{-# LANGUAGE OverloadedStrings #-}
module Links(extractLinks, linkToText, Link(..), c_extractLinks) where

import Text.XML
import qualified Data.Map as M
import Network.URI
import Data.Text (Text, unpack, append, pack, replace, strip)
import qualified Data.Text.Foreign as FTxt
import Data.Maybe

import Types
import Foreign.StablePtr
import Foreign.C.String
import Foreign.Marshal.Array
import Control.Monad (forM)
import Control.Exception (catch)

import System.Directory -- For locating links.xml, suggestions.gmni
import System.FilePath
import System.IO (hPrint, stderr) -- For error reporting

-- For suggestions.gmni
import Data.Trie.Text (Trie)
import qualified Data.Trie.Text as Trie
import Data.List (nub, intercalate)
import Control.Concurrent.MVar (readMVar)
import Control.Concurrent (forkIO)

data Link = Link {
    label :: Text,
    title :: Text,
    href :: URI
}

linkToText :: Link -> Text
linkToText (Link label' title' href') =
    rmWs label' +++ "\t" +++ rmWs title' +++ "\t" +++ pack (show href')

rmWs text = strip $ replace "\t" " " $ replace "\n" " " text

extractLinks :: Document -> [Link]
extractLinks (Document prologue root misc) =
    extractMisc (prologueBefore prologue) ++ extractMisc (prologueAfter prologue) ++
    extractEl [] root ++ extractMisc misc

extractMisc :: [Miscellaneous] -> [Link]
extractMisc (MiscInstruction (Instruction target dat):misc)
    | Just uri <- parseURIReference $ unpack target = Link dat "" uri : extractMisc misc
extractMisc (_:misc) = extractMisc misc
extractMisc [] = []

extractEl path el@(Element (Name "details" _ _) _ childs) =
    [Link (nodesText summary' $ nodesText childs "") "+" nullURI {
        uriFragment = '#':'.':intercalate "." (map show $ reverse path)
    } | NodeElement summary@(Element (Name "summary" _ _) _ summary') <- childs]
extractEl path el@(Element _ _ children) =
    extractElAttr el "href" ++
    extractElAttr el "longdesc" ++
    extractElAttr el "src" ++
    extractNodes (0:path) children

extractElAttr (Element _ attrs children) attr
        | Just val <- attr `M.lookup` attrs',
            Just uri <- parseURIReference $ unpack val = [Link label' title' uri]
        | otherwise = []
    where
        label' = nodesText children $ M.findWithDefault "" "rel" attrs'
        title' = fromMaybe "" $ M.lookup "title" attrs'
        attrs' = M.mapKeys nameLocalName attrs

extractNodes p@(n:path) (NodeElement el:nodes) = extractEl p el ++ extractNodes (succ n:path) nodes
extractNodes path (NodeInstruction instruct:nodes) =
    extractMisc [MiscInstruction instruct] ++ extractNodes path nodes
extractNodes path (_:nodes) = extractNodes path nodes
extractNodes _ [] = []

(+++) = append
nodesText :: [Node] -> Text -> Text
nodesText (NodeElement (Element _ attrs children):nodes) def =
    nodesText children def +++ nodesText nodes def
nodesText (NodeContent text:nodes) def = text +++ nodesText nodes def
nodesText (_:nodes) def = nodesText nodes def
nodesText [] def = def

linksFromPage :: Page -> [Link]
linksFromPage Page {
        url = url',
        pageTitle = title',
        html = html',
        apps = apps',
        backStack = back', forwardStack = forward'
    } = -- TODO internationalize!
        link' "reload" title' url' :
        link' "aggiorna" title' url' : -- Italian
        link' "ladda om" title' url' : -- Swedish (Svenska)
        link' "last på nytt" title' url' : -- Norwegian Bokmål
        link' "reload without cache" "Fetch again from server without checking for a local copy" uncached :
        link' "aggiorna senza cache" "" uncached : -- Italian
        link' "ladda om utan cache" "hämta från servern igen utan att kolla efter en lokal kopia" uncached : -- Swedish (Svenska)
        link' "last på nytt uten mellomlager" "Last siden på nytt uten å bruke lokal kopi" uncached : -- Norwegian Bokmål
        [link' l t u | (t, u) <- head' back', l <- backLabels] ++
        [link' l t u | (t, u) <- head' forward', l <- forwardLabels] ++
        [link' n desc $ URI "app:" Nothing id "" "" | Application n _ desc id <- apps'] ++
        extractLinks html'
    where
        uncached = url' { uriScheme = "nocache+" ++ uriScheme url' }
        backLabels = ["back", {- Italian -} "indietro", {- Swedish -} "tillbaka",
            {- Norwegian Bokmål -}"tilbake"]
        forwardLabels = ["forward", {- Italian -} "avanti", {- Swedish -} "framåt",
            {- Norwegian Bokmål -} "forover", "videre"]

head' (a:_) = [a]
head' [] = []
link' l t h = Link (pack l) (pack t) h

readBookmarks :: IO Document
readBookmarks = do
    dir <- getXdgDirectory XdgData "rhapsode"
    let file = dir </> "links.xml"
    exists <- doesFileExist file
    if exists then Text.XML.readFile def file `catch` handleInvalid else nodoc
  where
    handleInvalid err@(InvalidXMLFile _ _) = hPrint stderr err >> nodoc
    nodoc = return $ Document (Prologue [] Nothing []) (Element "empty" M.empty []) []

-- | Write out a file of most frequently encountered unvisited links.
-- Hopefully this'll help surfers rely less on YouTube, et al's hueristics.
updateSuggestions :: Page -> IO ()
updateSuggestions page = do
    hist <- readMVar $ visitedURLs page
    let links = extractLinks $ html page
    let domain = maybe "" show $ uriAuthority $ url page

    dir <- getXdgDirectory XdgData "rhapsode"
    let path = dir </> "suggestions.gmni"
    exists <- doesFileExist path
    suggestions <- if not exists then return [] else do
        file <- Prelude.readFile path
        return [line' | line <- lines file, line'@(_:uri':_) <- [words line], not (pack uri' `Trie.member` hist)]

    let suggestions' = suggestions ++ nub [["=>", uri', domain] | link <- links,
            let uri' = uriToString id (href link) "", not (pack uri' `Trie.member` hist)]

    createDirectoryIfMissing True dir
    Prelude.writeFile path $ unlines $ map unwords suggestions'

-- C API
foreign export ccall c_extractLinks :: StablePtr Page -> IO (CArray CString)

c_extractLinks c_page = do
    page <- deRefStablePtr c_page
    forkIO $ updateSuggestions page -- background process for useful navigation aid.
    bookmarks <- readBookmarks
    ret <- forM (linksFromPage page ++ extractLinks bookmarks) $ \link -> do
        c_label <- text2cstring $ strip $ label link
        c_title <- text2cstring $ strip $ title link
        c_href <- newCString $ uriToString id (href link) ""
        return [c_label, c_title, c_href]
    nil <- newCString " "
    newArray0 nil $ concat ret

text2cstring txt = FTxt.withCStringLen txt $ \s -> (peekCStringLen s >>= newCString)