All the examples I've seen that use aspect oriented programming for logging either log just class, method name and duration, and if they log parameters and return values they simply use ToString(). I need to have more control over what is logged. For example I want to skip passwords, or in some cases log all properties of an object but in other cases just the id property.
Any suggestions? I looked at AspectJ in Java and Unity interception in C# and could not find a solution.
You could try introducing parameter annotations to augment your parameters with some attributes. One of those attributes could signal to skip logging the parameter, another one could be used to specify a converter class for the string representation.
With the following annotations:
#Documented
#Retention(RetentionPolicy.RUNTIME)
#Target(ElementType.METHOD)
public #interface Log {
}
#Documented
#Retention(RetentionPolicy.RUNTIME)
#Target(ElementType.PARAMETER)
public #interface SkipLogging {
}
#Documented
#Retention(RetentionPolicy.RUNTIME)
#Target(ElementType.PARAMETER)
public #interface ToStringWith {
Class<? extends Function<?, String>> value();
}
the aspect could look like this:
import java.lang.reflect.Parameter;
import java.util.function.Function;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
import org.aspectj.lang.reflect.MethodSignature;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
public aspect LoggingAspect {
private final static Logger logger = LoggerFactory.getLogger(LoggingAspect.class);
pointcut loggableMethod(): execution(#Log * *..*.*(..));
before(): loggableMethod() {
MethodSignature signature = (MethodSignature) thisJoinPoint.getSignature();
Parameter[] parameters = signature.getMethod()
.getParameters();
String message = IntStream.range(0, parameters.length)
.filter(i -> this.isLoggable(parameters[i]))
.<String>mapToObj(i -> toString(parameters[i], thisJoinPoint.getArgs()[i]))
.collect(Collectors.joining(", ",
"method execution " + signature.getName() + "(", ")"));
Logger methodLogger = LoggerFactory.getLogger(
thisJoinPointStaticPart.getSignature().getDeclaringType());
methodLogger.debug(message);
}
private boolean isLoggable(Parameter parameter) {
return parameter.getAnnotation(SkipLogging.class) == null;
}
private String toString(Parameter parameter, Object value) {
ToStringWith toStringWith = parameter.getAnnotation(ToStringWith.class);
if (toStringWith != null) {
Class<? extends Function<?, String>> converterClass =
toStringWith.value();
try {
#SuppressWarnings("unchecked")
Function<Object, String> converter = (Function<Object, String>)
converterClass.newInstance();
String str = converter.apply(value);
return String.format("%s='%s'", parameter.getName(), str);
} catch (Exception e) {
logger.error("Couldn't instantiate toString converter for logging "
+ converterClass.getName(), e);
return String.format("%s=<error converting to string>",
parameter.getName());
}
} else {
return String.format("%s='%s'", parameter.getName(), String.valueOf(value));
}
}
}
Test code:
public static class SomethingToStringConverter implements Function<Something, String> {
#Override
public String apply(Something something) {
return "Something nice";
}
}
#Log
public void test(
#ToStringWith(SomethingToStringConverter.class) Something something,
String string,
#SkipLogging Class<?> cls,
Object object) {
}
public static void main(String[] args) {
// execution of this method should log the following message:
// method execution test(something='Something nice', string='some string', object='null')
test(new Something(), "some string", Object.class, null);
}
I used Java 8 Streams API in my answer for it's compactness, you could convert the code to normal Java code if you don't use Java 8 features or need better efficiency. It's just to give you an idea.
Related
I have created below JUnit5 parameterized test with ArgumentsSource for loading arguments for the test:
public class DemoModelValidationTest {
public ParamsProvider paramsProvider;
public DemoModelValidationTest () {
try {
paramsProvider = new ParamsProvider();
}
catch (Exception iaex) {
}
}
#ParameterizedTest
#ArgumentsSource(ParamsProvider.class)
void testAllConfigurations(int configIndex, String a) throws Exception {
paramsProvider.executeSimulation(configIndex);
}
}
and the ParamsProvider class looks like below:
public class ParamsProvider implements ArgumentsProvider {
public static final String modelPath = System.getProperty("user.dir") + File.separator + "demoModels";
YAMLDeserializer deserializedYAML;
MetaModelToValidationModel converter;
ValidationRunner runner;
List<Configuration> configurationList;
List<Arguments> listOfArguments;
public ParamsProvider() throws Exception {
configurationList = new ArrayList<>();
listOfArguments = new LinkedList<>();
deserializedYAML = new YAMLDeserializer(modelPath);
deserializedYAML.load();
converter = new MetaModelToValidationModel(deserializedYAML);
runner = converter.convert();
configurationList = runner.getConfigurations();
for (int i = 0; i < configurationList.size(); i++) {
listOfArguments.add(Arguments.of(i, configurationList.get(i).getName()));
}
}
public void executeSimulation(int configListIndex) throws Exception {
final Configuration config = runner.getConfigurations().get(configListIndex);
runner.run(config);
runner.getReporter().consolePrintReport();
}
#Override
public Stream<? extends Arguments> provideArguments(ExtensionContext context) {
return listOfArguments.stream().map(Arguments::of);
// return Stream.of(Arguments.of(0, "Actuator Power"), Arguments.of(1, "Error Logging"));
}}
In the provideArguments() method, the commented out code is working fine, but the first line of code
listOfArguments.stream().map(Arguments::of)
is returning the following error:
org.junit.platform.commons.PreconditionViolationException: Configuration error: You must configure at least one set of arguments for this #ParameterizedTest
I am not sure whether I am having a casting problem for the stream in provideArguments() method, but I guess it somehow cannot map the elements of listOfArguments to the stream, which can finally take the form like below:
Stream.of(Arguments.of(0, "Actuator Power"), Arguments.of(1, "Error Logging"))
Am I missing a proper stream mapping of listOfArguments?
provideArguments(…) is called before your test is invoked.
