Grammalecte  Diff

#!python3

import os
import traceback
import pkgutil
import re
from itertools import chain
from functools import wraps
import time

#import logging
#logging.basicConfig(filename="suggestions.log", level=logging.DEBUG)
import logging
logging.basicConfig(filename="suggestions.log", level=logging.DEBUG)

from . import str_transform as st
from . import char_player as cp
from .echo import echo


def timethis (func):
    "decorator for the execution time"
    @wraps(func)
    def wrapper (*args, **kwargs):
        fStart = time.time()
        result = func(*args, **kwargs)
        fEnd = time.time()
        print(func.__name__, fEnd - fStart)
        return result
    return wrapper


class SuggResult:

    def __init__ (self, sWord, nDistLimit=-1):
        self.sWord = sWord
        self.sCleanWord = cp.cleanWord(sWord)
        self.nDistLimit = nDistLimit  if nDistLimit >= 0  else  (len(sWord) // 3) + 1
        self.nMinDist = 1000
        self.aSugg = set()
        self.dSugg = { 0: [],  1: [],  2: [] }

    def addSugg (self, sSugg, nDeep=0):
        "add a suggestion"
        if sSugg not in self.aSugg:
            nDist = st.distanceDamerauLevenshtein(self.sCleanWord, cp.cleanWord(sSugg))
            if nDist <= self.nDistLimit:
                if nDist not in self.dSugg:
                    self.dSugg[nDist] = []
                self.dSugg[nDist].append(sSugg)
                logging.info((nDeep * "  ") + "__" + sSugg + "__")
                if nDist < self.nMinDist:
                    self.nMinDist = nDist
                self.nDistLimit = min(self.nDistLimit, self.nMinDist+2)

    def getSuggestions (self, nSuggLimit=10, nDistLimit=-1):
        "return a list of suggestions"
        lRes = []
        for lSugg in self.dSugg.values():
            lRes.extend(lSugg)
            if len(lRes) > nSuggLimit:
                break
        lRes = list(cp.filterSugg(lRes))
        if self.sWord.istitle():
            lRes = list(map(lambda sSugg: sSugg.title(), lRes))
        return lRes[:nSuggLimit]


class IBDAWG:
    """INDEXABLE BINARY DIRECT ACYCLIC WORD GRAPH"""

    def __init__ (self, sDicName):
        self.by = pkgutil.get_data(__package__, "_dictionaries/" + sDicName)
        if not self.by:

    @timethis
    def suggest (self, sWord, nMaxSugg=10):
        "returns a set of suggestions for <sWord>"
        sPfx, sWord, sSfx = cp.cut(sWord)
        nMaxDel = len(sWord) // 5
        nMaxHardRepl = max((len(sWord) - 5) // 4, 1)
        oSuggResult = SuggResult(sWord)
        aSugg = self._suggest(sWord, nMaxDel=nMaxDel, nMaxHardRepl=nMaxHardRepl)
        self._suggest(oSuggResult, sWord, nMaxDel=nMaxDel, nMaxHardRepl=nMaxHardRepl)
        if sWord.istitle():
            aSugg.update(self._suggest(sWord.lower(), nMaxDel=nMaxDel, nMaxHardRepl=nMaxHardRepl))
            self._suggest(oSuggResult, sWord.lower(), nMaxDel=nMaxDel, nMaxHardRepl=nMaxHardRepl)
            aSugg = set(map(lambda sSugg: sSugg.title(), aSugg))
        elif sWord.islower():
            aSugg.update(self._suggest(sWord.title(), nMaxDel=nMaxDel, nMaxHardRepl=nMaxHardRepl))
        aSugg = cp.filterSugg(aSugg)
            self._suggest(oSuggResult, sWord.title(), nMaxDel=nMaxDel, nMaxHardRepl=nMaxHardRepl)
        aSugg = oSuggResult.getSuggestions()
        sCleanWord = cp.cleanWord(sWord)
        aSugg = sorted(aSugg, key=lambda sSugg: cp.distanceDamerauLevenshtein(sCleanWord, cp.cleanWord(sSugg)))[:nMaxSugg]
        if sSfx or sPfx:
            # we add what we removed
            return list(map(lambda sSug: sPfx + sSug + sSfx, aSugg))
        return aSugg

    def _suggest (self, sRemain, nMaxDel=0, nMaxHardRepl=0, nDeep=0, iAddr=0, sNewWord="", bAvoidLoop=False):
    def _suggest (self, oSuggResult, sRemain, nMaxDel=0, nMaxHardRepl=0, nDeep=0, iAddr=0, sNewWord="", sAction="", bAvoidLoop=False):
        "returns a set of suggestions"
        # recursive function
        #logging.info((nDeep * "  ") + sNewWord + ":" + sRemain)
        #logging.info((nDeep * "  ") + sNewWord + ":" + sRemain + " · " + sAction)
        aSugg = set()
        if not sRemain:
            if int.from_bytes(self.byDic[iAddr:iAddr+self.nBytesArc], byteorder='big') & self._finalNodeMask:
                #logging.info((nDeep * "  ") + "__" + sNewWord + "__")
                aSugg.add(sNewWord)
                oSuggResult.addSugg(sNewWord)
            for sTail in self._getTails(iAddr):
                #logging.info((nDeep * "  ") + "__" + sNewWord+sTail + "__")
                aSugg.add(sNewWord+sTail)
            return aSugg
                oSuggResult.addSugg(sNewWord+sTail)
            return
        cCurrent = sRemain[0:1]
        for cChar, jAddr in self._getSimilarArcs(cCurrent, iAddr):
            aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, jAddr, sNewWord+cChar))
            self._suggest(oSuggResult, sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, jAddr, sNewWord+cChar, "*")
        if not bAvoidLoop: # avoid infinite loop
            if cCurrent == sRemain[1:2]:
                # same char, we remove 1 char without adding 1 to <sNewWord>
                aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord))
                self._suggest(oSuggResult, sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, cCurrent+"/2")
            else:
                # switching chars
                aSugg.update(self._suggest(sRemain[1:2]+sRemain[0:1]+sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                self._suggest(oSuggResult, sRemain[1:2]+sRemain[0:1]+sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, "><",True)
                # delete char
                if nMaxDel > 0:
                    aSugg.update(self._suggest(sRemain[1:], nMaxDel-1, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                    self._suggest(oSuggResult, sRemain[1:], nMaxDel-1, nMaxHardRepl, nDeep+1, iAddr, sNewWord, "-"+cCurrent, True)
            # Phonetic replacements
            for sRepl in cp.d1toX.get(cCurrent, ()):
                aSugg.update(self._suggest(sRepl + sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                self._suggest(oSuggResult, sRepl + sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, cCurrent+">"+sRepl, True)
            for sRepl in cp.d2toX.get(sRemain[0:2], ()):
                aSugg.update(self._suggest(sRepl + sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                self._suggest(oSuggResult, sRepl + sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain[0:2]+">"+sRepl, True)
            # Hard replacements
            if nDeep > 3 and nMaxHardRepl and len(sRemain) >= 2:
                for nVal, kAddr in self._getArcs1(iAddr):
                    if nVal in self.dCharVal:
                        cChar = self.dCharVal[nVal]
                        if cChar not in cp.d1to1.get(cCurrent, ""):
                            aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl-1, nDeep+1, kAddr, sNewWord+cChar, True))
                            self._suggest(oSuggResult, sRemain[1:], nMaxDel, nMaxHardRepl-1, nDeep+1, kAddr, sNewWord+cChar, "[["+cChar+"]]", True)
            # end of word
            if len(sRemain) == 2:
                for sRepl in cp.dFinal2.get(sRemain, ()):
                    aSugg.update(self._suggest(sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                    self._suggest(oSuggResult, sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain + " >> " + sRepl, True)
            elif len(sRemain) == 1:
                aSugg.update(self._suggest("", nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True)) # remove last char and go on
                self._suggest(oSuggResult, "", nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain + " [last char removed] ", True) # remove last char and go o
                for sRepl in cp.dFinal1.get(sRemain, ()):
                    aSugg.update(self._suggest(sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                    self._suggest(oSuggResult, sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain + " >> " + sRepl, True)
        return

    @timethis
    def suggest2 (self, sWord, nMaxSugg=10):
        "returns a set of suggestions for <sWord>"
        sPfx, sWord, sSfx = cp.cut(sWord)
        oSuggResult = SuggResult(sWord)
        self._suggest2(oSuggResult)
        aSugg = oSuggResult.getSuggestions()
        if sSfx or sPfx:
            # we add what we removed
            return list(map(lambda sSug: sPfx + sSug + sSfx, aSugg))
        return aSugg

    def _suggest2 (self, oSuggResult, nDeep=0, iAddr=0, sNewWord=""):
        #logging.info((nDeep * "  ") + sNewWord)
        if nDeep >= oSuggResult.nDistLimit:
            sCleanNewWord = cp.cleanWord(sNewWord)
            if st.distanceSift4(oSuggResult.sCleanWord[:len(sCleanNewWord)], sCleanNewWord) > oSuggResult.nDistLimit:
                return
        if int.from_bytes(self.byDic[iAddr:iAddr+self.nBytesArc], byteorder='big') & self._finalNodeMask:
            #logging.info((nDeep * "  ") + "__" + sNewWord + "__")
            oSuggResult.addSugg(sNewWord, nDeep)
        for cChar, jAddr in self._getCharArcs(iAddr):
            self._suggest2(oSuggResult, nDeep+1, jAddr, sNewWord+cChar)
        return

    def _getCharArcs (self, iAddr):
        "generator: yield all chars and addresses from node at address <iAddr>"
        for nVal, jAddr in self._getArcs(iAddr):
            if nVal < self.nChar:
                yield (self.dCharVal[nVal], jAddr)

    def _getSimilarArcs (self, cChar, iAddr):
        "generator: yield similar char of <cChar> and address of the following node"
        for c in cp.d1to1.get(cChar, [cChar]):
            if c in self.dChar:
                jAddr = self._lookupArcNode(self.dChar[c], iAddr)
                if jAddr: