My advice would be "don't even think of using a regex". This type of problem is really not in the regex domain because there are too many variables (no pun intended) to consider.

As an example, take your requirement #1 to see if there is a "return" statement within a "finally" block. For a start you need to make sure that you pick up the "return" that is the statement and not the word in the comment (see the line above in your example) - this involves coding into the pattern the ability to skip over comments that can occur anywhere within the source text.

Once you have that solved, you need to be able to find the limits of the "finally" block. If you look at your example, that means finding the "}" that matches the "{" that occurs immediately (accounting for any intervening whitespace and comments) after the "finally" keyword (again being careful not to pick up some extraneous character string that is not the required keyword). To do that you need to be able to count the "{"s and the corresponding "}"s and make sure they match. The vast majority of regex variants cannot do this - it is referred to as the "matching parentheses problem" - and, as far as I know, only the .NET and PCRE regex libraries have the necessary (incompatible) extensions to the pattern syntax to allow this [Perl 6 is supposed to have this ability as well but I'm not sure the current state of the Perl 6 for production use].

Also, considering your second problem, you will need to consider all of the possible code paths through the program. For example, if you have an "if/then/else" structure and the "object=null" statement occurs in the "else" clause, does this meet your needs or not as the code path through the "then" block would not set the object to null.

If you really want to do this, I would suggest that you write a parser for your target language (you may even find that one exists in some language such as Lex that you can use directly) and then start processing the semantic tree that it outputs.

Susan