Grammalecte  Check-in [18191569f4]

"""
Grammalecte: compile rules
"""

import re
import traceback
import json

import compile_rules_js_convert as jsconv
import compile_rules_graph as crg

dJSREGEXES = {}

sWORDLIMITLEFT  = r"(?<![\w.,–-])"   # r"(?<![-.,—])\b"  seems slower
sWORDLIMITRIGHT = r"(?![\w–-])"      # r"\b(?!-—)"       seems slower


def prepareFunction (s):
    "convert simple rule syntax to a string of Python code"
    s = s.replace("__also__", "bCondMemo")
    s = s.replace("__else__", "not bCondMemo")
    s = re.sub(r"isStart *\(\)", 'before("^ *$|, *$")', s)
    s = re.sub(r"isRealStart *\(\)", 'before("^ *$")', s)
    s = re.sub(r"isStart0 *\(\)", 'before0("^ *$|, *$")', s)
    s = re.sub(r"isRealStart0 *\(\)", 'before0("^ *$")', s)
    s = re.sub(r"isEnd *\(\)", 'after("^ *$|^,")', s)

            nState = 4
        elif nState == 4:
            nState = 0
    return sUp


def countGroupInRegex (sRegex):
    "returns the number of groups in <sRegex>"
    try:
        return re.compile(sRegex).groups
    except:
        traceback.print_exc()
        print(sRegex)
    return 0


def createRule (s, nIdLine, sLang, bParagraph, dOptPriority):
    "returns rule as list [option name, regex, bCaseInsensitive, identifier, list of actions]"
    global dJSREGEXES
    global nRULEWITHOUTNAME

    sLineId = str(nIdLine) + ("p" if bParagraph else "s")
    sRuleId = sLineId

    #### GRAPH CALL
    if s.startswith("@@@@"):
        if bParagraph:
            print("Error. Graph call can be made only after the first pass (sentence by sentence)")
            exit()
        return ["@@@@", s[4:], sLineId]

    #### OPTIONS
    sOption = False         # False or [a-z0-9]+ name
    nPriority = 4           # Default is 4, value must be between 0 and 9
    tGroups = None          # code for groups positioning (only useful for JavaScript)

        sRegex = sRegex.replace("(?i)", "")
        sRegex = uppercase(sRegex, sLang)
    else:
        print("# Unknown case mode [" + cCaseMode + "] at line " + sLineId)

    ## check regex
    try:
        re.compile(sRegex)
    except:
        print("# Regex error at line ", nIdLine)
        print(sRegex)
        traceback.print_exc()
        return None
    ## groups in non grouping parenthesis
    for x in re.finditer(r"\(\?:[^)]*\([[\w -]", sRegex):
        print("# Warning: groups inside non grouping parenthesis in regex at line " + sLineId)

    #### PARSE ACTIONS
    lActions = []
    nAction = 1
    for sAction in s.split(" <<- "):
        t = createAction(sRuleId + "_" + str(nAction), sAction, nGroup)
        nAction += 1
        if t:
            lActions.append(t)
    if not lActions:
        return None

    return [sOption, sRegex, bCaseInsensitive, sLineId, sRuleId, nPriority, lActions, tGroups]


def createAction (sIdAction, sAction, nGroup):
    "returns an action to perform as a tuple (condition, action type, action[, iGroup [, message, URL ]])"


    m = re.search(r"([-~=>])(\d*|)>>", sAction)
    if not m:
        print("# No action at line " + sIdAction)
        return None

    #### CONDITION
    sCondition = sAction[:m.start()].strip()
    if sCondition:
        sCondition = prepareFunction(sCondition)
        lFUNCTIONS.append(("_c_"+sIdAction, sCondition))
        for x in re.finditer(r"[.](?:group|start|end)[(](\d+)[)]", sCondition):
            if int(x.group(1)) > nGroup:
                print("# Error in groups in condition at line " + sIdAction + " ("+str(nGroup)+" groups only)")
        if ".match" in sCondition:
            print("# Error. JS compatibility. Don't use .match() in condition, use .search()")
        sCondition = "_c_"+sIdAction
    else:
        sCondition = None

            mURL = re.search("[|] *(https?://.*)", sMsg)
            if mURL:
                sURL = mURL.group(1).strip()
                sMsg = sMsg[:mURL.start(0)].strip()
            if sMsg[0:1] == "=":
                sMsg = prepareFunction(sMsg[1:])
                lFUNCTIONS.append(("_m_"+sIdAction, sMsg))
                for x in re.finditer(r"group[(](\d+)[)]", sMsg):
                    if int(x.group(1)) > nGroup:
                        print("# Error in groups in message at line " + sIdAction + " ("+str(nGroup)+" groups only)")
                sMsg = "=_m_"+sIdAction
            else:
                for x in re.finditer(r"\\(\d+)", sMsg):
                    if int(x.group(1)) > nGroup:
                        print("# Error in groups in message at line " + sIdAction + " ("+str(nGroup)+" groups only)")
                if re.search("[.]\\w+[(]", sMsg):
                    print("# Error in message at line " + sIdAction + ":  This message looks like code. Line should begin with =")

    if sAction[0:1] == "=" or cAction == "=":
        if "define" in sAction and not re.search(r"define\(\\\d+ *, *\[.*\] *\)", sAction):
            print("# Error in action at line " + sIdAction + ": second argument for define must be a list of strings")
        sAction = prepareFunction(sAction)
        sAction = sAction.replace("m.group(i[4])", "m.group("+str(iGroup)+")")
        for x in re.finditer(r"group[(](\d+)[)]", sAction):
            if int(x.group(1)) > nGroup:
                print("# Error in groups in replacement at line " + sIdAction + " ("+str(nGroup)+" groups only)")
    else:
        for x in re.finditer(r"\\(\d+)", sAction):
            if int(x.group(1)) > nGroup:
                print("# Error in groups in replacement at line " + sIdAction + " ("+str(nGroup)+" groups only)")
        if re.search("[.]\\w+[(]|sugg\\w+[(]", sAction):

        return [sCondition, cAction, ""]
    else:
        print("# Unknown action at line " + sIdAction)
        return None


def _calcRulesStats (lRules):
    "count rules and actions"
    d = {'=':0, '~': 0, '-': 0, '>': 0}
    for aRule in lRules:
        if aRule[0] != "@@@@":
            for aAction in aRule[6]:
                d[aAction[1]] = d[aAction[1]] + 1
    return (d, len(lRules))


def displayStats (lParagraphRules, lSentenceRules):
    "display rules numbers"
    print("  {:>18} {:>18} {:>18} {:>18}".format("DISAMBIGUATOR", "TEXT PROCESSOR", "GRAMMAR CHECKING", "REGEX"))
    d, nRule = _calcRulesStats(lParagraphRules)
    print("§ {:>10} actions {:>10} actions {:>10} actions  in {:>8} rules".format(d['='], d['~'], d['-'], nRule))
    d, nRule = _calcRulesStats(lSentenceRules)
    print("s {:>10} actions {:>10} actions {:>10} actions  in {:>8} rules".format(d['='], d['~'], d['-'], nRule))

            m = re.match("OPTGROUP/([a-z0-9]+):(.+)$", sLine)
            lStructOpt.append( (m.group(1), list(map(str.split, m.group(2).split(",")))) )
        elif sLine.startswith("OPTSOFTWARE:"):
            lOpt = [ [s, {}]  for s in sLine[12:].strip().split() ]  # don’t use tuples (s, {}), because unknown to JS
        elif sLine.startswith("OPT/"):
            m = re.match("OPT/([a-z0-9]+):(.+)$", sLine)
            for i, sOpt in enumerate(m.group(2).split()):
                lOpt[i][1][m.group(1)] = eval(sOpt)
        elif sLine.startswith("OPTPRIORITY/"):
            m = re.match("OPTPRIORITY/([a-z0-9]+): *([0-9])$", sLine)
            dOptPriority[m.group(1)] = int(m.group(2))
        elif sLine.startswith("OPTLANG/"):
            m = re.match("OPTLANG/([a-z][a-z](?:_[A-Z][A-Z]|)):(.+)$", sLine)
            sLang = m.group(1)[:2]
            dOptLabel[sLang] = { "__optiontitle__": m.group(2).strip() }

    print("  options defined for: " + ", ".join([ t[0] for t in lOpt ]))
    dOptions = { "lStructOpt": lStructOpt, "dOptLabel": dOptLabel, "sDefaultUILang": sDefaultUILang }
    dOptions.update({ "dOpt"+k: v  for k, v in lOpt })
    return dOptions, dOptPriority


def printBookmark (nLevel, sComment, nLine):
    "print bookmark within the rules file"
    print("  {:>6}:  {}".format(nLine, "  " * nLevel + sComment))


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

        sJSCallables += "    },\n"
    sJSCallables += "}\n"

    displayStats(lParagraphRules, lSentenceRules)

    print("Unnamed rules: " + str(nRULEWITHOUTNAME))

    dVars = {   "callables": sPyCallables,
                "callablesJS": sJSCallables,
                "gctests": sGCTests,
                "gctestsJS": sGCTestsJS,
                "paragraph_rules": mergeRulesByOption(lParagraphRules),
                "sentence_rules": mergeRulesByOption(lSentenceRules),
                "paragraph_rules_JS": jsconv.writeRulesToJSArray(mergeRulesByOption(lParagraphRulesJS)),
                "sentence_rules_JS": jsconv.writeRulesToJSArray(mergeRulesByOption(lSentenceRulesJS)) }
    dVars.update(dOptions)

    # compile graph rules
    dVars2 = crg.make(lGraphRule, dDEF, sLang, bJavaScript)
    dVars.update(dVars2)

    return dVars

"""
Grammalecte: compile rules
Create a Direct Acyclic Rule Graphs (DARGs)
"""

import re
import traceback
import json

import darg


dACTIONS = {}
dFUNCTIONS = {}


def prepareFunction (s, bTokenValue=False):
    "convert simple rule syntax to a string of Python code"
    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], dTags', s)
    s = re.sub(r"(tag_before|tag_after)[(][\\](\d+)", 'g_\\1(lToken[\\2+nTokenOffset], dTags', 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

    return s


def genTokenLines (sTokenLine, dDef):
    "tokenize a string and return a list of lines of tokens"
    lToken = sTokenLine.split()
    lTokenLines = None
    for sToken in 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]] + ")"

                    for aRule in lTokenLines:
                        aRule.append(sToken)
    for aRule in lTokenLines:
        yield aRule


def createRule (iLine, sRuleName, sTokenLine, iActionBlock, sActions, nPriority, dDef):
    "generator: create rule as list"
    # 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(")"):

                    dACTIONS[sActionId] = aAction
                    lResult = list(lToken)
                    lResult.extend(["##"+str(iLine), sActionId])
                    yield lResult


def changeReferenceToken (sText, dPos):
    "change group reference in <sText> with values in <dPos>"
    for i in range(len(dPos), 0, -1):
        sText = sText.replace("\\"+str(i), "\\"+str(dPos[i]))
    return sText


def checkTokenNumbers (sText, sActionId, nToken):
    "check if token references in <sText> greater than <nToken> (debugging)"
    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):
    "check if there is code in <sText> (debugging)"
    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):
    "create action rule as a list"
    # Option
    sOption = False
    m = re.match("/(\\w+)/", sAction)
    if m:
        sOption = m.group(1)
        sAction = sAction[m.end():].strip()
    # valid action?

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

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

"""
Convert Python code and regexes to JavaScript code
"""

import copy
import re
import json


def py2js (sCode):

        sRegex = sRegex + "i"
    if not lNegLookBeforeRegex:
        lNegLookBeforeRegex = None
    return (sRegex, lNegLookBeforeRegex)


def pyRuleToJS (lRule, dJSREGEXES, sWORDLIMITLEFT):
    "modify Python rules -> JS rules"
    lRuleJS = copy.deepcopy(lRule)
    # graph rules
    if lRuleJS[0] == "@@@@":
        return lRuleJS
    del lRule[-1] # tGroups positioning codes are useless for Python
    # error messages
    for aAction in lRuleJS[6]:
        if aAction[1] == "-":
            aAction[2] = aAction[2].replace(" ", " ") # nbsp --> nnbsp
            aAction[4] = aAction[4].replace("« ", "« ").replace(" »", " »").replace(" :", " :").replace(" :", " :")
    # js regexes
    lRuleJS[1], lNegLookBehindRegex = regex2js(dJSREGEXES.get(lRuleJS[3], lRuleJS[1]), sWORDLIMITLEFT)
    lRuleJS.append(lNegLookBehindRegex)
    return lRuleJS


def writeRulesToJSArray (lRules):
    "create rules as a string of arrays (to be bundled in a JSON string)"
    sArray = "[\n"
    for sOption, aRuleGroup in lRules:
        if sOption != "@@@@":
            sArray += '  ["' + sOption + '", [\n'  if sOption  else  "  [false, [\n"
            for sRegex, bCaseInsensitive, sLineId, sRuleId, nPriority, lActions, aGroups, aNegLookBehindRegex in aRuleGroup:
                sArray += '    [' + sRegex + ", "
                sArray += "true, " if bCaseInsensitive  else "false, "

                sArray += '    ["' + sGraphName + '", "' + sLineId + '"],\n"'
            sArray += "  ]],\n"
    sArray += "]"
    return sArray


def groupsPositioningCodeToList (sGroupsPositioningCode):
    "convert <sGroupsPositioningCode> to a list of codes (numbers or strings)"
    if not sGroupsPositioningCode:
        return None
    return [ int(sCode)  if sCode.isdigit() or (sCode[0:1] == "-" and sCode[1:].isdigit())  else sCode \
             for sCode in sGroupsPositioningCode.split(",") ]

def getConfig (sLang):
    "load config.ini in <sLang> at gc_lang/<sLang>, returns xConfigParser object"
    xConfig = configparser.SafeConfigParser()
    xConfig.optionxform = str
    try:
        xConfig.read_file(open("gc_lang/" + sLang + "/config.ini", "r", encoding="utf-8"))
    except FileNotFoundError:
        print("# Error. Can’t read config file [" + sLang + "]")
        exit()
    return xConfig


def createOptionsLabelProperties (dOptLbl):