[Feature Request]: support isAlpha et al. on codepoints

[vasslitvinov]

Summary of Feature

Description:

I want to be able to invoke the predicates isAlpha(), isDigit(), and so on, within a loop over a string's codepoints.

Currently these predicates are available only on whole strings and include an on-statement. So if I want to use them within my loop, there is a prohibitive amount of overhead of wrapping my codepoint into a string and performing an unnecessary on within isDigit().

Is this issue currently blocking your progress?

No.

Code Sample

Consider the implementation of Arkouda's Strings.isdecimal()

It checks whether each (unicode) character of myString either satisfies isDigit() or is a numeric subscript or superscript.

Currently the implementation does expensive computations if myString.isDigit() fails. Instead I would like for it to do simply:

for cp in myString.codepoints() do
  if ! isCodepointDigit(cp) &&
     // whatever other checks I need to do
     ! isNumericSubSuperScript(cp) then
    return false;
return true;

Ideally, if myString is long enough, I would like it to be a forall loop (see #19112) with a eureka exit (#12700).

Also if the config param useCachedNumCodepoints is true, I would like to check whether myString is an ASCII string and, if so, run much simpler/more efficient code. Currently I am reluctant to to use this route because string.isASCII() is O(string size) if this param is false and I should not be checking this param from Arkouda because it is undocumented / unstable.

[mppf]

There are already functions codepoint_isAlpha / codepoint_isDigit / etc that accept an int(32) which is the type yielded by string.codepoints(). But they are private and would need different names, at the very least, if they are to be public.

We could consider making them methods on integers (possibly, integers of any width) to support code like myInteger.isDigit() and for cp in myString.codepoints() do if ! isCodepointDigit(cp) ....

[vasslitvinov]

Right, the implementation is there already.

We could consider making them methods on integers

This might be not very intuitive. For example, one may reasonably expect 1.isDigit() to be true.