# 2011 Aug 14
# Copyright Terry Jan Reedy
# proof of concept for codepoint indexing on narrow builds
# by adjusting cp index to get corresponding code unit index

from bisect import bisect_left

class UTF16:
    def __init__(self, text, cpdex = None):
        self.text = text
        if cpdex is None: 
            cpdex = self.make_cpdex()
        # else like text is surrogate pair and cpdex is (0,)
        self.cpdex = cpdex
        self.len = len(text) - len(cpdex)
        
    def make_cpdex(self):
        cpdex = []  # codepoint indexes of upper chars
        i = 0  # codepoint index, not incremented for 2nd surrogates
        t_iter = iter(self.text)
        for c in t_iter:
            oc = ord(c)
            if 0xD800 <= oc <= 0xDBFF:  # first surrogate
                if 0xDC00 <= ord(next(t_iter)) <= 0xDFFF:
                    cpdex.append(i)
                else:
                    raise ValueError("hi surrogate not paired by lo surrogate")
            elif 0xDC00 <= oc <= 0xDFFF:
                raise ValueError("lo surrogate not paired by hi surrogate")
            i += 1
        return cpdex

    def __eq__(self, other):  # temporary kludge
        if type(other) is type(self):
            other = other.text
        # otherwise, it better be a (unicode) string
        return self.text == other
    
    def __len__(self): return self.len
            
    def __getitem__(self, cp):
        cu = cp + bisect_left(self.cpdex, cp)
        c = self.text[cu]
        if 0xD800 <= ord(c) <= 0xDBFF:
            return UTF16(self.text[cu:cu+2], (0,))
        else:
            return c

    def __repr__(self):
        return repr(self.text)
    __str__ = __repr__

tucs2 = 'A\U0001043cBC\U0001042f\U00010445DE\U00010428H'
tutf16= UTF16(tucs2)
tlist = ['A', '\U0001043c','B','C','\U0001042f','\U00010445',
         'D','E','\U00010428','H']
assert len(tutf16) == len(tlist)
tlis2 = [tutf16[i] for i in range(len(tlist))]
for c in tlis2: assert len(c) == 1
assert tlist == tlis2