Merge pull request #37 from prdoyle/bson

Add bsok-bson library

Author: prdoyle

bosk-bson/README.md

@@ -0,0 +1,7 @@
+## bosk-bson
+
+This is the subproject for the published `bosk-bson` library,
+containing facilities for manipulating MongoDB's binary JSON data structures.
+Outside of MongoDB itself, we also use BSON as a handy utility
+for divide up and re-combine large JSON documents,
+which is useful even outside of MongoDB.

bosk-bson/build.gradle

@@ -0,0 +1,47 @@
+
+plugins {
+    id 'bosk.development'
+    id 'bosk.maven-publish'
+    id 'info.solidsoft.pitest' version '1.15.0'
+    id 'com.github.spotbugs' version '5.1.5'
+}
+
+base {
+    archivesName = 'bosk-bson'
+}
+
+java {
+    toolchain {
+        languageVersion = JavaLanguageVersion.of(jdkVersion)
+    }
+}
+
+compileJava {
+    options.release = prodJavaVersion
+}
+
+compileTestJava {
+    options.release = null
+}
+
+dependencies {
+    api project(":bosk-core")
+    api 'org.mongodb:bson:5.1.2'
+    implementation 'com.github.spotbugs:spotbugs-annotations:4.8.6' // To stop warnings about When from MongoDB driver
+
+    testImplementation project(":bosk-logback")
+    testImplementation project(":bosk-testing")
+    testImplementation project(":lib-testing")
+}
+
+pitest {
+    pitestVersion = '1.15.0'
+    junit5PluginVersion = '1.2.0'
+    jvmArgs = ['-ea'] // Our unit tests check for assert statements
+    targetClasses = ['works.bosk.drivers.mongo.BsonSurgeon']
+    targetTests   = ['works.bosk.drivers.mongo.BsonSurgeonTest']
+    threads = 4
+    outputFormats = ['XML', 'HTML']
+    timestampedReports = false
+    //verbose = true
+}

bosk-mongo/src/main/java/works/bosk/drivers/mongo/BsonFormatException.java renamed to bosk-bson/src/main/java/works/bosk/bson/BsonFormatException.java

@@ -1,4 +1,4 @@
-package works.bosk.drivers.mongo;
+package works.bosk.bson;
 
 class BsonFormatException extends IllegalStateException {
 	public BsonFormatException(String s) { super(s); }