Your ParamsProvider class is instantiated by JUnit. Whatever you’re doing in desiralizeAndCreateValidationRunnerInstance should be done in the ParamsProvider constructor.
Also you’re already wrapping the values fro deserialised configurations to Arguments and you’re double wrapping them in providesArguments.
Do this:
#Override
public Stream<? extends Arguments> provideArguments(ExtensionContext context) {
return listOfArguments.stream();
}}
This might be a duplicate. But I cannot find a solution to my Problem.
I have a class
public class MyResponse implements Serializable {
private boolean isSuccess;
public boolean isSuccess() {
return isSuccess;
}
public void setSuccess(boolean isSuccess) {
this.isSuccess = isSuccess;
}
}
Getters and setters are generated by Eclipse.
In another class, I set the value to true, and write it as a JSON string.
System.out.println(new ObjectMapper().writeValueAsString(myResponse));
In JSON, the key is coming as {"success": true}.
I want the key as isSuccess itself. Is Jackson using the setter method while serializing? How do I make the key the field name itself?
This is a slightly late answer, but may be useful for anyone else coming to this page.
A simple solution to changing the name that Jackson will use for when serializing to JSON is to use the #JsonProperty annotation, so your example would become:
public class MyResponse implements Serializable {
private boolean isSuccess;
#JsonProperty(value="isSuccess")
public boolean isSuccess() {
return isSuccess;
}
public void setSuccess(boolean isSuccess) {
this.isSuccess = isSuccess;
}
}
This would then be serialised to JSON as {"isSuccess":true}, but has the advantage of not having to modify your getter method name.
Note that in this case you could also write the annotation as #JsonProperty("isSuccess") as it only has the single value element
I recently ran into this issue and this is what I found. Jackson will inspect any class that you pass to it for getters and setters, and use those methods for serialization and deserialization. What follows "get", "is" and "set" in those methods will be used as the key for the JSON field ("isValid" for getIsValid and setIsValid).
public class JacksonExample {
private boolean isValid = false;
public boolean getIsValid() {
return isValid;
}
public void setIsValid(boolean isValid) {
this.isValid = isValid;
}
}
Similarly "isSuccess" will become "success", unless renamed to "isIsSuccess" or "getIsSuccess"
Read more here: http://www.citrine.io/blog/2015/5/20/jackson-json-processor
Using both annotations below, forces the output JSON to include is_xxx:
#get:JsonProperty("is_something")
#param:JsonProperty("is_something")
When you are using Kotlin and data classes:
data class Dto(
#get:JsonProperty("isSuccess") val isSuccess: Boolean
)
You might need to add #param:JsonProperty("isSuccess") if you are going to deserialize JSON as well.
EDIT: If you are using swagger-annotations to generate documentation, the property will be marked as readOnly when using #get:JsonProperty. In order to solve this, you can do:
#JsonAutoDetect(isGetterVisibility = JsonAutoDetect.Visibility.NONE)
data class Dto(
#field:JsonProperty(value = "isSuccess") val isSuccess: Boolean
)
You can configure your ObjectMapper as follows:
mapper.setPropertyNamingStrategy(new PropertyNamingStrategy() {
#Override
public String nameForGetterMethod(MapperConfig<?> config, AnnotatedMethod method, String defaultName)
{
if(method.hasReturnType() && (method.getRawReturnType() == Boolean.class || method.getRawReturnType() == boolean.class)
&& method.getName().startsWith("is")) {
return method.getName();
}
return super.nameForGetterMethod(config, method, defaultName);
}
});
I didn't want to mess with some custom naming strategies, nor re-creating some accessors.
The less code, the happier I am.
This did the trick for us :
import com.fasterxml.jackson.annotation.JsonIgnoreProperties;
import com.fasterxml.jackson.annotation.JsonProperty;
#JsonIgnoreProperties({"success", "deleted"}) // <- Prevents serialization duplicates
public class MyResponse {
private String id;
private #JsonProperty("isSuccess") boolean isSuccess; // <- Forces field name
private #JsonProperty("isDeleted") boolean isDeleted;
}
Building upon Utkarsh's answer..
Getter names minus get/is is used as the JSON name.
public class Example{
private String radcliffe;
public getHarryPotter(){
return radcliffe;
}
}
is stored as { "harryPotter" : "whateverYouGaveHere" }
For Deserialization, Jackson checks against both the setter and the field name.
