Log line numbers of executed java code

Question

I am writing part of a PHP web application (which will be used in a high school bug finding contest) where the user must find bugs in a given Java program. As a part of this, when the Java program executes, we want to highlight the lines of the source of the Java program where the code has executed. To do this, all we need are the line numbers of the source that have been executed, that is, the code path (or is it called code coverage?). We will highlight the lines in the source file using the line numbers.

We will be using PHP's shell-exec() to execute the Java program and the tool to get the code path (whatever that will be). What is the easiest way of getting the line numbers of code path?

Thank you very much!

Here is a picture that describes what we would like

Answer 1

PHP interperts the code, which means it runs over the source each time you run the program. This has the benefit of blowing up as the code is read (which makes line number printouts trivial); however, it often is expensive in other ways, as you cannot optimize deeply (or do any pre-runtime error checking).

Java compiles its code into a JVM assembly language called "bytecode." This means that what is running doesn't generally have access to (or even use) the source code. That said, there are techniques. A compiled Java class has the ability to add "extra data" and one of those "extra data elements" is a line number table, which is an index allowing someone running the assembly to "look up" the line number as the compiler recorded it.

This generally works ok, with the considerations that: compilers often don't mark up every instruction, the source code may not be available, optimization might make certain inner chunks of code not function in ways that facilitate pointing to the input code text.

How code coverage tools "fix" this is that they generally insert into the code (at the assembly level) a large number of commands that effectively act as logging statements to a format that allows the tool to determine which path through the code was actually followed. This is then mapped back through the line number table as best as possible and then used to highlight lines within the original source file.

If you want something with finer resolution (something that can process which portion of a line was executed) then you need to dig deeper. Eventually you might even consider writing your own compiler (or compiler extension) which will store your own custom line number table that overcomes the shortcomings of the current solutions.

Tricks like throwing exceptions (as Shiven has mentioned) and parsing out the line number do work; however, they pollute your code with odd exception processing for items that really aren't exceptional, just to "get the line number". Due to the code clutter and the generally poorer runtime performance of exceptions, I tend to avoid such solutions (but they do work).

Anyway, hopefully this will give you a view as to why it doesn't always work exactly the same way as PHP.

Answer 2

Take a look at Cobertura. It computes coverage and stuff like that, and if it doesn't already do it, it should be relatively easy to add the line number collecting to it. There's a very hackish attempt to do that, but that's so slow that you may not be able to use it in production https://bitbucket.org/jowu/myriapod/wiki/Home