Grammalecte  Diff

def readFile (spf):
    print(" < Read lexicon: " + spf)
    if os.path.isfile(spf):
        with open(spf, "r", encoding="utf-8") as hSrc:
            for sLine in hSrc:
                sLine = sLine.strip()
                if sLine and not sLine.startswith("#"):
                    yield sLine
    else:
        raise OSError("# Error. File not found or not loadable: " + spf)



class DAWG:
    """DIRECT ACYCLIC WORD GRAPH"""
    # This code is inspired from Steve Hanov’s DAWG, 2011. (http://stevehanov.ca/blog/index.php?id=115)
    # We store suffix/affix codes and tags within the graph after the “real” word.
    # A word is a list of numbers [ c1, c2, c3 . . . cN, iAffix, iTags]
    # Each arc is an index in self.lArcVal, where are stored characters, suffix/affix codes for stemming and tags.
    # Important: As usual, the last node (after ‘iTags’) is tagged final, AND the node after ‘cN’ is ALSO tagged final.

    def __init__ (self, spfSrc, cStemming, sLangCode, sLangName="", sDicName=""):
        print("===== Direct Acyclic Word Graph - Minimal Acyclic Finite State Automaton =====")
        cStemming = cStemming.upper()
        if cStemming == "A":
            funcStemmingGen = st.defineAffixCode
        elif cStemming == "S":
            funcStemmingGen = st.defineSuffixCode
        elif cStemming == "N":
            funcStemmingGen = st.noStemming
        else:
            raise ValueError("# Error. Unknown stemming code: {}".format(cStemming))

        lEntry = []
        lChar = ['']; dChar = {}; nChar = 1; dCharOccur = {}
        lAff  = [];   dAff  = {}; nAff  = 0; dAffOccur = {}
        lTag  = [];   dTag  = {}; nTag  = 0; dTagOccur = {}
        nErr = 0
        
        # read lexicon

        for sLine in readFile(spfSrc):


            sFlex, sStem, sTag = sLine.split("\t")
            addWordToCharDict(sFlex)
            # chars
            for c in sFlex:
                if c not in dChar:
                    dChar[c] = nChar
                    lChar.append(c)
                    nChar += 1
                dCharOccur[c] = dCharOccur.get(c, 0) + 1
            # affixes to find stem from flexion
            aff = funcStemmingGen(sFlex, sStem)
            if aff not in dAff:
                dAff[aff] = nAff
                lAff.append(aff)
                nAff += 1
            dAffOccur[aff] = dCharOccur.get(aff, 0) + 1
            # tags
            if sTag not in dTag:
                dTag[sTag] = nTag
                lTag.append(sTag)
                nTag += 1
            dTagOccur[sTag] = dTagOccur.get(sTag, 0) + 1
            lEntry.append((sFlex, dAff[aff], dTag[sTag]))
        if not lEntry:
            raise ValueError("# Error. Empty lexicon")
        
        # Preparing DAWG
        print(" > Preparing list of words")
        lVal = lChar + lAff + lTag
        lWord = [ [dChar[c] for c in sFlex] + [iAff+nChar] + [iTag+nChar+nAff]  for sFlex, iAff, iTag in lEntry ]
        lEntry = None
        
        # Dictionary of arc values occurrency, to sort arcs of each node
        dValOccur = dict( [ (dChar[c], dCharOccur[c])  for c in dChar ] \
                        + [ (dAff[aff]+nChar, dAffOccur[aff]) for aff in dAff ] \
                        + [ (dTag[tag]+nChar+nAff, dTagOccur[tag]) for tag in dTag ] )
        #with open(spfSrc[:-8]+".valuesfreq.txt", 'w', encoding='utf-8') as hFreqDst:  # DEBUG
        #    for iKey, nOcc in sorted(dValOccur.items(), key=lambda t: t[1], reverse=True):
        #        hFreqDst.write("{}: {}\n".format(lVal[iKey], nOcc))
        #    hFreqDst.close()
        
        self.sFileName = spfSrc
        self.sLangCode = sLangCode
        self.sLangName = sLangName
        self.sDicName = sDicName
        self.nEntry = len(lWord)
        self.aPreviousEntry = []
        DawgNode.resetNextId()
        self.oRoot = DawgNode()

                sEntry = sWord + "\t" + self.funcStemming(sWord, self.lArcVal[nVal])
                for nMorphVal, _ in oNextNode.arcs.items():
                    if not zPattern or zPattern.search(self.lArcVal[nMorphVal]):
                        yield sEntry + "\t" + self.lArcVal[nMorphVal]


    # BINARY CONVERSION
    def createBinary (self, sPathFile, nCompressionMethod, bDebug=False):
        print(" > Write DAWG as an indexable binary dictionary [method: %d]" % nCompressionMethod)
        if nCompressionMethod == 1:
            self.nBytesArc = ( (self.nArcVal.bit_length() + 2) // 8 ) + 1   # We add 2 bits. See DawgNode.convToBytes1()
            self.nBytesOffset = 0
            self._calcNumBytesNodeAddress()
            self._calcNodesAddress1()
        elif nCompressionMethod == 2:

            self._calcNodesAddress3()
        else:
            print(" # Error: unknown compression method")
        print("   Arc values (chars, affixes and tags): {}  ->  {} bytes".format( self.nArcVal, len("\t".join(self.lArcVal).encode("utf-8")) ))
        print("   Arc size: {} bytes, Address size: {} bytes   ->   {} * {} = {} bytes".format( self.nBytesArc, self.nBytesNodeAddress, \
                                                                                                self.nBytesArc+self.nBytesNodeAddress, self.nArc, \
                                                                                                (self.nBytesArc+self.nBytesNodeAddress)*self.nArc ))
        self._writeBinary(sPathFile, nCompressionMethod)
        self._writeAsJSObject(sPathFile, nCompressionMethod)
        if bDebug:
            self._writeNodes(sPathFile, nCompressionMethod)

    def _calcNumBytesNodeAddress (self):
        "how many bytes needed to store all nodes/arcs in the binary dictionary"
        self.nBytesNodeAddress = 1
        while ((self.nBytesArc + self.nBytesNodeAddress) * self.nArc) > (2 ** (self.nBytesNodeAddress * 8)):
            self.nBytesNodeAddress += 1

                for oNextNode in self.lSortedNodes[i].arcs.values():
                    if 1 < (oNextNode.addr - self.lSortedNodes[i].addr) < self.nMaxOffset:
                        nSize -= nDiff
                if self.lSortedNodes[i].size != nSize:
                    self.lSortedNodes[i].size = nSize
                    bEnd = False

    def _writeAsJSObject (self, spfDst, nCompressionMethod, bInJSModule=False, bBinaryDictAsHexString=True):
        if not spfDst.endswith(".json"):
            spfDst += "."+str(nCompressionMethod)+".json"
        byDic = b""
        if nCompressionMethod == 1:
            byDic = self.oRoot.convToBytes1(self.nBytesArc, self.nBytesNodeAddress)
            for oNode in self.lMinimizedNodes:
                byDic += oNode.convToBytes1(self.nBytesArc, self.nBytesNodeAddress)
        elif nCompressionMethod == 2:
            byDic = self.oRoot.convToBytes2(self.nBytesArc, self.nBytesNodeAddress)
            for oNode in self.lSortedNodes:
                byDic += oNode.convToBytes2(self.nBytesArc, self.nBytesNodeAddress)
        elif nCompressionMethod == 3:
            byDic = self.oRoot.convToBytes3(self.nBytesArc, self.nBytesNodeAddress, self.nBytesOffset)
            for oNode in self.lSortedNodes:
                byDic += oNode.convToBytes3(self.nBytesArc, self.nBytesNodeAddress, self.nBytesOffset)

        with open(spfDst, "w", encoding="utf-8", newline="\n") as hDst:
            if bInJSModule:
                hDst.write('// JavaScript\n// Generated data (do not edit)\n\n"use strict";\n\nconst dictionary = ')
            hDst.write(json.dumps({
                            "sHeader": "/pyfsa/",
                            "sLangCode": self.sLangCode,
                            "sLangName": self.sLangName,
                            "sDicName": self.sDicName,
                            "sFileName": self.sFileName,
                            "sDate": self._getDate(),
                            "nEntry": self.nEntry,
                            "nChar": self.nChar,
                            "nAff": self.nAff,
                            "nTag": self.nTag,
                            "cStemming": self.cStemming,
                            "dChar": self.dChar,
                            "nNode": self.nNode,
                            "nArc": self.nArc,
                            "nArcVal": self.nArcVal,
                            "lArcVal": self.lArcVal,
                            "nCompressionMethod": nCompressionMethod,
                            "nBytesArc": self.nBytesArc,
                            "nBytesNodeAddress": self.nBytesNodeAddress,
                            "nBytesOffset": self.nBytesOffset,
                            # JavaScript is a pile of shit, so Mozilla’s JS parser don’t like file bigger than 4 Mb!
                            # So, if necessary, we use an hexadecimal string, that we will convert later in Firefox’s extension.
                            # https://github.com/mozilla/addons-linter/issues/1361
                            "sByDic": byDic.hex()  if bBinaryDictAsHexString  else [ e  for e in byDic ]

