* created SystemUtils to check for multi-threading permission
Revert "pmd: Remove hack related to multithreading in jdk5 with Eclipse - no longer relevant today"
This reverts commit 720bb3d3b41e2590d6f0e4b8542244192a56e022.
* the generated files are now located under target/generated-sources/javacc
* Note: The files that are left in
src/main/java/net/sourceforge/pmd/lang/java/ast
were initially generated by javacc, but were modified
(e.g. every AST*.java class)
/*
I'm doing a sort of State pattern thing here where
this goes into "discarding" mode when it hits an import or package
keyword and goes back into "accumulate mode" when it hits a semicolon.
This could probably be turned into some objects.
*/
I created a Discarder interface, and a couple implementations. One handles keywords/tokens like import and package. Another discards annotation blocks. Another turns CPD on and off based on Strings within an annotation e.g. @SuppressWarnings("CPD-START") and @SuppressWarnings("CPD-END).
There have been two features desired by myself and my colleagues strongly.
1. Ability to ignore sections of source code, such as switch/case statements or parameterized factories.
2. Ability to ignore annotations. More and more modern frameworks use annotations on classes and methods, which can be very redundant and trigger CPD matches.
Just to illustrate, here is a parametrized factory example that can cause a false positive if you ignore literals (String and int values):
if ("A".equals(input)) return new A(1);
if ("B".equals(input)) return new A(2);
if ("C".equals(input)) return new A(3);
if ("D".equals(input)) return new A(4);
if ("E".equals(input)) return new A(5);
False positives cause the CPD reports to be a burden to review and read. Many people start ignoring them when the first 20 entries are 'acceptable'.
For annotations, with J2EE (CDI, Transaction Handling, etc) and Spring (everything) become very redundant. Often classes or methods have the same 5-6 lines of annotations. This causes false positives.
For CPD-START and CPD-END, it would be much more desirable to use comments rather than annotations. However, the parser drops comments and they can not be processed by the JavaTokenizer. The current approach forces entire methods or classes to be skipped, rather than some blocks. However, I feel its an improvement in cases like parameterized factories or case/switch statements where I don't want them polluting my results.