For the Json String { "word1" : "example" }, both the below are valid.
public class Example{
private String word1;
public setword2( String pqr){
this.word1 = pqr;
}
}
public class Example2{
private String word2;
public setWord1(String pqr){
this.word2 = pqr ;
}
}
A more interesting question is which order Jackson considers for deserialization. If i try to deserialize { "word1" : "myName" } with
public class Example3{
private String word1;
private String word2;
public setWord1( String parameter){
this.word2 = parameter ;
}
}
I did not test the above case, but it would be interesting to see the values of word1 & word2 ...
Note: I used drastically different names to emphasize which fields are required to be same.
You can change primitive boolean to java.lang.Boolean (+ use #JsonPropery)
#JsonProperty("isA")
private Boolean isA = false;
public Boolean getA() {
return this.isA;
}
public void setA(Boolean a) {
this.isA = a;
}
Worked excellent for me.
If you are interested in handling 3rd party classes not under your control (like #edmundpie mentioned in a comment) then you add Mixin classes to your ObjectMapper where the property/field names should match the ones from your 3rd party class:
public class MyStack32270422 {
public static void main(String[] args) {
ObjectMapper om3rdParty = new ObjectMapper();
om3rdParty .addMixIn(My3rdPartyResponse.class, MixinMyResponse.class);
// add further mixins if required
String jsonString = om3rdParty.writeValueAsString(new My3rdPartyResponse());
System.out.println(jsonString);
}
}
class MixinMyResponse {
// add all jackson annotations here you want to be used when handling My3rdPartyResponse classes
#JsonProperty("isSuccess")
private boolean isSuccess;
}
class My3rdPartyResponse{
private boolean isSuccess = true;
// getter and setter here if desired
}
Basically you add all your Jackson annotations to your Mixin classes as if you would own the class. In my opinion quite a nice solution as you don't have to mess around with checking method names starting with "is.." and so on.
there is another method for this problem.
just define a new sub-class extends PropertyNamingStrategy and pass it to ObjectMapper instance.
here is a code snippet may be help more:
mapper.setPropertyNamingStrategy(new PropertyNamingStrategy() {
#Override
public String nameForGetterMethod(MapperConfig<?> config, AnnotatedMethod method, String defaultName) {
String input = defaultName;
if(method.getName().startsWith("is")){
input = method.getName();
}
//copy from LowerCaseWithUnderscoresStrategy
if (input == null) return input; // garbage in, garbage out
int length = input.length();
StringBuilder result = new StringBuilder(length * 2);
int resultLength = 0;
boolean wasPrevTranslated = false;
for (int i = 0; i < length; i++)
{
char c = input.charAt(i);
if (i > 0 || c != '_') // skip first starting underscore
{
if (Character.isUpperCase(c))
{
if (!wasPrevTranslated && resultLength > 0 && result.charAt(resultLength - 1) != '_')
{
result.append('_');
resultLength++;
}
c = Character.toLowerCase(c);
wasPrevTranslated = true;
}
else
{
wasPrevTranslated = false;
}
result.append(c);
resultLength++;
}
}
return resultLength > 0 ? result.toString() : input;
}
});
The accepted answer won't work for my case.
In my case, the class is not owned by me. The problematic class comes from 3rd party dependencies, so I can't just add #JsonProperty annotation in it.
To solve it, inspired by #burak answer above, I created a custom PropertyNamingStrategy as follow:
mapper.setPropertyNamingStrategy(new PropertyNamingStrategy() {
#Override
public String nameForSetterMethod(MapperConfig<?> config, AnnotatedMethod method, String defaultName)
{
if (method.getParameterCount() == 1 &&
(method.getRawParameterType(0) == Boolean.class || method.getRawParameterType(0) == boolean.class) &&
method.getName().startsWith("set")) {
Class<?> containingClass = method.getDeclaringClass();
String potentialFieldName = "is" + method.getName().substring(3);
try {
containingClass.getDeclaredField(potentialFieldName);
return potentialFieldName;
} catch (NoSuchFieldException e) {
// do nothing and fall through
}
}
return super.nameForSetterMethod(config, method, defaultName);
}
#Override
public String nameForGetterMethod(MapperConfig<?> config, AnnotatedMethod method, String defaultName)
{
if(method.hasReturnType() && (method.getRawReturnType() == Boolean.class || method.getRawReturnType() == boolean.class)
&& method.getName().startsWith("is")) {
Class<?> containingClass = method.getDeclaringClass();
String potentialFieldName = method.getName();
try {
containingClass.getDeclaredField(potentialFieldName);
return potentialFieldName;
} catch (NoSuchFieldException e) {
// do nothing and fall through
}
}
return super.nameForGetterMethod(config, method, defaultName);
}
});
Basically what this does is, before serializing and deserializing, it checks in the target/source class which property name is present in the class, whether it is isEnabled or enabled property.
Based on that, the mapper will serialize and deserialize to the property name that is exist.
I'm trying to translate a ParameterizedTest from JUnit4 to JUnit5 (sadly I'm not particularly skilled in testing).
