Languages should natively support something like FullForm. I'm trying to integrate clang AST with mathematica's FullForm with much frustration.

I think this is a useful question but not in it's current form at all, thanks to the legal issues around watermark removal. I was thinking of rewriting the question to instead talk about removing text from an image, and replacing the examples given with other images and removing any reference to watermarks... mathematica.stackexchange.com/questions/95512/…

Does anyone have any thoughts?

As it is, I've gone ahead and made the edits anyway.

I hope this shifts the focus away from watermark removal (legality dubious) and towards text removal, which is often what academic studies of inpainting look at.

I think there should already be questions about this, but I cannot find them ...

How can I create an expression type that formats in a special way, similar to InterpolatingFunction, FittedModel, etc. ?

Before I ask on the main site, I wanted to make sure I didn't miss anything obvious.

@Szabolcs Interpretation perhaps

@Pickett Or rather InterpretationBox

@Szabolcs Yes, you're right

@Pickett Do you know what does the SelectWithContents option of InterpretationBox, TagBox, etc. do? Builtin things that do this set it to True.

This is what I have so far:

MakeBoxes[
  array[arr_List],
  form : (StandardForm | TraditionalForm)
  ] := With[{boxes = RowBox[{"array", "[",
      ToBoxes[Panel@Row[
         {Length[arr], " elements"}
         ], form]
      , "]"}]},
  InterpretationBox[
   boxes,
   array[arr],
   SelectWithContents -> True
   ]
  ]

This is a dummy example for some "datatype" that looks like array[someList], it is formatted in a special way, and the formatted version is copyable and generally usable.

Also convertible to StandardForm/TraditionalForm/InputForm with keyboard shortcuts.

Seems to work, but I have a bad feeling that I didn't consider everything ...

I don't know, but it's an interesting question.

@VitaliyKaurov I started a few Kaggle entries a few years ago with Mathematica. But I ended up just taking a job building statistical models and data analysis instead. The problem with those competitions is that you don't have to work hard to get a good score at the beginning, but then you have to work harder and harder as the teams leap frog each other. So you probably end up working harder than the prize money is worth. Great deal for the host companies though.

So here it goes:

I posted because I am not confident about this and I won't be surprised if it turns out that my implementation is fragile and will break in some cases.

It's coming together nicely:

Can I get some help in re-opening this question. It doesn't involve epicycloids, and I have an answer I want to post that proves it.

@VitaliyKaurov one can prototype with Mathematica for Kaggle stuff, but, I think you can't win if you can't submit your solution in a free/open/whatever language like Python, R, Julia

and "eventually translating" the solution is not certain that will work, given the undocumented details of some mma functions

Winning is unlikely though so for fun its ok

@m_goldberg I voted to reopen of course, I look forward to seeing your answer.