zim-parser-0.2.0.0: Read and parse ZIM files

Copyright(c) Robbin C.
LicenseGPLv3
MaintainerRobbin C.
Stabilityunstable
Portabilityportable
Safe HaskellNone
LanguageHaskell2010

Codec.Archive.Zim.Parser

Contents

Description

This is a library for parsing ZIM (http://openzim.org) files. ZIM files contain offline web content (eg, Wikipedia) which can be browsed locally without an Internet connection.

The API is meant to be intuitive for normal use-cases.

To get content for "A/index.htm" from ZIM file "file.zim":

> mimeContent <- "file.zim" `getContent` Url "A/index.htm"
> :t mimeContent
mimeContent :: Maybe (B8.ByteString, BL.ByteString)
> print mimeContent
Just ("text/html", "<html><head>...</html>")

The above will open the file, parse the ZIM header, lookup the MIME type and content of the URL, close the file and return the MIME type and content as a pair. Note that content is a lazy bytestring.

The above operation should suffice for a simple webserver serving a ZIM file. For finer control, it is possible to cache and reuse the file handle and the ZIM header.

> hdl <- openBinaryFile "file.zim" ReadMode
> hdr <- getHeader hdl
> :t hdr
hdr :: ZimHeader
> (hdl, hdr) `getContent` Url "A/index.htm"
Just ("text/html", "<html><head>...</html>")

ZIM files of Wikimedia Foundation (Wikipedia, Wikibooks, etc) can be found at http://ftpmirror.your.org/pub/kiwix/zim.

Below is a full example of a Scotty web server that serves a ZIM file (specified on command line) on localhost port 3000:

{-# LANGUAGE OverloadedStrings #-}

import Control.Monad.IO.Class (liftIO)
import Data.Text.Lazy (toStrict, fromStrict)
import Data.Text.Encoding (decodeUtf8, encodeUtf8)
import System.Environment (getArgs)
import Network.HTTP.Types.Status (status404)
import Web.Scotty
import Codec.Archive.Zim.Parser (getMainPageUrl, getContent, Url(..))

main :: IO ()
main = do
    [fp] <- getArgs
    scotty 3000 $ do
      get "/" (redirectToZimMainPage fp)
      get (regex "^/(./.*)$") (serveZimUrl fp)
      notFound $ text "Invalid URL!"

redirectToZimMainPage :: FilePath -> ActionM ()
redirectToZimMainPage fp = do
    res <- liftIO $ getMainPageUrl fp
    case res of
      Nothing -> do
        status status404
        text "This ZIM file has no main page specified!"
      Just (Url url) -> redirect . fromStrict $ decodeUtf8 url

serveZimUrl :: FilePath -> ActionM ()
serveZimUrl fp = do
    url <- (encodeUtf8 . toStrict) <$> param "1"
    res <- liftIO $ fp `getContent` Url url
    case res of
      Nothing -> do
        liftIO . putStrLn $ "Invalid URL: " ++ show url
        status status404
        text $ "Invalid URL!"
      Just (mimeType, content) -> do
        liftIO . putStrLn $ "Serving: " ++ show url
        setHeader "Content-Type" (fromStrict $ decodeUtf8 mimeType)
        raw content

Feedback and contributions are welcome on http://github.com/robbinch/zim-parser.

Synopsis

Functions

getHeader :: RunZim h => h -> IO ZimHeader Source

getMimeList :: RunZim h => h -> IO MimeList Source

getMainPageUrl :: RunZim h => h -> IO (Maybe Url) Source

Returns URL of main page in ZIM. This URL can be used for redirecting to the actual page.

getCluster :: RunZim h => h -> ClusterNumber -> IO Cluster Source

getBlob :: RunZim h => h -> (ClusterNumber, BlobNumber) -> IO Blob Source

type MimeList = Array Int ByteString Source

List of Mime Types

mkNsTitle :: Char -> ByteString -> Title Source

Construct a Title with a Namespace.

mkNsTitlePrefix :: Char -> ByteString -> TitlePrefix Source

Construct a TitlePrefix with a Namespace.

mkNsUrl :: Char -> ByteString -> Url Source

Construct a Url with a Namespace.

class RunZim h Source

Instances of this class represent a Zim File and are able to perform ZIM operations (getMimeList, getContent, etc). Valid instances include a Handle to a ZIM file, a FilePath to a ZIM file, or a (Handle, ZimHeader) where ZimHeader is parsed previously (so it does not need to be reparsed).

Minimal complete definition

runZim

Instances

class ZimGetDE k where Source

Methods

getDE :: RunZim h => h -> k -> IO ZimDirEnt Source

Instances

class ZimSearchDE k where Source

Methods

searchDE :: RunZim h => h -> k -> IO [(Int, ZimDirEnt)] Source

Search for a Directory Entry on a RunZim. When searching for a:

Url
Returns either 0 (not found) or 1 element.
Title
Returns either 0 (not found) or 1 element.
TitlePrefix
Returns either 0 (not found) or 2 elements corresponding to lower and upper bound of titles containing the prefix.

Instances

class ZimGetContent k where Source

Methods

getContent :: RunZim h => h -> k -> IO (Maybe (ByteString, ByteString)) Source

Get (MIME type, Content). Note that Content is lazy.

Instances

Exceptions

data ZimException Source

Other than the below, ErrorCall can be thrown by LZMA library if there is a problem with decompression.

Constructors

ZimInvalidMagic

ZIM file has invalid magic number (anything other than 72173914).

ZimParseError String

There is an error in parsing.

ZimIncompleteInput

There is insufficient bytes required to parse.

ZimInvalidIndex Int

The given index (URL, title or cluster) is out of bounds for this ZIM file.

Instances

ZIM Header

data ZimHeader Source

See http://www.openzim.org/wiki/ZIM_file_format#Header for more details.

Constructors

ZimHeader 

Fields

zimMagicNumber :: Int

Magic Number of file (somewhat superfluous as getZimHeader will throw an exception if magic number is anything other than 72173914)

zimVersion :: Int

Version of ZIM header

zimUuid :: ByteString

UUID of file

zimArticleCount :: Int

Number of articles

zimClusterCount :: Int

Number of clusters

zimUrlPtrPos :: Integer

Position of sorted URL pointers

zimTitlePtrPos :: Integer

Position of sorted Title pointers

zimClusterPtrPos :: Integer

Position of Cluster pointers

zimMimeListPos :: Integer

Position of MIME list

zimMainPage :: Maybe Int

Index of main page

zimLayoutPage :: Maybe Int

Index of layout page

zimChecksumPos :: Integer

Position of MD5 checksum

Instances

ZIM Directory Entry

data ZimDirEntType Source

There are 4 types of directory entries. Most content in a ZIM file are usually ZimArticleEntry or ZimRedirectEntry.

Constructors

ZimArticleEntry 
ZimRedirectEntry 
ZimLinkTarget 
ZimDeletedEntry 

Instances

data ZimDirEnt Source

See http://www.openzim.org/wiki/ZIM_file_format#Directory_Entries for more details.

Constructors

ZimDirEnt 

Fields

zimDeType :: ZimDirEntType

Type of this Directory Entry

zimDeMimeType :: Int

Index into MIME list given by getZimMimeList

zimDeParameterLen :: Int

Parameter Length

zimDeNamespace :: Char

Namespace

zimDeRevision :: Int

Revision

zimDeRedirectIndex :: Maybe Int

Redirect Index (only applicable for ZimRedirectEntry)

zimDeClusterNumber :: Maybe Int

Content is stored in this Cluster Number (only applicable for ZimArticleEntry)

zimDeBlobNumber :: Maybe Int

Content is stored in this Blob Number (only applicable for ZimArticleEntry)

zimDeUrl :: ByteString

URL

zimDeTitle :: ByteString

Title

Instances

newtype UrlIndex Source

Wrapper for URL index

Constructors

UrlIndex Int 

Instances

newtype TitleIndex Source

Wrapper for Title index

Constructors

TitleIndex Int 

Instances

newtype ClusterNumber Source

Wrapper for Cluster number

Constructors

ClusterNumber Int 

Instances

newtype BlobNumber Source

Wrapper for Blob number

Constructors

BlobNumber Int 

Instances

newtype Cluster Source

Wrapper for Cluster

Constructors

Cluster 

Fields

unCluster :: ByteString
 

newtype Blob Source

Wrapper for Blob

Constructors

Blob 

Fields

unBlob :: ByteString
 

newtype Url Source

Wrapper for Url

Constructors

Url ByteString 

Instances

data Title Source

Wrapper for Title

Instances

data TitlePrefix Source

Wrapper for Title Prefix

Instances

ZIM file format

Following is a short summary of the ZIM file format. The authoritative reference is at http://www.openzim.org/wiki/ZIM_file_format.

1. ZIM header

This is an 80-byte header (see ZimHeader). Among other things, it contains file offsets to the below.

2. List of MIME types

This is a sequence of null-terminated strings (eg. text/html, text/javascript). The last string is zero length, so the end always consists of 2 consecutive null bytes.

3. List of URLs

This is a sequence of 8-byte file offsets, each pointing to a directory entry. This list is sorted by the directory entries' URL.

getZimDirEntByUrlIndex looks up this table to return a directory entry.

4. List of Titles

This is a sequence of 4-byte indices, each pointing to a URL above (which in turn point to a directory entry). This list is sorted by the directory entries' Title.

getZimDirEntByTitleIndex uses this table to return a directory entry.

5. Directory Entries

This is a sequence of Directory Entries (see ZimDirEnt). The first 2 bytes determine the type of this entry, which also determine the length. Contents include:

a. MIME type

This 2-byte field means:

0xffff
This directory entry is a ZimRedirectEntry.
0xfffe
This directory entry is a ZimLinkTarget.
0xfffd
This directory entry is a ZimDeletedEntry.
any other value
This directory entry is a ZimArticleEntry and this index into the MIME list from above determines its MIME type.

b. Namespace

This single character determines the directory entry's namespace. (eg. A for articles, I for images, etc.) The comprehensive list is at http://www.openzim.org/wiki/ZIM_file_format#Namespaces.

c. Cluster and Blob number

Only for ZimArticleEntry, this is the directory entry's Cluster and Blob number. The Cluster number is a 4-byte index into the list of Clusters below. The Blob number refers to a block inside the (decompressed) cluster. Together, they provide the content of this directory entry.

d. URL and Title

These 2 null-terminated strings represent the URL and Title of this directory entry respectively. If the Title is empty, it is taken to be the same as the URL.

6. List of Clusters

This is a list of 8-byte file offsets, each pointing to a cluster in the file. The end of a cluster is also the start of the next cluster. Therefore, the length of a cluster is the difference between the adjacent offsets. For the last cluster, the end is the Checksum file offset, as the Checksum is always the last 16 bytes of a ZIM file.

a. Compression Type

The first byte of the cluster determines if it is uncompressed (eg. PNG image) or compressed with LZMA (eg. HTML).

0 or 1
No compression
4
Compressed with LZMA

b. List of Blobs

This is a list of 4-byte offsets, each pointing inside this cluster. The end of a blob is also the start of the next blob. Therefore, the length of a blob is the difference between the adjacent offsets. The last offset points to the end of the data area so there is always one more offset than blobs.