diff --git a/api/src/main/java/com/google/common/flogger/LogContext.java b/api/src/main/java/com/google/common/flogger/LogContext.java index 4864c4ed..38887d21 100644 --- a/api/src/main/java/com/google/common/flogger/LogContext.java +++ b/api/src/main/java/com/google/common/flogger/LogContext.java @@ -517,7 +517,7 @@ protected boolean postProcess(@NullableDecl LogSiteKey logSiteKey) { * of arguments. */ private boolean shouldLog() { - // The log site may have already been injected via "withInjectedLogSite()". + // The log site may have already been injected via "withInjectedLogSite()" or similar. if (logSite == null) { // From the point at which we call inferLogSite() we can skip 1 additional method (the // shouldLog() method itself) when looking up the stack to find the log() method. @@ -571,9 +571,11 @@ private void logImpl(String message, Object... args) { @Override public final API withInjectedLogSite(LogSite logSite) { // First call wins (since auto-injection will typically target the log() method at the end of - // the chain and might not check for previous explicit injection). - if (this.logSite == null) { - this.logSite = checkNotNull(logSite, "log site"); + // the chain and might not check for previous explicit injection). In particular it MUST be + // allowed for a caller to specify the "INVALID" log site, and have that set the field here to + // disable log site lookup at this log statement (though passing "null" is a no-op). + if (this.logSite == null && logSite != null) { + this.logSite = logSite; } return api(); } diff --git a/api/src/main/java/com/google/common/flogger/LogSites.java b/api/src/main/java/com/google/common/flogger/LogSites.java index 5feaf3d9..3c367821 100644 --- a/api/src/main/java/com/google/common/flogger/LogSites.java +++ b/api/src/main/java/com/google/common/flogger/LogSites.java @@ -26,16 +26,83 @@ * */ public final class LogSites { + /** + * Returns a {@code LogSite} for the caller of the specified class. This can be used in + * conjunction with the {@link LoggingApi#withInjectedLogSite(LogSite)} method to implement + * logging helper methods. In some platforms, log site determination may be unsupported, and in + * those cases this method will always return the {@link LogSite#INVALID} instance. + *
+ * For example (in {@code MyLoggingHelper}): + *
{@code + * public static void logAndSomethingElse(String message, Object... args) { + * logger.atInfo() + * .withInjectedLogSite(callerOf(MyLoggingHelper.class)) + * .logVarargs(message, args); + * } + * }+ *
+ * This method should be used for the simple cases where the class in which the logging occurs is + * a public logging API. If the log statement is in a different class (not the public logging API) + * and the {@code LogSite} instance needs to be passed through several layers, consider using + * {@link #logSite()} instead to avoid too much "magic" in your code. + *
+ * You should also seek to ensure that any API used with this method "looks like a logging API". + * It's no good if a log entry contains a class and method name which doesn't correspond to + * anything the user can relate to. In particular, the API should probably always accept the log + * message or at least some of its parameters, and should always have methods with "log" in their + * names to make the connection clear. + *
+ * It is very important to note that this method can be very slow, since determining the log site + * can involve stack trace analysis. It is only recommended that it is used for cases where + * logging is expected to occur (e.g. {@code INFO} level or above). Implementing a helper method + * for {@code FINE} logging is usually unnecessary (it doesn't normally need to follow any + * specific "best practice" behavior). + *
+ * Note that even when log site determination is supported, it is not defined as to whether two + * invocations of this method on the same line of code will produce the same instance, equivalent + * instances or distinct instance. Thus you should never invoke this method twice in a single + * statement (and you should never need to). + *
+ * Note that this method call may be replaced in compiled applications via bytecode manipulation + * or other mechanisms to improve performance. + * + * @param loggingApi the logging API to be identified as the source of log statements (this must + * appear somewhere on the stack above the point at which this method is called). + * @return the log site of the caller of the specified logging API, or {@link LogSite#INVALID} if + * the logging API was not found. + */ + public static LogSite callerOf(Class> loggingApi) { + // Can't skip anything here since someone could pass in LogSite.class. + return Platform.getCallerFinder().findLogSite(loggingApi, 0); + } + /** * Returns a {@code LogSite} for the current line of code. This can be used in conjunction with * the {@link LoggingApi#withInjectedLogSite(LogSite)} method to implement logging helper * methods. In some platforms, log site determination may be unsupported, and in those cases this * method will always return the {@link LogSite#INVALID} instance. *
+ * For example (in {@code MyLoggingHelper}): + *
{@code + * public static void logAndSomethingElse(LogSite logSite, String message, Object... args) { + * logger.atInfo() + * .withInjectedLogSite(logSite) + * .logVarargs(message, args); + * } + * }+ * where callers would do: + *
{@code + * MyLoggingHelper.logAndSomethingElse(logSite(), "message..."); + * }+ *
+ * Because this method adds an additional parameter and exposes a Flogger specific type to the + * calling code, you should consider using {@link #callerOf(Class>)} for simple logging + * utilities. + *
* It is very important to note that this method can be very slow, since determining the log site * can involve stack trace analysis. It is only recommended that it is used for cases where - * logging is expected to occur (e.g. {@code WARNING} level or above). Implementing a helper - * method for {@code FINE} logging is usually unnecessary (it doesn't normally need to follow any + * logging is expected to occur (e.g. {@code INFO} level or above). Implementing a helper method + * for {@code FINE} logging is usually unnecessary (it doesn't normally need to follow any * specific "best practice" behavior). *
* Note that even when log site determination is supported, it is not defined as to whether two @@ -45,8 +112,11 @@ public final class LogSites { *
* Note that this method call may be replaced in compiled applications via bytecode manipulation
* or other mechanisms to improve performance.
+ *
+ * @return the log site of the caller of this method.
*/
public static LogSite logSite() {
+ // Don't call "callerOf()" to avoid making another stack entry.
return Platform.getCallerFinder().findLogSite(LogSites.class, 0);
}
diff --git a/api/src/main/java/com/google/common/flogger/LoggingApi.java b/api/src/main/java/com/google/common/flogger/LoggingApi.java
index e16a0636..f1ae7eec 100644
--- a/api/src/main/java/com/google/common/flogger/LoggingApi.java
+++ b/api/src/main/java/com/google/common/flogger/LoggingApi.java
@@ -627,7 +627,6 @@ protected final API noOp() {
@Override
public API withInjectedLogSite(LogSite logSite) {
- checkNotNull(logSite, "log site");
return noOp();
}
diff --git a/api/src/test/java/com/google/common/flogger/LogContextTest.java b/api/src/test/java/com/google/common/flogger/LogContextTest.java
index 2aedffaa..2264fa76 100644
--- a/api/src/test/java/com/google/common/flogger/LogContextTest.java
+++ b/api/src/test/java/com/google/common/flogger/LogContextTest.java
@@ -544,7 +544,6 @@ public void testStackTraceFormatting() {
// Print the stack trace via the expected method (ie, printStackTrace()).
Throwable cause = backend.getLogged(0).getMetadata().findValue(Key.LOG_CAUSE);
assertThat(cause).hasMessageThat().isEqualTo("SMALL");
- ;
StringWriter out = new StringWriter();
cause.printStackTrace(new PrintWriter(out));
Iterable