In JUnit4 I have the following class:
#RunWith(Parameterized.class)
public class AssertionTestCase {
private final TestInput testInput;
public AssertionTestCase(TestInput testInput) {
this.testInput = testInput;
}
#Parameterized.Parameters
public static Collection<Object[]> data() {
return AssertionTestCaseDataProvider.createDataCase();
}
#Test(timeout = 15 * 60 * 1000L)
public void testDailyAssertion() {
LOG.info("Testing input {}/{}", testInput.getTestCase(), testInput.getTestName());
//assert stuffs
}
}
in the AssertionTestCaseDataProvider class I have a simple method generating a collection of Object[]:
class AssertionTestCaseDataProvider {
static Collection<Object[]> createDataCase() {
final List<TestInput> testInputs = new ArrayList<>();
//create and populate testInputs
return testInputs.stream()
.map(testInput -> new Object[]{testInput})
.collect(Collectors.toList());
}
}
I've been trying to translate it using JUnit5 and obtained this:
class AssertionTestCase {
private final TestInput testInput;
public AssertionTestCase(TestInput testInput) {
this.testInput = testInput;
}
public static Collection<Object[]> data() {
return AssertionTestCaseDataProvider.createDataCase();
}
#ParameterizedTest
#MethodSource("data")
void testDailyAssertion() {
LOG.info("Testing input {}/{}", testInput.getTestCase(), testInput.getTestName());
// assert stuffs
}
}
I did not apply any change to the AssertionTestCaseDataProvider class.
Nevertheless, I'm getting the following error:
No ParameterResolver registered for parameter [com.xxx.xx.xxx.xxx.testinput.TestInput arg0] in constructor [public `com.xxx.xxx.xxx.xxx.xxx.AssertionTestCase(com.xxx.xxx.xxx.xxx.testinput.TestInput)]. org.junit.jupiter.api.extension.ParameterResolutionException: No ParameterResolver registered for parameter [com.xxx.xx.xxx.xxx.testinput.TestInput arg0] in constructor [public com.xxx.xxx.xxx.xxx.xxx.AssertionTestCase(com.xxx.xxx.xxx.xxx.testinput.TestInput)].`
I understand I'm probably not applying correctly JUnit5 when initializing the input collection for the test. Am I missing some annotations?
I've also tried to use #ArgumentSource instead of #MethodSource and implementing Argument for AssertionTestCaseDataProvider, with the same failing results.
It works in a bit another way in Junit5.
Test Method should have parameters, and provider method should return a Stream.
static Stream<Arguments> data(){
return Stream.of(
Arguments.of("a", 1),
Arguments.of("d", 2)
);
}
#ParameterizedTest
#MethodSource("data")
void testDailyAssertion(String a, int b) {
Assertions.assertAll(
() -> Assertions.assertEquals("a", a),
() -> Assertions.assertEquals(1, b)
);
}
In your case you can just return a Stream<TestInput>:
static Stream<TestInput> createDataCase() {
final List<TestInput> testInputs = new ArrayList<>();
//create and populate testInputs
return testInputs.stream();
}
and then in your testMethod:
#ParameterizedTest
#MethodSource("createDataCase")
void testDailyAssertion(TestInput testInput) {
{your assertions}
}
i've written a utility to monitor individual business transactions. For example, Alice calls a method which calls more methods and i want info on just Alice's call, separate from Bob's call to the same method.
Right now the entry point creates a Transaction object and it's passed as an argument to each method:
class Example {
public Item getOrderEntryPoint(int orderId) {
Transaction transaction = transactionManager.create();
transaction.trace("getOrderEntryPoint");
Order order = getOrder(orderId, transaction);
transaction.stop();
logger.info(transaction);
return item;
}
private Order getOrder(int orderId, Transaction t) {
t.trace("getOrder");
Order order = getItems(itemId, t);
t.addStat("number of items", order.getItems().size());
for (Item item : order.getItems()) {
SpecialOffer offer = getSpecialOffer(item, t);
if (null != offer) {
t.incrementStat("offers", 1);
}
}
t.stop();
return order;
}
private SpecialOffer getSpecialOffer(Item item, Transaction t) {
t.trace("getSpecialOffer(" + item.id + ")", TraceCategory.Database);
return offerRepository.getByItem(item);
t.stop();
}
}
This will print to the log something like:
Transaction started by Alice at 10:42
Statistics:
number of items : 3
offers : 1
Category Timings (longest first):
DB : 2s 903ms
code : 187ms
Timings (longest first):
getSpecialOffer(1013) : 626ms
getItems : 594ms
Trace:
getOrderEntryPoint (7ms)
getOrder (594ms)
getSpecialOffer(911) (90ms)
getSpecialOffer(1013) (626ms)
getSpecialOffer(2942) (113ms)
It works great but passing the transaction object around is ugly. Someone suggested AOP but i don't see how to pass the transaction created in the first method to all the other methods.
The Transaction object is pretty simple:
public class Transaction {
private String uuid = UUID.createRandom();
private List<TraceEvent> events = new ArrayList<>();
private Map<String,Int> stats = new HashMap<>();
}
class TraceEvent {
private String name;
private long durationInMs;
}
The app that uses it is a Web app, and this multi-threaded, but the individual transactions are on a single thread - no multi-threading, async code, competition for resources, etc.
