import re, unicodedata, sys

if sys.maxunicode == 65535:
    raise RuntimeError("need UCS-4 Python")

def gen_category(cats):
    for i in range(0, 0x110000):
        if unicodedata.category(unichr(i)) in cats:
            yield(i)

def gen_bidirectional(cats):
    for i in range(0, 0x110000):
        if unicodedata.bidirectional(unichr(i)) in cats:
            yield(i)

def compact_set(l):
    single = []
    tuple = []
    prev = None
    span = 0
    for e in l:
        if prev is None:
            prev = e
            span = 0
            continue
        if prev+span+1 != e:
            if span > 2:
                tuple.append((prev,prev+span+1))
            else:
                for i in range(prev, prev+span+1):
                    single.append(i)
            prev = e
            span = 0
        else:
            span += 1
    if span:
        tuple.append((prev,prev+span+1))
    else:
        single.append(prev)
    tuple = " + ".join(["range(%d,%d)" % t for t in tuple])
    if not single:
        return "set(%s)" % tuple
    if not tuple:
        return "set(%s)" % repr(single)
    return "set(%s + %s)" % (repr(single),tuple)

############## Read the tables in the RFC #######################

data = open("rfc3454.txt").readlines()

tables = []
curname = None
for l in data:
    l = l.strip()
    if not l:
        continue
    # Skip RFC page breaks
    if l.startswith("Hoffman & Blanchet") or\
       l.startswith("RFC 3454"):
        continue
    # Find start/end lines
    m = re.match("----- (Start|End) Table ([A-Z](.[0-9])+) -----", l)
    if m:
        if m.group(1) == "Start":
            if curname:
                raise RuntimeError("Double Start", (curname, l))
            curname = m.group(2)
            table = {}
            tables.append((curname, table))
            continue
        else:
            if not curname:
                raise RuntimeError("End without start", l)
            curname = None
            continue
    if not curname:
        continue
    # Now we are in a table
    fields = l.split(";")
    if len(fields) > 1:
        # Drop comment field
        fields = fields[:-1]
    if len(fields) == 1:
        fields = fields[0].split("-")
        if len(fields) > 1:
            # range
            try:
                start, end = fields
            except ValueError:
                raise RuntimeError("Unpacking problem", l)
        else:
            start = end = fields[0]
        start = int(start, 16)
        end = int(end, 16)
        for i in range(start, end+1):
            table[i] = i
    else:
        code, value = fields
        value = value.strip()
        if value:
            value = [int(v, 16) for v in value.split(" ")]
        else:
            # table B.1
            value = None
        table[int(code, 16)] = value

########### Generate compact Python versions of the tables #############

print """# This file is generated by mkstringprep.py. DO NOT EDIT.
\"\"\"Library that exposes various tables found in the StringPrep RFC 3454.

There are two kinds of tables: sets, for which a member test is provided,
and mappings, for which a mapping function is provided.
\"\"\"

import unicodedata
"""

print "assert unicodedata.unidata_version == %s" % repr(unicodedata.unidata_version)

# A.1 is the table of unassigned characters
# XXX Plane 15 PUA is listed as unassigned in Python.
name, table = tables[0]
del tables[0]
assert name == "A.1"
table = set(table.keys())
Cn = set(gen_category(["Cn"]))

# FDD0..FDEF are process internal codes
Cn -= set(range(0xFDD0, 0xFDF0))
# not a character
Cn -= set(range(0xFFFE, 0x110000, 0x10000))
Cn -= set(range(0xFFFF, 0x110000, 0x10000))

# assert table == Cn

print """
def in_table_a1(code):
    if unicodedata.category(code) != 'Cn': return False
    c = ord(code)
    if 0xFDD0