[PySlice]

Yeah, I guess it sucks if you work with Bignums. I would be happy if BigInteger and BigDecimal were promoted String-level language support: still implemented as classes, but with compiler support. They could add indexing [] for collections, but I can live with .get() and .put().

Some people seem to think that if they don't use, then it doesn't matter. In many business applications, you don't even need floating point (you just transfer data to the database and back), anyway.

Are floats and doubles "little features"?

In computer graphics, you need lots of floating point calculations. Are ints and longs "little features"?

In embedded software, you usually don't want to use resizable strings (C doesn't even have this feature built-in), because reallocations are expensive or unavailable. Are resizable strings (or containers) "little features"?

Different programs have different needs. If you haven't stumbled upon a good use case, doesn't mean that it's useless.

[smackfu]

This same logic can be used to defend any "little feature". Language design is all about where you draw those lines.

[PySlice]

That logic is just a starting point (just to avoid the "I don't need it, why can't you be just like me?"), not the whole thing.

Then you have to take into account a few other criteria, such as [1] internal coherence in the language, [2] difficulty in implementation (and explaining to others) [3] what your target audience is, and what they want, [4] what competitors are doing.

I think full operator overloading would hurt [2], but special-casing Bignums would not hurt [1], [2] at all (as I said, JSP EL already have +,-,*,/, etc. , String already have +, +=).

[3] This is what this thread is all about, no need to repeat :-D

And [4]... Well in this case Java is really playing catch up.