Re: Re: find with XPath over file system

  Hello Oleg,

  thank you for comments.

On Mon, 31 Jan 2005 23:58:46 -0800 (PST)
[email protected] wrote:

> 
> > find with XPath over file system
> > http://uucode.com/texts/xfind/index.html
> 
> That is quite neat!
> 
> > Actions (for example, printing file names) are executed after
> > collecting files, not during tree traversal.
> 
> Why?

  I can't do it with txpath.

> 
> The converters (or transformers, whatever they are called) in sxpath
> can be quite general. Any function node->nodelist should do.  So you
> can print as you traverse. The last snippet in
> 
> http://sourceforge.net/mailarchive/forum.php?forum_id=599&viewmonth=200411&viewday=1&style=flat
> 
> can be a good example. Actually, it would be even better if we write
> it as
> 
> ((sxpath `(// (div 1)
> 	       ,(lambda (node . rest) (write (strings-of-node node)))))
> 	       doc)
> 

  Yes, it's simple with sxpath. But I use txpath, and

(txpath "//div[1]")

  gives one solid lambda which can't be affected.

  By the way, it's one of the reasons why I'd like to implement a method
to compile XPath to Scheme source code: it would be possible to modify
behaviour of XPath evaluation.

> 
> > There are some differences between the file system tree and the XML
> > tree. At first, the paths /bin and ../../../../bin may point to the
> > same directory, but they look different for users. At second, we can
> > lazily walk from the base directory to the root and at the same time
> > descent from the root. A special attention should be paid to a meeting
> > point. But I decided to ignore it due to the first issue, so nodes
> > from relative and absolute XPaths are never equal.
> 
> If both paths point within UFS, then taking stat() (or lstat) of both
> and comparing stat.dev and stat.ino fields should tell if they are the
> same. In UFS, the combination (stat.dev, stat.ino) is a unique
> identifier.

  Yes, it's easy to check if two stat'ed files are the same. But what we
should print to user:

/bin

  or

../../../../bin?

  Well, it's not very important. Actually, the problem appeared due
to my fault. I found the issue too late to fix it. I'll fix the bug
in the new version.

> Things are quite more problematic for NFS-mounted (and
> remounted) and foreign-mounted file systems.
> 
> 
> > In my applications, I'd like to represent tree-like structures as the
> > real Scheme lists to Guile. There are two issues here:
> >
> >  - structures are big enough. Making a full copy of data is overhead,
> >    so I'd like to instantiate tree branches on demand.
> >  - Scheme list functions such as car, cdr, for-each etc should
> >    work on exposed data.
> >
> > It's impossible to satisfy the both requirements. The problem is that
> > list functions check the type of arguments, but we have to use a
> > custom type for lazy instantiation of internal data as Scheme data.
> 
> It is true that R5RS permits an implementation to do an implicit
> `forcing' of a delayed computation

  It's interesting. I can't remember such things in R5RS. Would you please
point me to it?

> -- although only Gambit (with a
> special compilation flag) actually does that.

  Thank you, now I'll know it.

> But surely one can
> accomplish `lazy' evaluation just by defining special functions like
> fcar (to be a composition of car and force), fcdr, fnull, fmap,
> ffor-each. Or may want to use the stream SRFI, 
> 	http://srfi.schemers.org/srfi-40/
> 	http://srfi.schemers.org/srfi-40/post-mail-archive/msg00006.html

  When using special functions, I have to modify source code of libraries.
In my development, the sx-library is used without any modifications.

> 
> 
> 	Cheers,
> 	Oleg
> 


-- 
Oleg


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl