Fix incorrect class name generation from file path when path contains /java

[sweetStreet]

The original implementation of this method led to incorrect class name generation when the file path contained the substring /java.

Previously, a file path like "com/amazon/sqs/javamessaging/AmazonSQSExtendedClient.java" would incorrectly generate the class name "com.amazon.sqsmessaging.AmazonSQSExtendedClient". This was due to the premature removal of the /java segment.

[I-Al-Istannen]

That is true :) Won't really happen in the problem domain this codebase is used in, but still worth fixing. I guess a nice stripSuffix method is missing in the JDK string…

How did you find this? Just searched on github for the two replacement patterns? It looks like you are also working on automatic program testing/analysis, so that sounds interesting :)

[sweetStreet]

I see. Thanks for the feedback! Good point, even though we can use replaceAll(".java$", "") You're right, haha. I simply searched for the pattern. It's a byproduct of my automatic test generation project. We proposed inline tests that check the correctness of statements. To automatically generate them, our tool executes unit tests and collects the runtime values of variables in a statement. One value I manually inspected is a file path that contains "/java", so I think there might be some projects in GitHub that do not consider this case. However, for this pull request, a unit test should be enough.

…

On Thu, Dec 21, 2023 at 4:44 PM I-Al-Istannen ***@***.***> wrote: Merged #39 <#39> into master. — Reply to this email directly, view it on GitHub <#39 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEKHIQHCKTDNFBCMAJR7WB3YKS3UHAVCNFSM6AAAAABA646ZR2VHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJRGMZDAOBQHE3DGOA> . You are receiving this because you authored the thread.Message ID: ***@***.*** com>

[I-Al-Istannen]

Ah, neat. Have you seen https://arxiv.org/pdf/2212.11077.pdf or https://arxiv.org/pdf/2302.04547.pdf? How do you collect runtime values? I wrote some PoC code for that a while ago.

The other part that would interest me is how you create the tests. Do you synthesize java code or reconstruct the objects and values from JSON/XML/Protobuf/whatever? I.e. do you generate a test like

Bar foo = new Bar(x, y);
assert(foo.whatever() == 20);

or

Bar foo = fromJson("constant");
assert(foo.whatever() == fromJson("expected"));

[sweetStreet]

Thanks for recommending these two papers!! The idea of using runtime information to augment code diff is interesting. We collected runtime values and serialized them like these papers (with instrumentation). The difference is that your repo uses bytecode instrumentation through ASM while mine uses source code level instrumentation with JavaParser ( https://github.com/EngineeringSoftware/exli-internal/tree/main/java/raninline ). For primitive types and String, we directly stored the constants (like the first example), but for other objects, we dumped them into XML files (like the second example). The limitation of saving objects into files may make the tests not readable.

…

On Fri, Dec 22, 2023 at 5:33 AM I-Al-Istannen ***@***.***> wrote: Ah, neat. Have you seen https://arxiv.org/pdf/2212.11077.pdf or https://arxiv.org/pdf/2302.04547.pdf? How do you collect runtime values? I wrote some PoC code <https://github.com/I-Al-Istannen/traceaman> for that a while ago. The other part that would interest me is how you create the tests. Do you synthesize java code or reconstruct the objects and values from JSON/XML/Protobuf/whatever? I.e. do you generate a test like Bar foo = new Bar(x, y);assert(foo.whatever() == 20); or Bar foo = fromJson("constant");assert(foo.whatever() == fromJson("expected")); — Reply to this email directly, view it on GitHub <#39 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEKHIQGLDXPBOMWFGN7TT7TYKVVXJAVCNFSM6AAAAABA646ZR2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRXGU3TSOBTHE> . You are receiving this because you authored the thread.Message ID: ***@***.***>