                        }, ensure_ascii=False))







            if bInJSModule:
                hDst.write(";\n\nexports.dictionary = dictionary;\n")

    def _writeBinary (self, sPathFile, nCompressionMethod):
        """
        Format of the binary indexable dictionary:
        Each section is separated with 4 bytes of \0
        
        - Section Header:
            /pyfsa/[compression method]
                * compression method is an ASCII string

        - Section Values:
                * a list of strings encoded in binary from utf-8, each value separated with a tabulation
        
        - Section Word Graph (nodes / arcs)
                * A list of nodes which are a list of arcs with an address of the next node.
                  See DawgNode.convToBytes() for details.
        """

        if not sPathFile.endswith(".bdic"):
            sPathFile += "."+str(nCompressionMethod)+".bdic"
        with open(sPathFile, 'wb') as hDst:
            # header
            hDst.write("/pyfsa/{}/".format(nCompressionMethod).encode("utf-8"))
            hDst.write(b"\0\0\0\0")
            # infos

                hDst.write(self.oRoot.convToBytes2(self.nBytesArc, self.nBytesNodeAddress))
                for oNode in self.lSortedNodes:
                    hDst.write(oNode.convToBytes2(self.nBytesArc, self.nBytesNodeAddress))
            elif nCompressionMethod == 3:
                hDst.write(self.oRoot.convToBytes3(self.nBytesArc, self.nBytesNodeAddress, self.nBytesOffset))
                for oNode in self.lSortedNodes:
                    hDst.write(oNode.convToBytes3(self.nBytesArc, self.nBytesNodeAddress, self.nBytesOffset))
            hDst.close()


    def _getDate (self):
        return time.strftime("%Y.%m.%d, %H:%M")

    def _writeNodes (self, sPathFile, nCompressionMethod):
        "for debugging only"
        print(" > Write nodes")

                for oNode in self.lSortedNodes:
                    hDst.write(oNode.getTxtRepr2(self.nBytesArc, self.nBytesNodeAddress, self.lArcVal)+"\n")
            if nCompressionMethod == 3:
                hDst.write(self.oRoot.getTxtRepr3(self.nBytesArc, self.nBytesNodeAddress, self.nBytesOffset, self.lArcVal)+"\n")
                #hDst.write( ''.join( [ "%02X " %  z  for z in self.oRoot.convToBytes3(self.nBytesArc, self.nBytesNodeAddress, self.nBytesOffset) ] ).strip() )
                for oNode in self.lSortedNodes:
                    hDst.write(oNode.getTxtRepr3(self.nBytesArc, self.nBytesNodeAddress, self.nBytesOffset, self.lArcVal)+"\n")
            hDst.close()
    
    def writeResults (self, sPathFile):
        bFileExits = os.path.isfile("_lexicons.res.txt")
        with open("_lexicons.res.txt", "a", encoding='utf-8', newline="\n") as hDst:
            sFormat1 = "{:<12} {:>12} {:>5} {:>8} {:>8} {:>6} {:>8} {:>9} {:>9} {:>15} {:>12} {:>12}\n"
            sFormat2 = "{:<12} {:>12,} {:>5,} {:>8,} {:>8} {:>6,} {:>8,} {:>9,} {:>9,} {:>15,} {:>12,} {:>12,}\n"
            if not bFileExits:
                hDst.write(sFormat1.format("Lexicon", "Entries", "Chars", "Affixes", "Stemming", "Tags", "Values", "Nodes", "Arcs", "Lexicon (Kb)", "Dict (Kb)", "LT Dict (Kb)"))
            hDst.write(sFormat2.format(self.sLangName, self.nEntry, self.nChar, self.nAff, self.cStemming + "FX", self.nTag, self.nArcVal, \
                                       self.nNode, self.nArc, os.path.getsize(self.sFileName), os.path.getsize(sPathFile), \
                                       os.path.getsize("cfsa/dict/{}.dict".format(self.sLangName)) if os.path.isfile("cfsa/dict/{}.dict".format(self.sLangName)) else 0))
            hDst.close()



class DawgNode:
    NextId = 0
    NextPos = 1 # (version 2)