Hi David
Thanks for your insightful reply.
So you said, in order no to hijack XQuery talk, I'll response only to
the question on RELAX NG, and I'll leave rest of my thoughts to my blog.
The way for RELAX NG to work as a search pattern is just like how RegEx
works for string. RELAX NG can be seen as regular expression on nodes.
Of course when using RELAX NG for search, one does not use the whole
schema, instead he just write a rule that matches some nodes. There are
pros and cons comparing to XPath. The strength of RELAX NG kind of node
pattern is probably in searching for a range of nodes (with wildcard
node in between). Of course, such usage goes beyond RELAX NG. But if we
can have RegEx for string, we can also have RegEx for nodes. RELAX NG is
not designed for that kind of usage, but can be the basis for such work.
Regards
Henry
On 24/11/2011 9:26 PM, David Lee wrote:
Comments Below.
I dont want to co-opt this thread/group into a 'my language vs your
language'
especially since its XQuery talk :)
But some concepts are really compelling.
(below)
*From:*[email protected] [mailto:[email protected]] *On
Behalf Of *Henry Luo
*Sent:* Thursday, November 24, 2011 1:01 AM
*To:* [email protected]
*Cc:* 'Geert Josten'; [email protected]; [email protected]
*Subject:* Re: [xquery-talk] [ANN] Candle 0.10 Beta Release - a new
scripting language for XML and more
=============
·Converting the input and output of shell commands into structured
hierarchical data, instead of just bytes and lines in abstract syntax.
I've determined this is in general difficult to do. If by 'shell
commands' you mean reusing existing ones. I have found 'shell
commands' tend to need to be rewritten to effectively and efficiently
do this. I had a goal once of a general purpose abstraction to
convert line-oriented commands into XML oriented commands (both in and
out), and while early work showed it does 'work' (you can use various
general purpose text/xml transformation filters) the results are
non-ideal. Many command simply dont provide for (in or out) the
richness of the data they are processing in text mode, and/or are hard
to parse. And also, unlike the early Unix days ... running a
subprocess is expensive.
=============
·Converting the semantics of shell commands from text based to node
based, as being done in xmlsh; more can be added - like xdiff (diff on
XML tree), xgrep2 (search based on RELAX NG kind of node pattern,
instead of xpath). The more we have, the more value for people to
convert to this new x-shell platform.
================
RelaxNG as a search pattern ? This had not occurred to me . Could you
give an example of how RelaxNG could work as a search pattern ?
·Extending the XML data model to the file level, so that advanced
features like schema, XPath, XQuery and XSLT can be applied to files
and dir nodes as well.
=======================
I have mixed feelings on this. I did once (and may still do ) have
a vision of a lazy-evaluation node based model of the filesystem. The
"xls" command provides a snapshot but its not efficient for large
trees if your just poking around.
A lazy-load node based representation of the filesystem would be
useful and I've seen several examples (but can't recall any of them
now). Actually "cd" provides a basics of a cursor based node
traversal mechanism. But its not what one thinks of much ...
But why I have mixed feelings is that File Systems are both evolving
and becoming less relevant. Or more heterogeneous. For example "The
Web" is becoming a filesystem For example in xmlsh I've tried
(without 100% success) at treating any URL as a "file" in places
normally files would be expected. E.g "cat < http://www.mysite.com";
works 'as expected'. But I have not implemented "cd" on any but file
schemes. But have thought about it. What would it look like if you
did "cd http://www.mysite.com; ls" . What commands can handle this
? What limits of each kind of scheme (ftp,http etc. ) can actually do
this and what features of get/list/put do they support ? Its a mess.
Add to that the emergence of XML Databases and NoSQL and distributed
files store (like amazon S3) ... Equal access to all those is a goal
I have in mind. Why can't XQuery fn:doc() work with an S3 'file' ?
Well it can with some heavy lifting. But that heavy lifting is what
keeps people from actually doing it. I'd like to see the
normalization of 'The space of all resources' not just filesystems.
REST is trying to do that but independent work evolves so fast its
hard to keep up. And REST doesn't work with the local filesystem ! Arg.
-------------------
·Rather than inventing another DSL for shell scripting, Candle extend
XQuery to become a general-purpose scripting language. The advantage
is that users faces only one language, instead of many DSLs (XPath,
XSLT, XQuery, XProc, xmlsh).
I personally think this is a red herring. "Extending a DSL" and
"Implementing a new DSL" are the same thing, IMHO. Both require the
users to learn new things. There is a degree of difference between
an 'entirely new DSL' and "A DSL similar to but 'extended' from one
you already know' ... But if we are to reuse existing DSL's like
XPath, XSLT, XQuery, XProc etc. the end result is going to be a hybrid
at best.
I'll leave the discussion on pipelining to another day. It takes a
whole conference to discuss it, and still there isn't a satisfactory
agreement.
-David
----------------------------------------
David A. Lee
[email protected] <mailto:[email protected]>
http://www.xmlsh.org
_______________________________________________
[email protected]
http://x-query.com/mailman/listinfo/talk
_______________________________________________
[email protected]
http://x-query.com/mailman/listinfo/talk
