#!/usr/bin/python

import sys
import re
import io

definitionSet = set()
definitionToSourceLocationMap = dict()
touchSet = set()
# exclude some stuff, mostly because they are some kind of definition of external file formats
excludedSourceFiles = set([
    "include/svx/msdffdef.hxx",
    "sw/source/filter/ww8/fields.hxx",
    "sw/source/filter/inc/wwstyles.hxx",
    "sw/inc/toxe.hxx",
    "sw/inc/poolfmt.hxx",
    "sw/inc/hintids.hxx",
    "vcl/inc/unx/XIM.h",
    ])
excludedTypes = set([
    "SwVarFormat", "RES_FIELDS", "SwFillOrder", "SwIoDetect", "SwDocumentSettingsPropertyHandles",
    "SalGenericDataType", "SwDateSubFormat", "XclFutureRecType", "ds_status", "MediaCommand",
    "EmfPlusHatchStyle"
    ])

# clang does not always use exactly the same numbers in the type-parameter vars it generates
# so I need to substitute them to ensure we can match correctly.
normalizeTypeParamsRegex = re.compile(r"type-parameter-\d+-\d+")
def normalizeTypeParams( line ):
    return normalizeTypeParamsRegex.sub("type-parameter-?-?", line)

# The parsing here is designed to avoid grabbing stuff which is mixed in from gbuild.
# I have not yet found a way of suppressing the gbuild output.
with io.open("loplugin.unusedenumvalues.log", "rb", buffering=1024*1024) as txt:
    for line in txt:
        tokens = line.strip().split("\t")
        if tokens[0] == "definition:":
            funcInfo = (normalizeTypeParams(tokens[1]), tokens[2])
            definitionSet.add(funcInfo)
            definitionToSourceLocationMap[funcInfo] = tokens[3]
        elif tokens[0] == "touch:":
            callInfo = (normalizeTypeParams(tokens[1]), tokens[2])
            touchSet.add(callInfo)

# Invert the definitionToSourceLocationMap
# If we see more than one method at the same sourceLocation, it's being autogenerated as part of a template
# and we should just ignore
sourceLocationToDefinitionMap = {}
for k, v in definitionToSourceLocationMap.iteritems():
    sourceLocationToDefinitionMap[v] = sourceLocationToDefinitionMap.get(v, [])
    sourceLocationToDefinitionMap[v].append(k)
for k, definitions in sourceLocationToDefinitionMap.iteritems():
    if len(definitions) > 1:
        for d in definitions:
            definitionSet.remove(d)
    
untouchedSet = set()
for d in definitionSet:
    clazz = d[0] + " " + d[1]
    if d in touchSet:
        continue
    srcLoc = definitionToSourceLocationMap[d]
    srcLocWithoutLineNo = srcLoc.split(":")[0]
    # ignore external source code
    if (srcLoc.startswith("external/")):
        continue
    # ignore build folder
    if (srcLoc.startswith("workdir/")):
        continue
    # ignore our stable/URE/UNO api
    if (srcLoc.startswith("include/com/")
        or srcLoc.startswith("include/cppu/")
        or srcLoc.startswith("include/cppuhelper/")
        or srcLoc.startswith("include/osl/")
        or srcLoc.startswith("include/rtl/")
        or srcLoc.startswith("include/sal/")
        or srcLoc.startswith("include/salhelper/")
        or srcLoc.startswith("include/systools/")
        or srcLoc.startswith("include/typelib/")
        or srcLoc.startswith("include/uno/")):
        continue
    if srcLocWithoutLineNo in excludedSourceFiles or d[0] in excludedTypes:
        continue
    # structure definitions
    if srcLoc.startswith("lotuswordpro/"):
        continue
        
    # used in templates to find the last member of an enum
    if d[1] == "LAST" or d[1].endswith("_END"):
        continue
    # used to aid in alignment of enum values
    if d[1].endswith("FORCE_EQUAL_SIZE"):
        continue
        
    untouchedSet.add((clazz, srcLoc))

# sort the results using a "natural order" so sequences like [item1,item2,item10] sort nicely
def natural_sort_key(s, _nsre=re.compile('([0-9]+)')):
    return [int(text) if text.isdigit() else text.lower()
            for text in re.split(_nsre, s)]

# sort results by name and line number
tmp1list = sorted(untouchedSet, key=lambda v: natural_sort_key(v[1]))

# print out the results
with open("loplugin.unusedenumvalues.report-untouched", "wt") as f:
    for t in tmp1list:
        f.write( t[1] + "\n" )
        f.write( "    " + t[0] + "\n" )