Grammalecte  Artifact [8db3f5f069]

Artifact 8db3f5f069d37dddba89d93f589dd49c384753dc70e5183d947f2184d2338664:


# Create a Direct Acyclic Rule Graph (DARG)

import re
import traceback
import json

import darg


dACTIONS = {}
dFUNCTIONS = {}


def prepareFunction (s, bTokenValue=False):
    s = s.replace("__also__", "bCondMemo")
    s = s.replace("__else__", "not bCondMemo")
    s = re.sub(r"(morph|analyse|displayInfo)[(]\\(\d+)", 'g_\\1(lToken[\\2+nTokenOffset]', s)
    s = re.sub(r"(select|exclude|define)[(][\\](\d+)", 'g_\\1(lToken[\\2+nTokenOffset]', s)
    s = re.sub(r"(tag_before|tag_after)[(][\\](\d+)", 'g_\\1(lToken[\\2+nTokenOffset]', s)
    s = re.sub(r"(switchGender|has(?:Mas|Fem)Form)[(]\\(\d+)", '\\1(lToken[\\2+nTokenOffset]["sValue"]', s)
    s = re.sub(r"(morph|analyse)\(>1", 'g_\\1(lToken[nLastToken+1]', s)                     # next token
    s = re.sub(r"(morph|analyse)\(<1", 'g_\\1(lToken[nTokenOffset]', s)                     # previous token
    s = re.sub(r"[\\](\d+)\.is(upper|lower|title)\(\)", 'lToken[\\1+nTokenOffset]["sValue"].is(\\2)()', s)
    s = re.sub(r"\bspell *[(]", '_oSpellChecker.isValid(', s)
    if bTokenValue:
        # token values are used as parameter
        s = re.sub(r"[\\](\d+)", 'lToken[\\1+nTokenOffset]["sValue"]', s)
    else:
        # tokens used as parameter
        s = re.sub(r"[\\](\d+)", 'lToken[\\1+nTokenOffset]', s)
    return s


def genTokenLines (sTokenLine, dDef):
    "tokenize a string and return a list of lines of tokens"
    lToken = sTokenLine.split()
    lTokenLines = None
    for i, sToken in enumerate(lToken):
        # optional token?
        bNullPossible = sToken.startswith("?") and sToken.endswith("¿")
        if bNullPossible:
            sToken = sToken[1:-1]
        # token with definition?
        if sToken.startswith("({") and sToken.endswith("})") and sToken[1:-1] in dDef:
            sToken = "(" + dDef[sToken[1:-1]] + ")"
        elif sToken.startswith("{") and sToken.endswith("}") and sToken in dDef:
            sToken = dDef[sToken]
        if ( (sToken.startswith("[") and sToken.endswith("]")) or (sToken.startswith("([") and sToken.endswith("])")) ):
            # multiple token
            bSelectedGroup = sToken.startswith("(") and sToken.endswith(")")
            if bSelectedGroup:
                sToken = sToken[1:-1]
            lNewToken = sToken[1:-1].split("|")
            if not lTokenLines:
                lTokenLines = [ [s]  for s  in lNewToken ]
                if bNullPossible:
                    lTokenLines.extend([ []  for i  in range(len(lNewToken)+1) ])
            else:
                lNewTemp = []
                if bNullPossible:
                    for aRule in lTokenLines:
                        for sElem in lNewToken:
                            aNewRule = list(aRule)
                            aNewRule.append(sElem)
                            lNewTemp.append(aNewRule)
                else:
                    sElem1 = lNewToken.pop(0)
                    for aRule in lTokenLines:
                        for sElem in lNewToken:
                            aNewRule = list(aRule)
                            aNewRule.append("(" + sElem + ")"  if bSelectedGroup  else sElem)
                            lNewTemp.append(aNewRule)
                        aRule.append("(" + sElem1 + ")"  if bSelectedGroup  else sElem1)
                lTokenLines.extend(lNewTemp)
        else:
            # simple token
            if not lTokenLines:
                lTokenLines = [[sToken], []]  if bNullPossible  else [[sToken]]
            else:
                if bNullPossible:
                    lNewTemp = []
                    for aRule in lTokenLines:
                        lNew = list(aRule)
                        lNew.append(sToken)
                        lNewTemp.append(lNew)
                    lTokenLines.extend(lNewTemp)
                else:
                    for aRule in lTokenLines:
                        aRule.append(sToken)
    for aRule in lTokenLines:
        yield aRule


def createRule (iLine, sRuleName, sTokenLine, iActionBlock, sActions, nPriority, dDef):
    # print(iLine, "//", sRuleName, "//", sTokenLine, "//", sActions, "//", nPriority)
    for lToken in genTokenLines(sTokenLine, dDef):
        # Calculate positions
        dPos = {}   # key: iGroup, value: iToken
        iGroup = 0
        for i, sToken in enumerate(lToken):
            if sToken.startswith("(") and sToken.endswith(")"):
                lToken[i] = sToken[1:-1]
                iGroup += 1
                dPos[iGroup] = i + 1    # we add 1, for we count tokens from 1 to n (not from 0)

        # Parse actions
        for iAction, sAction in enumerate(sActions.split(" <<- "), 1):
            sAction = sAction.strip()
            if sAction:
                sActionId = sRuleName + "__b" + str(iActionBlock) + "_a" + str(iAction) + "_" + str(len(lToken))
                aAction = createAction(sActionId, sAction, nPriority, len(lToken), dPos)
                if aAction:
                    dACTIONS[sActionId] = aAction
                    lResult = list(lToken)
                    lResult.extend(["##"+str(iLine), sActionId])
                    yield lResult


def changeReferenceToken (sText, dPos):
    for i in range(len(dPos), 0, -1):
        sText = sText.replace("\\"+str(i), "\\"+str(dPos[i]))
    return sText


def checkTokenNumbers (sText, sActionId, nToken):
    for x in re.finditer(r"\\(\d+)", sText):
        if int(x.group(1)) > nToken:
            print("# Error in token index at line " + sActionId + " ("+str(nToken)+" tokens only)")
            print(sText)


def checkIfThereIsCode (sText, sActionId):
    if re.search("[.]\\w+[(]|sugg\\w+[(]|\\([0-9]|\\[[0-9]", sText):
        print("# Warning at line " + sActionId + ":  This message looks like code. Line should probably begin with =")
        print(sText)


def createAction (sActionId, sAction, nPriority, nToken, dPos):
    # Option
    sOption = False
    m = re.match("/(\\w+)/", sAction) 
    if m:
        sOption = m.group(1)
        sAction = sAction[m.end():].strip()
    # valid action?
    m = re.search("(?P<action>[-~=/])(?P<start>\\d+|)(?P<end>:\\d+|)>> ", sAction)
    if not m:
        print(" # Error. No action found at: ", sActionId)
        print("   ==", sAction, "==")
        return None
    # Condition
    sCondition = sAction[:m.start()].strip()
    if sCondition:
        sCondition = prepareFunction(sCondition)
        sCondition = changeReferenceToken(sCondition, dPos)    
        dFUNCTIONS["g_c_"+sActionId] = sCondition
        sCondition = "g_c_"+sActionId
    else:
        sCondition = ""
    # Action
    cAction = m.group("action")
    sAction = sAction[m.end():].strip()
    sAction = changeReferenceToken(sAction, dPos)
    if not m.group("start"):
        iStartAction = 1
        iEndAction = 0
    else:
        iStartAction = int(m.group("start"))
        iEndAction = int(m.group("end")[1:])  if m.group("end")  else iStartAction
    if dPos:
        try:
            iStartAction = dPos[iStartAction]
            iEndAction = dPos[iEndAction]
        except:
            print("# Error. Wrong groups in: " + sActionId)

