New sorting is still strange

[rlwhitcomb]

Even with "natorder-1.1.3.jar":

> s = [ "$", "#", "1", "0", "2" ]
s = ["$", "#", "1", "0", "2"] -> [ "$", "#", "1", "0", "2" ]
> sort s
sort s -> [ "0", "#", "$", "1", "2" ]

I would expect [ "#", "$", "0", "1", "2" ] instead

[rlwhitcomb]

This is tracking issue rlwhitcomb/utilities#547.

[616slayer616]

I had a look and the problem is that "0" is kind of special since it there can be leading zeroes.

The method isSpaceOrSeparatorOrEmpty could be adjusted like this:
return isSpaceOrSeparator(c) || c == '0' || c == '#' || c == '$';
To get it to work. But I don't know what other implications there could be and you would basically have to add all special characters.

The example with the one character strings is very basic and this library is not intended for this purpose.
It makes zero sense to compare single characters with this.

Do you have longer examples which illustrate this issue?
I could test those in this branch: https://github.com/616slayer616/natorder/tree/issue%2316

[rlwhitcomb]

The original reason this came up was in my Calculator language I define special variables for function parameters (for the case of variable number of arguments) named "$#" (for total number of args), "$*" (for the array of arguments), plus "$0", "$1, "$2", etc. for each individual argument. I was trying to sort them and saw the strange order, so basically any prefix to these "#", "*" would cause it. I don't know how common this type of naming would be, but ...

> a = [ '$*', '$#', '$0', '$1', '$2', '$10' ]
a = ['$*', '$#', '$0', '$1', '$2', '$10'] -> [ "$*", "$#", "$0", "$1", "$2", "$10" ]
> sort a
sort a -> [ "$0", "$#", "$*", "$1", "$2", "$10" ]
> a = [ "abc0", "abc#", "abc*", "abc1", "abc2", "abc10", "abc20" ]
a = ["abc0", "abc#", "abc*", "abc1", "abc2", "abc10", "abc20"] -> [ "abc0", "abc#", "abc*", "abc1", "abc2", "abc10", "abc20" ]
> sort a
sort a -> [ "abc0", "abc#", "abc*", "abc1", "abc2", "abc10", "abc20" ]

My thought to fix it was to basically sort symbols in their ASCII order compared to numbers, so the test you mentioned could just be `... || c < '0' (Note: I haven't tried that yet to be sure).

[rlwhitcomb]

Further testing just now shows this weirdness too:

a = ["abc0def", "abc#def", "abc*def", "abc1def", "abc2def", "abc10def", "abc20def"] -> [ "abc0def", "abc#def", "abc*def", "abc1def", "abc2def", "abc10def", "abc20def" ]
> sort a
sort a -> [ "abc#def", "abc*def", "abc1def", "abc2def", "abc10def", "abc20def", "abc0def" ]

where the "0" gets put to the end ...

[616slayer616]

I can't get it to work for everything.
I can do minuscule changes to the method to achieve your desired results but then other tests will fail.
Could introducing "flaovrs" be a solution here?
master...issue#16
"SPECIAL" worked for your first examples and "ZERO" works for your "zero in the middle of a string" example.
If you like it maybe you have better naming ideas

[rlwhitcomb]

That solution sounds like probably the only way to get it to "work", but I'm afraid for my use case it probably won't fly, because in order to set the "right" sort flavor I'd have to analyze the data beforehand, which isn't feasible.... Unfortunately.

[krasa]

It seems similar to this: paour/natorder#14