Ask HN: Would you pay for a code similarity detection tool?

Clickable link for the image: https://dl.dropboxusercontent.com/u/30225560/ase.png

Your example looks similar to what ReSharper does: http://blog.jetbrains.com/dotnet/2009/12/11/resharper-50-pre...

I see this being useful in court when project A accuses project B in stealing major portion of their code and relabeling it as created by B + changing the license to incompatible.

Some exist already:

http://www.harukizaemon.com/simian/ http://dickgrune.com/Programs/similarity_tester/ https://github.com/silviocesare/Clonewise

Personally no.

Sometimes I deliberately choose not to abstract away simple logic (as in your screenshot) for performance reasons, or just to reduce a complex dependency chain (I'm all for code reuse, but I do also believe in a balance when writing portable code). And the instances where the required logic is more complex, I'd know to be looking for a module before writing my own code.

Because of these reasons, I couldn't even see myself using this tool if it was free.

However, this does sound an interesting project and I think you should still proceed with it regardless of my feedback as, even if it doesn't because a profitable exercise, I could see this becoming a future must-have feature for IDEs - eg for code refactoring. In fact maybe you could extend this tool to analyse repeated code within a project and suggest abstracting that out to a function (that's a tool I probably would use on larger code bases!)

As a professor, I always missed an easy-to-use and modern tool for that. I didn't research this topic much, but not finding anything easy and ready to use is probably a market opportunity (if there is a market).

Then more code obfuscation takes place, and the scanner needs additional layers of deobfuscation and this repeats itself until it takes longer to compare code then to write a new application from scratch.

Your lefthand example works on "i" as a global variable - it has additional effects compared to the example on the right.

Tools like this are common within Java as Java and similar are easier to this kind of analysis on.

IntelliJ IDEA is long reigning champ in this area. (Which is a product I would pay for.)

I'm not sure if I would buy it, but it could be a very nice way of introducing to a new language/framework. When learning a new environment, one doesn't know all the functions and it's very easy to write code that is already implemented.

So, maybe try asking someone who is into programming training programs.

Tools are very hard to sell. Especially very specialized tools. And tools for programmers often come with the end-user price of 0, so justifying anything above that is hard.

For me personally, I don't see the need to even use such a tool, let alone paying for it. Many programmers seem to focus on code, but in my experience, that's usually not the problem you have when things go bad.

Mirror, in case the OP's Dropbox account hits its bandwidth limit: http://i.imgur.com/ciEUQth.png

Would love to see this integrated into an IDE such that it is capable of detecting whether or not what you're typing (or something close to it) already exists. For companies with large codebases this is a major enough problem that a decent solution would have a compelling enough ROI to shell out some cash.

Personally, no. Even if it was an IDE feature, I would rather just learn it if I use it frequently.

> basically an open REST service

Be careful with that, it restricts your customers to a certain lightmillisecond distance from your server - if the plugin is intended to be real-time.

There is a project called jsinspect which does what you are talking about. I dont know if yours would be a bit more robust than this one.. But just wanted to provide this - so you know what you are competing with.

https://github.com/danielstjules/jsinspect

Generally, no. There is no money to be made making tools.

I was just reading about plagiarism detection recently http://theory.stanford.edu/~aiken/moss/

How does your techniques compare to Winnowing? http://theory.stanford.edu/~aiken/publications/papers/sigmod...

I would not stuff like that exist for free and I rarely use that kind of tools.

How did you choose to work on this idea?

There's a YC company doing something very much like this, but against compiled binaries.

Well there are plagiarism detection tools that education institutions use to detect cheating in CS classes.

There are several machine learning solutions that can look for your IP in either source code or binary form even on an abstract (algorithm) level.

Many tools exist like that, the real question is what exactly your target audience is?

If you are scanning public repo's what service does the tool actually provides, and no detecting code similarity isn't the answer here.

Selling something isn't really an issue as long as it has a clear purpose which i don't think your idea has at this moment in time.

Would you pay for it yourself? I'm struggling to see what a normal developer would use it for. I think your main customer will be high schools, colleges and universities. Not so much professional developers.

Commercial consumption of this idea is around verifying licensing. My employer runs a tool against our internal repositories looking for code it's aware of on the internet. When it flags a match, a human looks at our use of the publicly available code and verifies our compliance with the license that came with the code.

I, a software developer, wouldn't buy such a thing, but there's certainly enterprise/corporate demand for it.

Have you looked at Code Climate?

https://codeclimate.com/

(I work mostly on Python, so this is my opinion based on that)

Most of the code that I write that would be duplicating the functionality in a library would be doing so because we don't need the extra functionality. For example, I need to pluralise a small set of words, so I write 3-4 clause if-statement and append some "s" characters because I don't see the need to use python-inflection. The latter is massively more complex, so unlikely to be detected as the same thing.

Sure, there might be a few matches, but I suspect they will mostly be helper functions within libraries, rather than the public API of libraries.

I would prefer not to send code to a web service for detection, although not totally against it. In many companies this would not be allowed, either through policy, firewalls, exfiltration detection systems, or lack of internet on development machines (I know people who work, or have worked myself, in all of these situations).

Something I think would be far more valuable, and possibly more realistic as well, is local detection that highlights possibly duplicated code in a codebase. I find little snippets (1-2 lines) that have been duplicated on a fairly regular basis, and if I could identify those to be extracted out into re-usable methods, that would be amazing.

Would I pay money for it? Probably not, it's not that much of a problem, and all those sorts of tools are usually open source anyway. Unfortunately that's my expectation now.