My attempt at an annotation:
#Around("execution(* *(..)) && #annotation(Trace)")
public Object around(ProceedingJoinPoint point) {
String methodName = MethodSignature.class.cast(point.getSignature()).getMethod().getName();
//--- Where do i get this call's instance of TRANSACTION from?
if (null == transaction) {
transaction = TransactionManager.createTransaction();
}
transaction.trace(methodName);
Object result = point.proceed();
transaction.stop();
return result;
Introduction
Unfortunately, your pseudo code does not compile. It contains several syntactical and logical errors. Furthermore, some helper classes are missing. If I did not have spare time today and was looking for a puzzle to solve, I would not have bothered making my own MCVE out of it, because that would actually have been your job. Please do read the MCVE article and learn to create one next time, otherwise you will not get a lot of qualified help here. This was your free shot because you are new on SO.
Original situation: passing through transaction objects in method calls
Application helper classes:
package de.scrum_master.app;
public class Item {
private int id;
public Item(int id) {
this.id = id;
}
public int getId() {
return id;
}
#Override
public String toString() {
return "Item[id=" + id + "]";
}
}
package de.scrum_master.app;
public class SpecialOffer {}
package de.scrum_master.app;
public class OfferRepository {
public SpecialOffer getByItem(Item item) {
if (item.getId() < 30)
return new SpecialOffer();
return null;
}
}
package de.scrum_master.app;
import java.util.ArrayList;
import java.util.List;
public class Order {
private int id;
public Order(int id) {
this.id = id;
}
public List<Item> getItems() {
List<Item> items = new ArrayList<>();
int offset = id == 12345 ? 0 : 1;
items.add(new Item(11 + offset, this));
items.add(new Item(22 + offset, this));
items.add(new Item(33 + offset, this));
return items;
}
}
Trace classes:
package de.scrum_master.trace;
public enum TraceCategory {
Code, Database
}
package de.scrum_master.trace;
class TraceEvent {
private String name;
private TraceCategory category;
private long durationInMs;
private boolean finished = false;
public TraceEvent(String name, TraceCategory category, long startTime) {
this.name = name;
this.category = category;
this.durationInMs = startTime;
}
public long getDurationInMs() {
return durationInMs;
}
public void setDurationInMs(long durationInMs) {
this.durationInMs = durationInMs;
}
public boolean isFinished() {
return finished;
}
public void setFinished(boolean finished) {
this.finished = finished;
}
#Override
public String toString() {
return "TraceEvent[name=" + name + ", category=" + category +
", durationInMs=" + durationInMs + ", finished=" + finished + "]";
}
}
Transaction classes:
Here I tried to mimic your own Transaction class with as few as possible changes, but there was a lot I had to add and modify in order to emulate a simplified version of your trace output. This is not thread-safe and the way I am locating the last unfinished TraceEvent is not nice and only works cleanly if there are not exceptions. But you get the idea, I hope. The point is to just make it basically work and subsequently get log output similar to your example. If this was originally my code, I would have solved it differently.
package de.scrum_master.trace;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.UUID;
public class Transaction {
private String uuid = UUID.randomUUID().toString();
private List<TraceEvent> events = new ArrayList<>();
private Map<String, Integer> stats = new HashMap<>();
public void trace(String message) {
trace(message, TraceCategory.Code);
}
public void trace(String message, TraceCategory category) {
events.add(new TraceEvent(message, category, System.currentTimeMillis()));
}
public void stop() {
TraceEvent event = getLastUnfinishedEvent();
event.setDurationInMs(System.currentTimeMillis() - event.getDurationInMs());
event.setFinished(true);
}
private TraceEvent getLastUnfinishedEvent() {
return events
.stream()
.filter(event -> !event.isFinished())
.reduce((first, second) -> second)
.orElse(null);
}
public void addStat(String text, int size) {
stats.put(text, size);
}
public void incrementStat(String text, int increment) {
Integer currentCount = stats.get(text);
if (currentCount == null)
currentCount = 0;
stats.put(text, currentCount + increment);
}
#Override
public String toString() {
return "Transaction {" +
toStringUUID() +
toStringStats() +
toStringEvents() +
"\n}\n";
}
private String toStringUUID() {
return "\n uuid = " + uuid;
}
private String toStringStats() {
String result = "\n stats = {";
for (Entry<String, Integer> statEntry : stats.entrySet())
result += "\n " + statEntry;
return result + "\n }";
}
private String toStringEvents() {
String result = "\n events = {";
for (TraceEvent event : events)
result += "\n " + event;
return result + "\n }";
}
}
package de.scrum_master.trace;
public class TransactionManager {
public Transaction create() {
return new Transaction();
}
}
Example driver application:
package de.scrum_master.app;
import de.scrum_master.trace.TraceCategory;
import de.scrum_master.trace.Transaction;
import de.scrum_master.trace.TransactionManager;
public class Example {
private TransactionManager transactionManager = new TransactionManager();
private OfferRepository offerRepository = new OfferRepository();
public Order getOrderEntryPoint(int orderId) {
Transaction transaction = transactionManager.create();
transaction.trace("getOrderEntryPoint");
sleep(100);
Order order = getOrder(orderId, transaction);
transaction.stop();
System.out.println(transaction);
return order;
}
private Order getOrder(int orderId, Transaction t) {
t.trace("getOrder");
sleep(200);
Order order = new Order(orderId);
t.addStat("number of items", order.getItems().size());
for (Item item : order.getItems()) {
SpecialOffer offer = getSpecialOffer(item, t);
if (null != offer)
t.incrementStat("special offers", 1);
}
t.stop();
return order;
}
private SpecialOffer getSpecialOffer(Item item, Transaction t) {
t.trace("getSpecialOffer(" + item.getId() + ")", TraceCategory.Database);
sleep(50);
SpecialOffer specialOffer = offerRepository.getByItem(item);
t.stop();
return specialOffer;
}
private void sleep(long millis) {
try {
Thread.sleep(millis);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
public static void main(String[] args) {
new Example().getOrderEntryPoint(12345);
new Example().getOrderEntryPoint(23456);
}
}
If you run this code, the output is as follows:
Transaction {
uuid = 62ec9739-bd32-4a56-b6b3-a8a13624961a
stats = {
special offers=2
number of items=3
}
events = {
TraceEvent[name=getOrderEntryPoint, category=Code, durationInMs=561, finished=true]
TraceEvent[name=getOrder, category=Code, durationInMs=451, finished=true]
TraceEvent[name=getSpecialOffer(11), category=Database, durationInMs=117, finished=true]
TraceEvent[name=getSpecialOffer(22), category=Database, durationInMs=69, finished=true]
TraceEvent[name=getSpecialOffer(33), category=Database, durationInMs=63, finished=true]
}
}
Transaction {
uuid = a420cd70-96e5-44c4-a0a4-87e421d05e87
stats = {
special offers=2
number of items=3
}
events = {
TraceEvent[name=getOrderEntryPoint, category=Code, durationInMs=469, finished=true]
TraceEvent[name=getOrder, category=Code, durationInMs=369, finished=true]
TraceEvent[name=getSpecialOffer(12), category=Database, durationInMs=53, finished=true]
TraceEvent[name=getSpecialOffer(23), category=Database, durationInMs=63, finished=true]
TraceEvent[name=getSpecialOffer(34), category=Database, durationInMs=53, finished=true]
}
}
AOP refactoring
Preface
Please note that I am using AspectJ here because two things about your code would never work with Spring AOP because it works with a delegation pattern based on dynamic proxies:
self-invocation (internally calling a method of the same class or super-class)
intercepting private methods
Because of these Spring AOP limitations I advise you to either refactor your code so as to avoid the two issues above or to configure your Spring applications to use full AspectJ via LTW (load-time weaving) instead.
As you noticed, my sample code does not use Spring at all because AspectJ is completely independent of Spring and works with any Java application (or other JVM languages, too).
Refactoring idea
Now what should you do in order to get rid of passing around tracing information (Transaction objects), polluting your core application code and tangling it with trace calls?
You extract transaction tracing into an aspect taking care of all trace(..) and stop() calls.
Unfortunately your Transaction class contains different types of information and does different things, so you cannot completely get rid of context information about how to trace for each affected method. But at least you can extract that context information from the method bodies and transform it into a declarative form using annotations with parameters.
These annotations can be targeted by an aspect taking care of handling transaction tracing.
Added and updated code, iteration 1
Annotations related to transaction tracing:
package de.scrum_master.trace;
import static java.lang.annotation.ElementType.METHOD;
import static java.lang.annotation.RetentionPolicy.RUNTIME;
import java.lang.annotation.Retention;
import java.lang.annotation.Target;
#Retention(RUNTIME)
#Target(METHOD)
public #interface TransactionEntryPoint {}
package de.scrum_master.trace;
import static java.lang.annotation.ElementType.METHOD;
import static java.lang.annotation.RetentionPolicy.RUNTIME;
import java.lang.annotation.Retention;
import java.lang.annotation.Target;
#Retention(RUNTIME)
#Target(METHOD)
public #interface TransactionTrace {
String message() default "__METHOD_NAME__";
TraceCategory category() default TraceCategory.Code;
String addStat() default "";
String incrementStat() default "";
}
Refactored application classes with annotations:
package de.scrum_master.app;
import java.util.ArrayList;
import java.util.List;
import de.scrum_master.trace.TransactionTrace;
public class Order {
private int id;
public Order(int id) {
this.id = id;
}
#TransactionTrace(message = "", addStat = "number of items")
public List<Item> getItems() {
List<Item> items = new ArrayList<>();
int offset = id == 12345 ? 0 : 1;
items.add(new Item(11 + offset));
items.add(new Item(22 + offset));
items.add(new Item(33 + offset));
return items;
}
}
Nothing much here, only added an annotation to getItems(). But the sample application class changes massively, getting much cleaner and simpler:
package de.scrum_master.app;
import de.scrum_master.trace.TraceCategory;
import de.scrum_master.trace.TransactionEntryPoint;
import de.scrum_master.trace.TransactionTrace;
public class Example {
private OfferRepository offerRepository = new OfferRepository();
#TransactionEntryPoint
#TransactionTrace
public Order getOrderEntryPoint(int orderId) {
sleep(100);
Order order = getOrder(orderId);
return order;
}
#TransactionTrace
private Order getOrder(int orderId) {
sleep(200);
Order order = new Order(orderId);
for (Item item : order.getItems()) {
SpecialOffer offer = getSpecialOffer(item);
// Do something with special offers
}
return order;
}
#TransactionTrace(category = TraceCategory.Database, incrementStat = "specialOffers")
private SpecialOffer getSpecialOffer(Item item) {
sleep(50);
SpecialOffer specialOffer = offerRepository.getByItem(item);
return specialOffer;
}
private void sleep(long millis) {
try {
Thread.sleep(millis);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
public static void main(String[] args) {
new Example().getOrderEntryPoint(12345);
new Example().getOrderEntryPoint(23456);
}
}
See? Except for a few annotations there is nothing left of the transaction tracing logic, the application code only takes care of its core concern. If you also remove the sleep() method which only makes the application slower for demonstration purposes (because we want some nice statistics with measured times >0 ms), the class gets even more compact.
But of course we need to put the transaction tracing logic somewhere, more precisely modularise it into an AspectJ aspect:
Transaction tracing aspect:
package de.scrum_master.trace;
import java.lang.reflect.Array;
import java.util.Arrays;
import java.util.Collection;
import java.util.stream.Collectors;
import org.aspectj.lang.JoinPoint;
import org.aspectj.lang.ProceedingJoinPoint;
import org.aspectj.lang.annotation.After;
import org.aspectj.lang.annotation.Around;
import org.aspectj.lang.annotation.Aspect;
import org.aspectj.lang.annotation.Pointcut;
import org.aspectj.lang.reflect.MethodSignature;
#Aspect("percflow(entryPoint())")
public class TransactionTraceAspect {
private static TransactionManager transactionManager = new TransactionManager();
private Transaction transaction = transactionManager.create();
#Pointcut("execution(* *(..)) && #annotation(de.scrum_master.trace.TransactionEntryPoint)")
private static void entryPoint() {}
#Around("execution(* *(..)) && #annotation(transactionTrace)")
public Object doTrace(ProceedingJoinPoint joinPoint, TransactionTrace transactionTrace) throws Throwable {
preTrace(transactionTrace, joinPoint);
Object result = joinPoint.proceed();
postTrace(transactionTrace);
addStat(transactionTrace, result);
incrementStat(transactionTrace, result);
return result;
}
private void preTrace(TransactionTrace transactionTrace, ProceedingJoinPoint joinPoint) {
String traceMessage = transactionTrace.message();
if ("".equals(traceMessage))
return;
MethodSignature signature = (MethodSignature) joinPoint.getSignature();
if ("__METHOD_NAME__".equals(traceMessage)) {
traceMessage = signature.getName() + "(";
traceMessage += Arrays.stream(joinPoint.getArgs()).map(arg -> arg.toString()).collect(Collectors.joining(", "));
traceMessage += ")";
}
transaction.trace(traceMessage, transactionTrace.category());
}
private void postTrace(TransactionTrace transactionTrace) {
if ("".equals(transactionTrace.message()))
return;
transaction.stop();
}
private void addStat(TransactionTrace transactionTrace, Object result) {
if ("".equals(transactionTrace.addStat()) || result == null)
return;
if (result instanceof Collection)
transaction.addStat(transactionTrace.addStat(), ((Collection<?>) result).size());
else if (result.getClass().isArray())
transaction.addStat(transactionTrace.addStat(), Array.getLength(result));
}
private void incrementStat(TransactionTrace transactionTrace, Object result) {
if ("".equals(transactionTrace.incrementStat()) || result == null)
return;
transaction.incrementStat(transactionTrace.incrementStat(), 1);
}
#After("entryPoint()")
public void logFinishedTransaction(JoinPoint joinPoint) {
System.out.println(transaction);
}
}
Let me explain what this aspect does:
#Pointcut(..) entryPoint() says: Find me all methods in the code annotated by #TransactionEntryPoint. This pointcut is used in two places:
#Aspect("percflow(entryPoint())") says: Create one aspect instance for each control flow beginning at a transaction entry point.
#After("entryPoint()") logFinishedTransaction(..) says: Execute this advice (AOP terminology for a method linked to a pointcut) after an entry point methods is finished. The corresponding method just prints the transaction statistics just like in the original code at the end of Example.getOrderEntryPoint(..).
#Around("execution(* *(..)) && #annotation(transactionTrace)") doTrace(..)says: Wrap methods annotated by TransactionTrace and do the following (method body):
add new trace element and start measuring time
execute original (wrapped) method and store result
update trace element with measured time
add one type of statistics (optional)
increment another type of statistics (optional)
return wrapped method's result to its caller
The private methods are just helpers for the #Around advice.
The console log when running the updated Example class and active AspectJ is:
Transaction {
uuid = 4529d325-c604-441d-8997-45ca659abb14
stats = {
specialOffers=2
number of items=3
}
events = {
TraceEvent[name=getOrderEntryPoint(12345), category=Code, durationInMs=468, finished=true]
TraceEvent[name=getOrder(12345), category=Code, durationInMs=366, finished=true]
TraceEvent[name=getSpecialOffer(Item[id=11]), category=Database, durationInMs=59, finished=true]
TraceEvent[name=getSpecialOffer(Item[id=22]), category=Database, durationInMs=50, finished=true]
TraceEvent[name=getSpecialOffer(Item[id=33]), category=Database, durationInMs=51, finished=true]
}
}
Transaction {
uuid = ef76a996-8621-478b-a376-e9f7a729a501
stats = {
specialOffers=2
number of items=3
}
events = {
TraceEvent[name=getOrderEntryPoint(23456), category=Code, durationInMs=452, finished=true]
TraceEvent[name=getOrder(23456), category=Code, durationInMs=351, finished=true]
TraceEvent[name=getSpecialOffer(Item[id=12]), category=Database, durationInMs=50, finished=true]
TraceEvent[name=getSpecialOffer(Item[id=23]), category=Database, durationInMs=50, finished=true]
TraceEvent[name=getSpecialOffer(Item[id=34]), category=Database, durationInMs=50, finished=true]
}
}
You see, it looks almost identical to the original application.