bosk-bson/src/main/java/works/bosk/bson/BsonFormatter.java

@@ -0,0 +1,332 @@
+package works.bosk.bson;
+
+import java.lang.reflect.Type;
+import java.net.URLDecoder;
+import java.nio.charset.StandardCharsets;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.Iterator;
+import java.util.List;
+import java.util.NoSuchElementException;
+import java.util.function.Function;
+import java.util.function.UnaryOperator;
+import java.util.regex.Pattern;
+import org.bson.BsonDocument;
+import org.bson.BsonDocumentWriter;
+import org.bson.BsonReader;
+import org.bson.BsonValue;
+import org.bson.codecs.Codec;
+import org.bson.codecs.DecoderContext;
+import org.bson.codecs.DocumentCodecProvider;
+import org.bson.codecs.EncoderContext;
+import org.bson.codecs.ValueCodecProvider;
+import org.bson.codecs.configuration.CodecRegistries;
+import org.bson.codecs.configuration.CodecRegistry;
+import works.bosk.BoskDiagnosticContext;
+import works.bosk.BoskInfo;
+import works.bosk.Listing;
+import works.bosk.Path;
+import works.bosk.Reference;
+import works.bosk.SerializationPlugin;
+import works.bosk.SideTable;
+import works.bosk.exceptions.InvalidTypeException;
+
+import static works.bosk.Path.validSegment;
+import static works.bosk.ReferenceUtils.rawClass;
+
+/**
+ * Utility class for encoding and decoding bosk state trees as BSON.
+ */
+public class BsonFormatter {
+	protected static final UnaryOperator<String> DECODER;
+	protected static final UnaryOperator<String> ENCODER;
+	protected final CodecRegistry simpleCodecs;
+	protected final Function<Type, Codec<?>> preferredBoskCodecs;
+	protected final Function<Reference<?>, SerializationPlugin.DeserializationScope> deserializationScopeFunction;
+
+	public BsonFormatter(BoskInfo<?> boskInfo, BsonPlugin bsonPlugin) {
+		this.simpleCodecs = CodecRegistries.fromProviders(
+			bsonPlugin.codecProviderFor(boskInfo),
+			new ValueCodecProvider(),
+			new DocumentCodecProvider());
+		this.preferredBoskCodecs = type -> bsonPlugin.getCodec(type, rawClass(type), simpleCodecs, boskInfo);
+		this.deserializationScopeFunction = bsonPlugin::newDeserializationScope;
+	}
+
+	public static String dottedFieldNameSegment(String segment) {
+		return ENCODER.apply(validSegment(segment));
+	}
+
+	public static String undottedFieldNameSegment(String dottedSegment) {
+		return DECODER.apply(dottedSegment);
+	}
+
+	/**
+	 * @param refLength behave as though <code>ref</code> were truncated to this length without actually having to do it
+	 */
+	private static <T> void buildDottedFieldNameOf(Reference<T> ref, int startLength, int refLength, ArrayList<String> segments) {
+		if (ref.path().length() > startLength) {
+			Reference<?> enclosingReference = ref.enclosingReference(Object.class);
+			BsonFormatter.buildDottedFieldNameOf(enclosingReference, startLength, refLength, segments);
+			if (ref.path().length() <= refLength) {
+				if (Listing.class.isAssignableFrom(enclosingReference.targetClass())) {
+					segments.add("ids");
+				} else if (SideTable.class.isAssignableFrom(enclosingReference.targetClass())) {
+					segments.add("valuesById");
+				}
+				segments.add(BsonFormatter.dottedFieldNameSegment(ref.path().lastSegment()));
+			}
+		}
+	}
+
+	static {
+		DECODER = s->{
+			return URLDecoder.decode(s, StandardCharsets.UTF_8);
+		};
+
+		ENCODER = s->{
+			// Selective percent-encoding of characters MongoDB doesn't like.
+			// Standard percent-encoding doesn't handle the period character, which
+			// we want, so if we're already diverging from the standard, we might
+			// as well do something that suits our needs.
+			// Good to stay compatible with standard percent-DEcoding, though.
+			StringBuilder sb = new StringBuilder();
+			for (int i = 0; i < s.length(); ) {
+				int cp = s.codePointAt(i);
+				switch (cp) {
+					case '%': // For percent-encoding
+					case '+': case ' ': // These two are affected by URLDecoder
+					case '$': // MongoDB treats these specially
+					case '.': // MongoDB separator for dotted field names
+					case 0:   // Can MongoDB handle nulls? Probably. Do we want to find out? Not really.
+					case '|': // (These are reserved for internal use)
+					case '!':
+					case '~':
+					case '[':
+					case ']':
+						appendPercentEncoded(sb, cp);
+						break;
+					default:
+						sb.appendCodePoint(cp);
+						break;
+				}
+				i += Character.charCount(cp);
+			}
+			return sb.toString();
+		};
+	}
+
+	private static void appendPercentEncoded(StringBuilder sb, int cp) {
+		assert 0 <= cp && cp <= 255;
+		sb
+			.append('%')
+			.append(hexCharForDigit(cp / 16))
+			.append(hexCharForDigit(cp % 16));
+	}
+
+	/**
+	 * An uppercase version of {@link Character#forDigit} with a radix of 16.
+	 */
+	private static char hexCharForDigit(int value) {
+		if (value < 10) {
+			return (char)('0' + value);
+		} else {
+			return (char)('A' + value - 10);
+		}
+	}
+
+	/**
+	 * @param ref the bosk node whose field name is required
+	 * @param startingRef the bosk node corresponding to the MongoDB document in which {@code ref} is located
+	 * @return MongoDB field name corresponding to {@code ref} within the document for {@code startingRef}, starting with {@link DocumentFields#state state}
+	 * @see #referenceTo(String, Reference)
+	 */
+	public static <T> String dottedFieldNameOf(Reference<T> ref, Reference<?> startingRef) {
+		assert startingRef.encloses(ref);
+		return dottedFieldNameOf(ref, ref.path().length(), startingRef);
+	}
+
+	/**
+	 * @param refLength behave as though <code>ref</code> were truncated to this length, without actually having to do it
+	 * @return MongoDB field name corresponding to the given Reference
+	 * @see #referenceTo(String, Reference)
+	 */
+	static <T> String dottedFieldNameOf(Reference<T> ref, int refLength, Reference<?> startingRef) {
+		// TODO: Is this required? It's currently only really called by tests
+		ArrayList<String> segments = dottedFieldNameSegments(ref, refLength, startingRef);
+		return String.join(".", segments.toArray(new String[0]));
+	}
+
+	/**
+	 * @return Reference corresponding to the given field name within a document representing {@code startingReference}
+	 * @see #dottedFieldNameOf
+	 */
+	@SuppressWarnings("unchecked")
+	public static <T> Reference<T> referenceTo(String dottedName, Reference<?> startingReference) throws InvalidTypeException {
+		Reference<?> ref = startingReference;
+		Iterator<String> iter = Arrays.asList(dottedName.split(Pattern.quote("."))).iterator();
+		BsonFormatter.skipField(ref, iter, DocumentFields.state.name()); // The entire Bosk state is in this field
+		while (iter.hasNext()) {
+			if (Listing.class.isAssignableFrom(ref.targetClass())) {
+				BsonFormatter.skipField(ref, iter, "ids");
+			} else if (SideTable.class.isAssignableFrom(ref.targetClass())) {
+				BsonFormatter.skipField(ref, iter, "valuesById");
+			}
+			if (iter.hasNext()) {
+				String segment = undottedFieldNameSegment(iter.next());
+				ref = ref.then(Object.class, segment);
+			}
+		}
+		return (Reference<T>) ref;
+	}
+
+	private static void skipField(Reference<?> ref, Iterator<String> iter, String expectedName) {
+		String actualName;
+		try {
+			actualName = iter.next();
+		} catch (NoSuchElementException e) {
+			throw new IllegalStateException("Expected '" + expectedName + "' for " + ref.targetClass().getSimpleName() + "; encountered end of dotted field name");
+		}
+		if (!expectedName.equals(actualName)) {
+			throw new IllegalStateException("Expected '" + expectedName + "' for " + ref.targetClass().getSimpleName() + "; was: " + actualName);
+		}
+	}
+
+	/**
+	 * A <em>BSON path</em> is a pipe-delimited sequence of BSON fields.
+	 *
+	 * <p>
+	 * Note that the BSON structure is such that the BSON path's segments may
+	 * not exactly match those of {@code ref.path()}.
+	 *
+	 * <p>
+	 * This computes the BSON path for {@code ref} within a document representing {@code startingRef}.
+	 *
+	 * @param ref reference to the node whose path is desired
+	 * @param startingRef reference to the node corresponding to the document containing the {@code ref} node
+	 * @return the BSON path leading to {@code ref} within a document representing {@code startingRef}
+	 */
+	public static String docBsonPath(Reference<?> ref, Reference<?> startingRef) {
+		return "|" + String.join("|", docSegments(ref, startingRef));
+	}
+
+	/**
+	 * @return list of field names suitable for {@link #lookup} to find the document corresponding
+	 * to <code>docRef</code> inside a document corresponding to <code>rootRef</code>
+	 */
+	private static List<String> docSegments(Reference<?> docRef, Reference<?> rootRef) {
+		ArrayList<String> allSegments = dottedFieldNameSegments(docRef, docRef.path().length(), rootRef);
+		return allSegments
+			.subList(1, allSegments.size()); // Skip the "state" field
+	}
+
+	protected Codec<?> codecFor(Type type) {
+		// BsonPlugin gives better codecs than CodecRegistry, because BsonPlugin is aware of generics,
+		// so we always try that first. The CodecSupplier protocol uses "null" to indicate that another
+		// CodecSupplier should be used, so we follow that protocol and fall back on the CodecRegistry.
+		// TODO: Should this logic be in BsonPlugin? It has nothing to do with MongoDriver really.
+		Codec<?> result = preferredBoskCodecs.apply(type);
+		if (result == null) {
+			return simpleCodecs.get(rawClass(type));
+		} else {
+			return result;
+		}
+	}
+
+	/**
+	 * Converts a bosk state node to BSON.
+	 *
+	 * <p>
+	 * A common way to call this is, for a given {@link Reference} {@code ref}, is:
+	 * <pre>
+	 *     ref.value(), ref.targetType()
+	 * </pre>
+	 * @param object the bosk state node to convert
+	 * @param type the type of {@code object}
+	 * @see #bsonValue2object(BsonValue, Reference)
+	 */
+	@SuppressWarnings("unchecked")
+	public <T> BsonValue object2bsonValue(T object, Type type) {
+		rawClass(type).cast(object);
+		Codec<T> objectCodec = (Codec<T>) codecFor(type);
+		BsonDocument document = new BsonDocument();
+		try (BsonDocumentWriter writer = new BsonDocumentWriter(document)) {
+			// To support arbitrary values, not just whole documents, we put the result INSIDE a document.
+			writer.writeStartDocument();
+			writer.writeName("value");
+			objectCodec.encode(writer, object, EncoderContext.builder().build());
+			writer.writeEndDocument();
+		}
+		return document.get("value");
+	}
+
+	/**
+	 * Converts a BSON value to a bosk state node.
+	 * @param bson the value to convert
+	 * @param target the bosk location of the resulting node
+	 * @see #object2bsonValue(Object, Type)
+	 */
+	@SuppressWarnings("unchecked")
+	public <T> T bsonValue2object(BsonValue bson, Reference<T> target) {
+		Codec<T> objectCodec = (Codec<T>) codecFor(target.targetType());
+		BsonDocument document = new BsonDocument();
+		document.append("value", bson);
+		try (
+			@SuppressWarnings("unused") SerializationPlugin.DeserializationScope scope = deserializationScopeFunction.apply(target);
+			BsonReader reader = document.asBsonReader()
+		) {
+			reader.readStartDocument();
+			reader.readName("value");
+			return objectCodec.decode(reader, DecoderContext.builder().build());
+		}
+	}
+
+	/**
+	 * The fields of the main MongoDB document.  Case-sensitive.
+	 *
+	 * <p>
+	 * No field name should be a prefix of any other.
+	 */
+	public enum DocumentFields {
+		/**
+		 * The location of this document's state within the conceptual giant document.
+		 * A pipe-separated list of BSON field names.
+		 */
+		path,
+
+		/**
+		 * The BSON-encoded portion of the bosk state represented by this document.
+		 */
+		state,
+
+		/**
+		 * An ever-increasing 64-bit long that is incremented every time the document changes.
+		 */
+		revision,
+
+		/**
+		 * The contents of {@link BoskDiagnosticContext} corresponding to this
+		 * document's last update.
+		 */
+		diagnostics,
+	}
+
+	static <T> ArrayList<String> dottedFieldNameSegments(Reference<T> ref, int refLength, Reference<?> startingRef) {
+		assert startingRef.path().matchesPrefixOf(ref.path()): "'" + ref + "' must be under '" + startingRef + "'";
+		ArrayList<String> segments = new ArrayList<>();
+		segments.add(DocumentFields.state.name());
+		buildDottedFieldNameOf(ref, startingRef.path().length(), refLength, segments);
+		return segments;
+	}
+
+	/**
+	 * @param elementRefLength behave as though <code>elementRef</code> were truncated to this length, without actually having to do it
+	 * @return MongoDB field name corresponding to the object that contains the given element
+	 * @see #referenceTo(String, Reference)
+	 */
+	static <T> List<String> containerSegments(Reference<T> elementRef, int elementRefLength, Reference<?> startingRef) {
+		List<String> elementSegments = dottedFieldNameSegments(elementRef, elementRefLength, startingRef);
+		return elementSegments.subList(0, elementSegments.size()-1); // Trim off the element itself
+	}
+
+}