Re: Linking Emacs with libxml2

[joakim]

Chad Brown <address@hidden> writes:

> On Sep 6, 2010, at 11:44 AM, Lennart Borgman wrote:
>
>> On Mon, Sep 6, 2010 at 5:21 PM, Lars Magne Ingebrigtsen <address@hidden> 
>> wrote:
>>> Apparently libxml2 comes with a parser for "real world" HTML, which is
>>> very intriguing:
>>> 
>>> http://www.xmlsoft.org/html/libxml-HTMLparser.html>> 
>>> If Emacs provided a native interface to this function, we could say
>>> 
>>> (parse-html "file.html")
>>> => (:html (:head ...) (:body ...))
>>> 
>>> and get a nice parse tree out very fast.  (Parsing HTML from Emacs Lisp
>>> is rather slow.)
>>> 
>>> Has this been discussed before and rejected?  It seems like an obvious
>>> idea, and would enable both easier extraction of data from HTML files,
>>> as well as writing a (simple) HTML renderer in Emacs Lisp.
>> 
>> It was discussed before here:
>> 
>>  http://lists.gnu.org/archive/html/emacs-devel/2007-06/msg01147.html> 
>> Wasn't there a problem with linking to external libraries at that time?
>
> Yes, there was.  The FSF [lawyers] recently determined that it would be 
> possible to use external libraries with some explicit marking of legal 
> status, 
> along the lines of what is used in GCC.  Looking back through the mail 
> archives, it seems that practical implementation is stuck waiting on an FFI 
> design/implementation.  I thought that one had been sketched out, but I'm 
> not finding it in the archives, so perhaps I am confused.   
>

We dont need a generic FFI to link with libxml2, so thats not hindering
adoption. What we would need is an interface specification for xml
parsing in Emacs, and a specific implementation for libxml2.

The problem is rather that adding more dependencies complicates
maintenance of Emacs, and that maintaining a process buffer style
interface is perceived as simpler. Given that sentiment, strong
arguments needs to be presented in favour of in-process handling.

FWIW my personal opinion is that linking with well established external
libraries is no big deal.

> *Chad

-- 
Joakim Verona