JsonParser p) throws IOException, FormatError {
 		final JsonResource jr = new JsonResource();
 		jr.size = -1;
@@ -195,7 +232,7 @@ public class JsonFormat {
 				if (fieldName.startsWith("x-")) {
 					if (jr.attr == null)
 						jr.attr = new HashMap<>(4);
-					jr.attr.put(fieldName.substring(2).intern(), readText(p));
+					jr.attr.put(fieldName.substring(2), readText(p));
 				} else {
 					throw new FormatError("Unknown property: " + fieldName, p);
 				}
@@ -225,7 +262,8 @@ public class JsonFormat {
 	}
 
 	private byte[] readBytes(JsonParser p) throws IOException, FormatError {
-		return Utils.base64decode(readText(p));
+		assertNextToken(p, JsonToken.VALUE_STRING);
+		return p.getBinaryValue();
 	}
 
 	/*
@@ -400,7 +438,7 @@ public class JsonFormat {
 			if (src.startsWith("external:")) {
 				res.put("src", "x-" + src.substring("external:".length()));
 			} else if (src.endsWith(".bin")
-					&& src.equals(Utils.bytesToHex(Utils.base64decode(res.get("sha256").asText())) + ".bin")) {
+				&& src.equals(Utils.bytesToHex(Utils.base64decode(res.get("sha256").asText())) + ".bin")) {
 				res.remove("src");
 			} else {
 				throw new FormatError("Unrecognized value for 'resource.src' in version 3 format: " + Utils.repr(src));

cdstar-backend-nio/src/main/java/de/gwdg/cdstar/pool/nio/json/JsonResource.java

@@ -9,9 +9,8 @@ import com.fasterxml.jackson.annotation.JsonInclude.Include;
 @JsonInclude(Include.NON_NULL)
 public final class JsonResource {
 	public String id; // Unique tracking ID that never changes
-	public String src; // Source filename (relative to object path), null, or
-						// prefixed URI
-	public String name; // Given name of this file
+	public String src; // Null or prefixed URI
+	public String name; // Given name for this file
 	public String type; // Content-Type compatible mime type
 	public String enc; // Content-Encoding compatible encoding name. May be
 						// null if no encoding is applied.

cdstar-runtime/src/main/java/de/gwdg/cdstar/runtime/client/FileImpl.java

@@ -20,6 +20,7 @@ import de.gwdg.cdstar.runtime.client.exc.InvalidFileName;
 
 class FileImpl implements CDStarFile {
 
+	private static final String DEFAULT_MEDIA_TYPE = "application/octet-stream";
 	static final String RESOURCE_PREFIX = "data/";
 	final ArchiveImpl archive;
 	final Resource resource;
@@ -139,7 +140,7 @@ class FileImpl implements CDStarFile {
 	public String getMediaType() {
 		final String type = resource.getMediaType();
 		if (type == null)
-			return "application/octet-stream";
+			return DEFAULT_MEDIA_TYPE;
 		return type;
 	}
 
@@ -155,6 +156,8 @@ class FileImpl implements CDStarFile {
 		archive.checkPermission(ArchivePermission.CHANGE_FILES);
 		if (Utils.equal(getMediaType(), type) && Utils.equal(getContentEncoding(), coding))
 			return;
+		if (Utils.equal(type, DEFAULT_MEDIA_TYPE))
+			type = null; // Do not store the obvious
 		resource.setMediaType(type, coding);
 		archive.markContentModified();
 	}