" Description: Omni completion tokenizer
" Maintainer:  Vissale NEANG
" Last Change: 26 sept. 2007
" TODO: Generic behaviour for Tokenize()

" From the C++ BNF
let s:cppKeyword = ['asm', 'auto', 'bool', 'break', 'case', 'catch', 'char', 'class', 'const', 'const_cast', 'continue', 'default', 'delete', 'do', 'double', 'dynamic_cast', 'else', 'enum', 'explicit', 'export', 'extern', 'false', 'float', 'for', 'friend', 'goto', 'if', 'inline', 'int', 'long', 'mutable', 'namespace', 'new', 'operator', 'private', 'protected', 'public', 'register', 'reinterpret_cast', 'return', 'short', 'signed', 'sizeof', 'static', 'static_cast', 'struct', 'switch', 'template', 'this', 'throw', 'true', 'try', 'typedef', 'typeid', 'typename', 'union', 'unsigned', 'using', 'virtual', 'void', 'volatile', 'wchar_t', 'while', 'and', 'and_eq', 'bitand', 'bitor', 'compl', 'not', 'not_eq', 'or', 'or_eq', 'xor', 'xor_eq']

let s:reCppKeyword = '\C\<'.join(s:cppKeyword, '\>\|\<').'\>'

" The order of items in this list is very important because we use this list to build a regular
" expression (see below) for tokenization
let s:cppOperatorPunctuator = ['->*', '->', '--', '-=', '-', '!=', '!', '##', '#', '%:%:', '%=', '%>', '%:', '%', '&&', '&=', '&', '(', ')', '*=', '*', ',', '...', '.*', '.', '/=', '/', '::', ':>', ':', ';', '?', '[', ']', '^=', '^', '{', '||', '|=', '|', '}', '~', '++', '+=', '+', '<<=', '<%', '<:', '<<', '<=', '<', '==', '=', '>>=', '>>', '>=', '>']

" We build the regexp for the tokenizer
let s:reCComment = '\/\*\|\*\/'
let s:reCppComment = '\/\/'
let s:reComment = s:reCComment.'\|'.s:reCppComment
let s:reCppOperatorOrPunctuator = escape(join(s:cppOperatorPunctuator, '\|'), '*./^~[]')


" Tokenize a c++ code
" a token is dictionary where keys are:
"   -   kind = cppKeyword|cppWord|cppOperatorPunctuator|unknown|cComment|cppComment|cppDigit
"   -   value = 'something'
"   Note: a cppWord is any word that is not a cpp keyword
function! omni#cpp#tokenizer#Tokenize(szCode)
    let result = []

    " The regexp to find a token, a token is a keyword, word or
    " c++ operator or punctuator. To work properly we have to put 
    " spaces and tabs to our regexp.
    let reTokenSearch = '\(\w\+\)\|\s\+\|'.s:reComment.'\|'.s:reCppOperatorOrPunctuator
    " eg: 'using namespace std;'
    "      ^    ^
    "  start=0 end=5
    let startPos = 0
    let endPos = matchend(a:szCode, reTokenSearch)
    let len = endPos-startPos
    while endPos!=-1
        " eg: 'using namespace std;'
        "      ^    ^
        "  start=0 end=5
        "  token = 'using'
        " We also remove space and tabs
        let token = substitute(strpart(a:szCode, startPos, len), '\s', '', 'g')

        " eg: 'using namespace std;'
        "           ^         ^
        "       start=5     end=15
        let startPos = endPos
        let endPos = matchend(a:szCode, reTokenSearch, startPos)
        let len = endPos-startPos

        " It the token is empty we continue
        if token==''
            continue
        endif

        " Building the token
        let resultToken = {'kind' : 'unknown', 'value' : token}

        " Classify the token
        if token =~ '^\d\+'
            " It's a digit
            let resultToken.kind = 'cppDigit'
        elseif token=~'^\w\+$'
            " It's a word
            let resultToken.kind = 'cppWord'

            " But maybe it's a c++ keyword
            if match(token, s:reCppKeyword)>=0
                let resultToken.kind = 'cppKeyword'
            endif
        else
            if match(token, s:reComment)>=0
                if index(['/*','*/'],token)>=0
                    let resultToken.kind = 'cComment'
                else
                    let resultToken.kind = 'cppComment'
                endif
            else
                " It's an operator
                let resultToken.kind = 'cppOperatorPunctuator'
            endif
        endif

        " We have our token, let's add it to the result list
        call extend(result, [resultToken])
    endwhile

    return result
endfunc