# -*- coding: utf-8 -*-
"""
Created on Mon Sep 16 20:15:26 2024

@author: Arnd Helmut Hafner
"""
#Adds the original punctuation marks("┘") to the transcribtion of "爲獄等状"
#Manually corrected results can be found in Igokutojo(Zenjoho).txt

import sys # for sys.exit after error
import re
from re import compile, search

def testtitle(line):
    """
    Tests title lines for read-functions

    Parameters
    ----------
    line : str

    Returns
    -------
    title : Boolean.

    """

    title = False
    condition = compile("^([〇一二三四五六七八九]+　)|待考殘𥳑")
    if re.search(condition, line):
        title = True
    return(title)

def testnumber(line):
    """
    Reads slip number of text-file with original punctuation

    Parameters
    ----------
    line : str

    Returns
    -------
    slipn : str
    """
    
    m = re.search('[^\t]+\t([^\t]+)', line)
    if m:return(m.group(1))
    else:sys.exit("mismatch in testnaumber for {}".format(line))

def readori(filename):
    """
    Reads the text-file with the original punctuation marks
    
    Parameters
    ----------
    filename : str
    
    Returns
    -------
    textori : dict
        {slipnumber: line}
    """
    
    slipn = ""
    textori = {}
    datafile = open(filename, 'r', encoding='utf-8-sig')
    for line in datafile:
        line = re.sub('\n','',line)
        if line == '' or testtitle(line):continue
        if "\t" in line:
            slipn = testnumber(line)
            continue
        textori[slipn] = line
    datafile.close()
    return(textori)

def insertpunc(newline,oldline,slipn,logfile):
    
    templine = newline
    ambig = False#置換に疑問が生じた時に置換を中止してログファイルに警告を出す
    pm = '。，、：；？！'#句読点
    hit = re.findall('.┘.',oldline)
    r = 0
    if re.search('┘.┘',oldline):
        logfile.write('Slip {} might have some misrepresentations because of vincinity of punctuation marks\n'.format(slipn))
    while ambig == False and r < len(hit):
        A,B = hit[r].split('┘')
        reg = compile(A + '([^' + pm + A +']*?)([' +pm + ']*)' + B)
        if len(re.findall(reg,newline)) > 1:
            logfile.write('Slip {} might have some misrepresentations because of text similarities\n'.format(slipn))
        m = re.search(reg,newline)
        if m:
            okikaemae = m.group(0)
            if m.group(1) and m.group(2):
                okikaego = A + m.group(1) + '┘（' + m.group(2) + '）' + B
            elif m.group(2):
                okikaego = A + '┘（' + m.group(2) + '）' + B
            elif m.group(1):
                okikaego =  A + m.group(1) + '┘' + B
            else:
                okikaego = A + '┘' + B
            newline = newline.replace(okikaemae, okikaego,1)
            r += 1
        else:
            ambig = True
        
        
    if ambig :
        logfile.write("""Replacement for slip {} failed.\nOldline is: {}\nNewline is: {}\nReplacement result is: {}\n""".format(
                      slipn,oldline,templine,newline))
        return(templine)
    return(newline)

def substpunct(slipn, line, textori,logfile):
    """
    Comparing text-lines and inserting original punctuation marks

    Parameters
    ----------
    slipn : str
    line : str
    textori : dict
        {slipnumber: line(ori)}
    logfile : IOWrapper

    Returns
    -------
    newline : str
        'slipn + \t + line'

    """
    if slipn not in textori.keys():
        logfile.write('slipnumber not found in textori:{}\n'.format(slipn))
    elif '┘' in textori[slipn]:
        line = insertpunc(line,textori[slipn],slipn,logfile)
    return('{}\t{}'.format(slipn,line))
        

if __name__ == "__main__":
    #reading text-file with original punctuation
    textori = readori("Igokutojo(Hakubun).txt")
    #reading text-file with new punctuation and adding original punctuation
    datafile = open('Igokutojo(Zenjoho).txt', 'r', encoding='utf-8-sig')
    outputfile = open('output.txt','w', encoding='utf-8-sig')
    logfile = open('log.txt','w', encoding='utf-8-sig')
    for line in datafile:
        line = re.sub('\n','',line)
        #checking and outputing title-lines and empty lines
        if line == '' or testtitle(line):
            outputfile.write('{}\n'.format(line))
            continue
        #discerning slipnumbers from text and substituting punctuation marks
        m = re.search('^([^\t]+)\t([^\t]+)$',line)
        if m:
            outputfile.write('{}\n'.format(substpunct(m.group(1),m.group(2),textori,logfile)))
        else:
            sys.exit("mismatch in discerning slipnumbers for {}".format(line))
    datafile.close()
    outputfile.close()
    logfile.close()