Grammalecte  Diff

"""
Grammalecte: compile rules
"""

import re
import traceback
import json

import compile_rules_js_convert as jsconv
import compile_rules_graph as crg


dDEF = {}
lFUNCTIONS = []

aRULESET = set()     # set of rule-ids to check if there is several rules with the same id
nRULEWITHOUTNAME = 0

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)
    s = re.sub(r"isRealEnd *\(\)", 'after("^ *$")', s)
    s = re.sub(r"isEnd0 *\(\)", 'after0("^ *$|^,")', s)
    s = re.sub(r"isRealEnd0 *\(\)", 'after0("^ *$")', s)
    s = re.sub(r"(select|exclude)[(][\\](\d+)", '\\1(dTokenPos, m.start(\\2), m.group(\\2)', s)
    s = re.sub(r"define[(][\\](\d+)", 'define(dTokenPos, m.start(\\1)', s)
    s = re.sub(r"(morph|morphex|displayInfo)[(][\\](\d+)", '\\1((m.start(\\2), m.group(\\2))', s)
    s = re.sub(r"(morph|morphex|displayInfo)[(]", '\\1(dTokenPos, ', s)
    s = re.sub(r"(sugg\w+|switch\w+)\(@", '\\1(m.group(i[4])', s)
    s = re.sub(r"word\(\s*1\b", 'nextword1(s, m.end()', s)                                  # word(1)
    s = re.sub(r"word\(\s*-1\b", 'prevword1(s, m.start()', s)                               # word(-1)
    s = re.sub(r"word\(\s*(\d)", 'nextword(s, m.end(), \\1', s)                             # word(n)
    s = re.sub(r"word\(\s*-(\d)", 'prevword(s, m.start(), \\1', s)                          # word(-n)
    s = re.sub(r"before\(\s*", 'look(s[:m.start()], ', s)                                   # before(s)
    s = re.sub(r"after\(\s*", 'look(s[m.end():], ', s)                                      # after(s)
    s = re.sub(r"textarea\(\s*", 'look(s, ', s)                                             # textarea(s)
    s = re.sub(r"before_chk1\(\s*", 'look_chk1(dTokenPos, s[:m.start()], 0, ', s)           # before_chk1(s)
    s = re.sub(r"after_chk1\(\s*", 'look_chk1(dTokenPos, s[m.end():], m.end(), ', s)        # after_chk1(s)
    s = re.sub(r"textarea_chk1\(\s*", 'look_chk1(dTokenPos, s, 0, ', s)                     # textarea_chk1(s)
    s = re.sub(r"/0", 'sx[m.start():m.end()]', s)                                           # /0
    s = re.sub(r"before0\(\s*", 'look(sx[:m.start()], ', s)                                 # before0(s)
    s = re.sub(r"after0\(\s*", 'look(sx[m.end():], ', s)                                    # after0(s)
    s = re.sub(r"textarea0\(\s*", 'look(sx, ', s)                                           # textarea0(s)
    s = re.sub(r"before0_chk1\(\s*", 'look_chk1(dTokenPos, sx[:m.start()], 0, ', s)         # before0_chk1(s)
    s = re.sub(r"after0_chk1\(\s*", 'look_chk1(dTokenPos, sx[m.end():], m.end(), ', s)      # after0_chk1(s)
    s = re.sub(r"textarea0_chk1\(\s*", 'look_chk1(dTokenPos, sx, 0, ', s)                   # textarea0_chk1(s)
    s = re.sub(r"isEndOfNG\(\s*\)", 'isEndOfNG(dTokenPos, s[m.end():], m.end())', s)        # isEndOfNG(s)
    s = re.sub(r"isNextNotCOD\(\s*\)", 'isNextNotCOD(dTokenPos, s[m.end():], m.end())', s)  # isNextNotCOD(s)
    s = re.sub(r"isNextVerb\(\s*\)", 'isNextVerb(dTokenPos, s[m.end():], m.end())', s)      # isNextVerb(s)
    s = re.sub(r"\bspell *[(]", '_oSpellChecker.isValid(', s)
    s = re.sub(r"[\\](\d+)", 'm.group(\\1)', s)
    return s


def uppercase (s, sLang):
    "(flag i is not enough): converts regex to uppercase regex: 'foo' becomes '[Ff][Oo][Oo]', but 'Bar' becomes 'B[Aa][Rr]'."

            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)
    cCaseMode = 'i'         # i: case insensitive,  s: case sensitive,  u: uppercasing allowed
    cWordLimitLeft = '['    # [: word limit, <: no specific limit
    cWordLimitRight = ']'   # ]: word limit, >: no specific limit
    m = re.match("^__(?P<borders_and_case>[[<]\\w[]>])(?P<option>/[a-zA-Z0-9]+|)(?P<ruleid>\\(\\w+\\)|)(?P<priority>![0-9]|)__ *", s)

    #### REGEX TRIGGER
    i = s.find(" <<-")
    if i == -1:
        print("# Error: no condition at line " + sLineId)
        return None
    sRegex = s[:i].strip()
    s = s[i+4:]

    # JS groups positioning codes
    m = re.search("@@\\S+", sRegex)
    if m:
        tGroups = jsconv.groupsPositioningCodeToList(sRegex[m.start()+2:])
        sRegex = sRegex[:m.start()].strip()
    # JS regex
    m = re.search("<js>.+</js>i?", sRegex)

        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 checkReferenceNumbers (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 (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))
        checkReferenceNumbers(sCondition, sIdAction, nGroup)


        if ".match" in sCondition:
            print("# Error. JS compatibility. Don't use .match() in condition, use .search()")
        sCondition = "_c_"+sIdAction
    else:
        sCondition = None

    #### iGroup / positioning
    iGroup = int(m.group(2)) if m.group(2) else 0
    if iGroup > nGroup:
        print("# Selected group > group number in regex at line " + sIdAction)

    #### ACTION
    sAction = sAction[m.end():].strip()
    cAction = m.group(1)
    if cAction == "-":
        ## error
        iMsg = sAction.find(" # ")
        if iMsg == -1:
            sMsg = "# Error. Error message not found."
            sURL = ""
            print(sMsg + " Action id: " + sIdAction)
        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()
            checkReferenceNumbers(sMsg, sIdAction, nGroup)
            if sMsg[0:1] == "=":
                sMsg = prepareFunction(sMsg[1:])
                lFUNCTIONS.append(("_m_"+sIdAction, sMsg))



                sMsg = "=_m_"+sIdAction
            else:


                checkIfThereIsCode(sMsg, sIdAction)



    checkReferenceNumbers(sAction, sIdAction, nGroup)
    if sAction[0:1] == "=" or cAction == "=":


        sAction = prepareFunction(sAction)
        sAction = sAction.replace("m.group(i[4])", "m.group("+str(iGroup)+")")



    else:
        checkIfThereIsCode(sAction, sIdAction)

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

    if not sAction:
        print("# Error in action at line " + sIdAction + ":  This action is empty.")
        return None

    if cAction == "-":
        ## error detected --> suggestion


        if sAction[0:1] == "=":
            lFUNCTIONS.append(("_s_"+sIdAction, sAction[1:]))
            sAction = "=_s_"+sIdAction
        elif sAction.startswith('"') and sAction.endswith('"'):
            sAction = sAction[1:-1]
        if not sMsg:
            print("# Error in action at line " + sIdAction + ":  the message is empty.")
        return [sCondition, cAction, sAction, iGroup, sMsg, sURL]
    elif cAction == "~":
        ## text processor


        if sAction[0:1] == "=":
            lFUNCTIONS.append(("_p_"+sIdAction, sAction[1:]))
            sAction = "=_p_"+sIdAction
        elif sAction.startswith('"') and sAction.endswith('"'):
            sAction = sAction[1:-1]
        return [sCondition, cAction, sAction, iGroup]
    elif cAction == "=":
        ## disambiguator
        if sAction[0:1] == "=":
            sAction = sAction[1:]
        if "define" in sAction and not re.search(r"define\(dTokenPos, *m\.start.*, \[.*\] *\)", sAction):
            print("# Error in action at line " + sIdAction + ": second argument for define must be a list of strings")
            print(sAction)
        lFUNCTIONS.append(("_d_"+sIdAction, sAction))
        sAction = "_d_"+sIdAction
        return [sCondition, cAction, sAction]



    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

    print("> read rules file...")
    try:
        lRules = open(spLang + "/rules.grx", 'r', encoding="utf-8").readlines()
    except:
        print("Error. Rules file in project [" + sLang + "] not found.")
        exit()

    # removing comments, zeroing empty lines, creating definitions, storing tests, merging rule lines
    print("  parsing rules...")


    lRuleLine = []
    lTest = []
    lOpt = []
    bGraph = False
    lGraphRule = []

    for i, sLine in enumerate(lRules, 1):
        if sLine.startswith('#END'):
            # arbitrary end
            printBookmark(0, "BREAK BY #END", i)
            break
        elif sLine.startswith("#"):
            # comment
            pass
        elif sLine.startswith("DEF:"):

            # definition



            m = re.match("DEF: +([a-zA-Z_][a-zA-Z_0-9]*) +(.+)$", sLine.strip())
            if m:
                dDEF["{"+m.group(1)+"}"] = m.group(2)
            else:
                print("Error in definition: ", end="")
                print(sLine.strip())
        elif sLine.startswith("TEST:"):
            # test
            lTest.append("{:<8}".format(i) + "  " + sLine[5:].strip())
        elif sLine.startswith("TODO:"):
            # todo
            pass
        elif sLine.startswith(("OPTGROUP/", "OPTSOFTWARE:", "OPT/", "OPTLANG/", "OPTDEFAULTUILANG:", "OPTLABEL/", "OPTPRIORITY/")):
            # options
            lOpt.append(sLine)


        elif sLine.startswith("!!"):
            # bookmark
            m = re.match("!!+", sLine)
            nExMk = len(m.group(0))
            if sLine[nExMk:].strip():
                printBookmark(nExMk-2, sLine[nExMk:-3].strip(), i)
        # Graph rules
        elif sLine.startswith("@@@@GRAPH:"):
            # rules graph call
            m = re.match(r"@@@@GRAPH: *(\w+)", sLine.strip())
            if m:
                printBookmark(1, "____ GRAPH: " + m.group(1) + " ____", i)
                lRuleLine.append([i, "@@@@"+m.group(1)])
                bGraph = True
            lGraphRule.append([i, sLine])
            bGraph = True
        elif sLine.startswith("@@@@END_GRAPH"):
            #lGraphRule.append([i, sLine])
            bGraph = False
        elif re.match("@@@@ *$", sLine):
            pass
        elif bGraph:
            lGraphRule.append([i, sLine])
        # Regex rules
        elif re.match("[  \t]*$", sLine):
            # empty line
            pass
        elif sLine.startswith(("    ", "\t")):
            # rule (continuation)
            lRuleLine[-1][1] += " " + sLine.strip()
        else:
            # new rule
            lRuleLine.append([i, sLine.strip()])

    # generating options files
    print("  parsing options...")
    try:
        dOptions, dOptPriority = prepareOptions(lOpt)
    except:

                        lSentenceRulesJS.append(jsconv.pyRuleToJS(aRule, dJSREGEXES, sWORDLIMITLEFT))

    # 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 lFUNCTIONS:

        if sFuncName.startswith("_c_"): # condition
            sParams = "s, sx, m, dTokenPos, sCountry, bCondMemo"
        elif sFuncName.startswith("_m_"): # message
            sParams = "s, m"
        elif sFuncName.startswith("_s_"): # suggestion
            sParams = "s, m"
        elif sFuncName.startswith("_p_"): # preprocessor
            sParams = "s, m"
        elif sFuncName.startswith("_d_"): # disambiguator
            sParams = "s, m, dTokenPos"
        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"

    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, dOptPriority, bJavaScript)
    dVars.update(dVars2)

    return dVars