    if cAction == "-":
        ## error
        iMsg = sAction.find(" # ")
        if iMsg == -1:
            sMsg = "# Error. Error message not found."
            sURL = ""
            print(sMsg + " Action id: " + sActionId)
        else:
            sMsg = sAction[iMsg+3:].strip()
            sAction = sAction[:iMsg].strip()
            sURL = ""
            mURL = re.search("[|] *(https?://.*)", sMsg)
            if mURL:
                sURL = mURL.group(1).strip()
                sMsg = sMsg[:mURL.start(0)].strip()
            checkTokenNumbers(sMsg, sActionId, nToken)
            if sMsg[0:1] == "=":
                sMsg = prepareFunction(sMsg[1:], True)
                dFUNCTIONS["g_m_"+sActionId] = sMsg
                sMsg = "=g_m_"+sActionId
            else:
                checkIfThereIsCode(sMsg, sActionId)
            
    # checking consistancy
    checkTokenNumbers(sAction, sActionId, nToken)

    if cAction == ">":
        ## no action, break loop if condition is False
        return [sOption, sCondition, cAction, ""]

    if not sAction:
        print("# Error in action at line " + sActionId + ":  This action is empty.")

    if sAction[0:1] != "=":
        checkIfThereIsCode(sAction, sActionId)

    if cAction == "-":
        ## error detected --> suggestion
        if sAction[0:1] == "=":
            sAction = prepareFunction(sAction, True)
            dFUNCTIONS["g_s_"+sActionId] = sAction[1:]
            sAction = "=g_s_"+sActionId
        elif sAction.startswith('"') and sAction.endswith('"'):
            sAction = sAction[1:-1]
        if not sMsg:
            print("# Error in action at line " + sActionId + ":  The message is empty.")
        return [sOption, sCondition, cAction, sAction, iStartAction, iEndAction, nPriority, sMsg, sURL]
    elif cAction == "~":
        ## text processor
        if sAction[0:1] == "=":
            dFUNCTIONS["g_p_"+sActionId] = sAction[1:]
            sAction = "=g_p_"+sActionId
        elif sAction.startswith('"') and sAction.endswith('"'):
            sAction = sAction[1:-1]
        return [sOption, sCondition, cAction, sAction, iStartAction, iEndAction]
    elif cAction == "/":
        ## tags
        return [sOption, sCondition, cAction, sAction, iStartAction, iEndAction]
    elif cAction == "=":
        ## disambiguator
        if sAction[0:1] == "=":
            sAction = sAction[1:]
        if "define" in sAction and not re.search(r"define\(\\\d+ *, *\[.*\] *\)", sAction):
            print("# Error in action at line " + sActionId + ": second argument for <define> must be a list of strings")
        sAction = prepareFunction(sAction)
        dFUNCTIONS["g_d_"+sActionId] = sAction
        sAction = "g_d_"+sActionId
        return [sOption, sCondition, cAction, sAction]
    else:
        print("# Unknown action at line " + sActionId)
        return None


def make (lRule, dDef, sLang, bJavaScript):
    "compile rules, returns a dictionary of values"
    # for clarity purpose, don’t create any file here

    # removing comments, zeroing empty lines, creating definitions, storing tests, merging rule lines
    print("  parsing rules...")
    lTokenLine = []
    sActions = ""
    nPriority = 4
    dAllGraph = {}
    sGraphName = ""
    iActionBlock = 0

    for i, sLine in lRule:
        sLine = sLine.rstrip()
        if "\t" in sLine:
            # tabulation not allowed
            print("Error. Tabulation at line: ", i)
            exit()
        elif sLine.startswith("@@@@GRAPH: "):
            # rules graph call
            m = re.match(r"@@@@GRAPH: *(\w+)", sLine.strip())
            if m:
                sGraphName = m.group(1)
                if sGraphName in dAllGraph:
                    print("Error. Group name " + sGraphName + " already exists.")
                    exit()
                dAllGraph[sGraphName] = []
            else:
                print("Error. Graph name not found at line", i)
                exit()
        elif sLine.startswith("__") and sLine.endswith("__"):
            # new rule group
            m = re.match("__(\\w+)(!\\d|)__", sLine)
            if m:
                sRuleName = m.group(1)
                iActionBlock = 1
                nPriority = int(m.group(2)[1:]) if m.group(2)  else 4
            else:
                print("Error at rule group: ", sLine, " -- line:", i)
                break
        elif re.search("    +<<- ", sLine):
            # actions
            sActions += " " + sLine.strip()
        elif re.match("[  ]*$", sLine):
            # empty line to end merging
            if not lTokenLine:
                continue
            if not sActions:
                print("Error. No action found at line:", i)
                exit()
            if not sGraphName:
                print("Error. All rules must belong to a named graph. Line: ", i)
                exit()
            for j, sTokenLine in lTokenLine:
                dAllGraph[sGraphName].append((j, sRuleName, sTokenLine, iActionBlock, sActions, nPriority))
            lTokenLine.clear()
            sActions = ""
            iActionBlock += 1
        elif sLine.startswith(("    ")):
            # tokens
            lTokenLine.append([i, sLine.strip()])
        else:
            print("Unknown line:")
            print(sLine)

    # processing rules
    print("  preparing rules...")
    for sGraphName, lRuleLine in dAllGraph.items():
        lPreparedRule = []
        for i, sRuleGroup, sTokenLine, iActionBlock, sActions, nPriority in lRuleLine:
            for lRule in createRule(i, sRuleGroup, sTokenLine, iActionBlock, sActions, nPriority, dDef):
                lPreparedRule.append(lRule)
        # Graph creation
        oDARG = darg.DARG(lPreparedRule, sLang)
        dAllGraph[sGraphName] = oDARG.createGraph()
        # Debugging
        #print("\nGRAPH:", sGraphName)
        #for e in lPreparedRule:
        #    print(e)
        #for k, v in dAllGraph[sGraphName].items():
        #    print(k, "\t", v)

    # creating file with all functions callable by rules
    print("  creating callables...")
    sPyCallables = "# generated code, do not edit\n"
    #sJSCallables = "// generated code, do not edit\nconst oEvalFunc = {\n"
    for sFuncName, sReturn in dFUNCTIONS.items():
        if sFuncName.startswith("g_c_"): # condition
            sParams = "lToken, nTokenOffset, nLastToken, sCountry, bCondMemo, dTags"
        elif sFuncName.startswith("g_m_"): # message
            sParams = "lToken, nTokenOffset"
        elif sFuncName.startswith("g_s_"): # suggestion
            sParams = "lToken, nTokenOffset"
        elif sFuncName.startswith("g_p_"): # preprocessor
            sParams = "lToken"
        elif sFuncName.startswith("g_d_"): # disambiguator
            sParams = "lToken, nTokenOffset"
        else:
            print("# Unknown function type in [" + sFuncName + "]")
            continue
        sPyCallables += "def {} ({}):\n".format(sFuncName, sParams)
        sPyCallables += "    return " + sReturn + "\n"
        #sJSCallables += "    {}: function ({})".format(sFuncName, sParams) + " {\n"
        #sJSCallables += "        return " + jsconv.py2js(sReturn) + ";\n"
        #sJSCallables += "    },\n"
    #sJSCallables += "}\n"

    # Debugging
    print("\nActions:")
    for sActionName, aAction in dACTIONS.items():
        print(sActionName, aAction)
    print("\nFunctions:")
    print(sPyCallables)

    # Result
    d = {
        "graph_callables": sPyCallables,
        "rules_graphs": dAllGraph,
        "rules_actions": dACTIONS
    }
    return d