Idea for further simplification, iteration 2
When reading method Example.getOrder(int orderId) I was wondering why you are calling order.getItems(), looping over it and calling getSpecialOffer(item) inside the loop. In your sample code you do not use the results for anything other than updating the transaction trace object. I am assuming that in your real code you do something with the order and with the special offers in that method.
But just in case you really do not need those calls inside that method, I suggest
you factor the calls out right into the aspect, getting rid of the TransactionTrace annotation parameters String addStat() and String incrementStat().
The Example code would get even simpler and
the annotation #TransactionTrace(message = "", addStat = "number of items") in class would go away, too.
I am leaving this refactoring to you if you think it makes sense.
I am receiving this error when I compile
The type XXX must implement the inherited abstract method
I have three files
A default implementation [com.SafeReaderIMPL.java]
public class SafeReaderIMPL implements ISafeReader {
private boolean successfulRead;
public SafeReaderIMPL() {
successfulRead = true;
}
protected void fail() {
successfulRead = false;
}
#Override
public boolean isSuccessfulRead() {
return successfulRead;
}
}
An interface file [com.ISafeReader.java]
public interface ISafeReader {
public boolean isSuccessfulRead();
}
An apsect (using annotations) [com.SafeReaderAspect.java]
#Aspect
public class SafeReaderAspect {
#DeclareParents(value = "com.BadReader", defaultImpl = SafeReaderIMPL.class)
public ISafeReader implementedInterface;
#AfterThrowing(pointcut = "execution(* *.*(..)) && this(m)", throwing = "e")
public void handleBadRead(JoinPoint joinPoint, ISafeReader m, Throwable e) {
((SafeReaderIMPL)m).fail();
}
}
And a Test Class [com.BadReader]
public class BadReader {
public void fail() throws Throwable {
throw new Throwable();
}
}
I compile the first three files in a separate jar using
ajc -source 1.8 -sourceroots . -outjar aspectLib.jar
I then compile the second file using the aspectLib like so
ajc -source 1.8 -sourceroots . -aspectpath ./aspectLib.jar -outjar common.jar
When I go to compile the second jar I get the error. I am using the latest stable version of AspectJ 1.8.3
BadReader.java:10 [error] The type BadReader must implement the
inherited abstract method ISafeReader.isSuccessfulRead() public class
BadReader {
^^^^^^^^
The problem is not two-step compilation as such, but the fact that #DeclareParents in #AspectJ syntax in not 100% compatible with declare parents in native syntax. Actually, #DeclareParents for introducing default interface implementations is superseded by #DeclareMixin (see this bug ticket), but the downside of the mixin approach is that you do not have a real A implements B scenario there, i.e. you cannot cast as you wish in your after-throwing advice, so this is also not a good option in your case.
So what do you do if you want to keep the two-step compilation approach? Just use native syntax:
Interface:
package com;
public interface ISafeReader {
boolean isSuccessfulRead();
}
Default implementation:
package com;
public class SafeReaderIMPL implements ISafeReader {
private boolean successfulRead;
public SafeReaderIMPL() { successfulRead = true; }
public void fail() { successfulRead = false; }
#Override public boolean isSuccessfulRead() { return successfulRead; }
}
ITD aspect:
package com;
public aspect SafeReaderAspect {
declare parents : com.BadReader implements SafeReaderIMPL;
after(ISafeReader safeReader) throwing : execution(* *(..)) && this(safeReader) {
System.out.println(thisJoinPoint + " - calling 'fail()' before rethrowing error");
((SafeReaderIMPL) safeReader).fail();
}
}
ITD target class with sample main method:
package com;
public class BadReader {
public void doSomething() {
throw new RuntimeException("my error");
}
public static void main(String[] args) {
BadReader badReader = new BadReader();
System.out.println("badReader.isSuccessfulRead() = " + badReader.isSuccessfulRead());
try { badReader.doSomething(); }
catch(Throwable t) { System.out.println(t); }
System.out.println("badReader.isSuccessfulRead() = " + badReader.isSuccessfulRead());
}
}
Now you can use the two-stage compilation approach.
Console output:
badReader.isSuccessfulRead() = true
execution(void com.BadReader.doSomething()) - calling 'fail()' before rethrowing error
java.lang.RuntimeException: my error
badReader.isSuccessfulRead() = false
The problem is due to the two-step compilation. During the second step, ajc needs the source code of SafeReaderIMPL to be able to weave BadReader, but it cannot find it into aspectLib.jar
In fact, if you try compiling in a single step (I did), it compiles and runs.
Unfortunately I don't know a way to fix this without providing the source code during the second compile step, which I suppose would render the whole two-step approach a bit pointless.