haskell
haskell@gentoo.org
This package provides a general purpose Haskell XML library using Expat to
do its parsing (<http://expat.sourceforge.net/> - a fast stream-oriented XML
parser written in C). It is extensible to any string type, with @String@,
@ByteString@ and @Text@ provided out of the box.
Basic usage: Parsing a tree (/Tree/), formatting a tree (/Format/).
Other features: Helpers for processing XML trees (/Proc/), trees annotated with
XML source location (/Annotated/), extended XML trees with comments,
processing instructions, etc (/Extended/), XML cursors (/Cursor/),
SAX-style parse (/SAX/), and access to the low-level interface in case speed
is paramount (/Internal.IO/).
The design goals are speed, speed, speed, interface simplicity and modularity.
For introduction and examples, see the /Text.XML.Expat.Tree/ module. For benchmarks,
<http://haskell.org/haskellwiki/Hexpat/>
If you want to do interactive I\/O, an obvious option is to use lazy parsing
with one of the lazy I\/O functions such as hGetContents. However, this can be
problematic in some applications because it doesn't handle I\/O errors properly
and can give no guarantee of timely resource cleanup. In these cases, chunked
I\/O is a better approach: Take a look at the /hexpat-enumerator/ package.
/IO/ is filed under /Internal/ because it's low-level and most users won't want
it. The other /Internal/ modules are re-exported by /Annotated/, /Tree/ and /Extended/,
so you won't need to import them directly.
Credits to Iavor Diatchki and the @xml@ (XML.Light) package for /Proc/ and /Cursor/.
Thanks to the many contributors.
BOUND VS. UNBOUND THREADS: GHC (at least versions 6.12.X) will spawn threads
if you call a safe FFI callback from an unbound thread. This can get out of
control in a busy application. To avoid this, from version 0.19.1 we now delegate
processing to a single worker thread if the calling thread is not bound.
This essentially means that hexpat currently won't exploit multicores very well.
It also means that hexpat may be more efficient on threads spawned with forkOS
(to give you a bound thread) rather than forkIO.
ChangeLog: 0.15 changes intended to fix a (rare) \"error: a C finalizer called back into Haskell.\"
that seemed only to happen only on ghc6.12.X; 0.15.1 Fix broken Annotated parse;
0.16 switch from mtl to transformers; 0.17 fix mapNodeContainer & rename some things.;
0.18 rename defaultEncoding to overrideEncoding. 0.18.3 formatG and indent were demanding list
items more than once (inefficient in chunked processing); 0.19 add Extended.hs;
0.19.1 fix a memory leak introduced in 0.19, delegate parsing to bound thread
if unbound (see note above); 0.19.2 include expat source code so \'cabal install\' just works
on Linux, Mac and Windows (thanks Jacob Stanley); 0.19.3 fix misconfiguration of expat
which broke entity parsing; 0.19.4 bump version constraint for text; 0.19.5 bump text
to < 0.12 and fix text-0.10.0.1 breakage; 0.19.6 dependency breakage with List;
0.19.7 ghc-7.2.1 compatibility; 0.19.8 fix space leak on lazy parse under ghc-7.2.1