Using s-expressions instead of XML | Hacker News Reader

Using s-expressions instead of XML

blog.fandle.com

29 pointsjazzdev23 comments

I'll just leave this here: http://okmij.org/ftp/Scheme/xml.html

It is still cool that he came up with this on his own.

Yea although he goes a little further with scheme.

To simplify things a bit, I would suggest the use of angle brackets instead of parentheses:

     <tagname <@ attr "value" attr2 "value2">
       <tagname2>
       <tagname3 "data">>

Going a bit further, we could require the `@`, and place it after the attributes list. This format has fewer special cases:

     <tagname attr "value" attr2 "value2" @
       <tagname2 @>
       <tagname3 @"data">>

Finally, we could remove the quotes around the raw strings ("data"), while inverting the brackets:

     >tagname attr "value" attr2 "value2" @
       >tagname2 @<
       >tagname3 @data<<

This may look twisted, but we stay very close to true S-expressions, while using the same escape characters we have in plain XML.

Close is not enough -- part of the attraction of s-expressions is that they are part of the language.

Thus you can use the language's escaping mechanisms, or put the escaping in your sxml->xml filter.

Preserving XML escapes is not actually a virtue. If you enter your data using S-expressions and validate on transformation, you will never have malformed XML.

(Not in any way that matters, anyway: http://validator.w3.org/check?uri=http%3A%2F%2Fwingolog.org%...)

edit: note also that in this context, s-expressions are more about language and less about parentheses; one can have them in python too: http://wingolog.org/pub/original/rename-to-index.py.txt, http://wingolog.org/software/original/

I think you may have been voted down because the form you have made is not close to S-expressions at all. Sexps are a direct way of encoding a tree of symbols, values and other trees - nothing else.

To give an example of why this is powerful, suppose I decide I need two different namespace of attributes. In your example, this is hard. I can either do:

  >tagname attr "value" @ attr "value2" !<

which special cases the parsing, and is hard to understand, or

  >tagname ns1:attr "value" ns2:attr "value2" @<

which makes it hard to process attributes (xml is like this)

Sexps have an easy, obvious solution

  (tagname (@ attr "value") (@other attr "value2"))

which I'd argue is clearer, more backward compatible, and doesn't require changing the parser.

Now this is only one example, but the point of sexps is they don't have an special cases, and they move the semantics out of the syntax and into the processing.

[meta] Can someone tell me which line did I cross? I thought my comment above was more than noise.

XML is not S-Expressions http://www.prescod.net/xml/sexprs.html

Nice!

And a good time to remember Erik Naggum's epic S-exps vs. XML rant:

http://harmful.cat-v.org/software/xml/s-exp_vs_XML

Robbery is not just another way of making a living, rape is not just another way of satisfying basic human needs, torture is not just another way of interrogation. And XML is not just another way of writing S-exps. There are some things in life that you do not do if you want to be a moral being and feel proud of what you have accomplished.

It distresses me that someone could type that up and think it reasonable and worth sharing.

If you find that so distressing, I wonder how you feel about this little tidbit from Swift:

  I have been assured by a very knowing American of my
  acquaintance in London, that a young healthy child well
  nursed, is, at a year old, a most delicious nourishing and
  wholesome food, whether stewed, roasted, baked, or boiled;
  and I make no doubt that it will equally serve in a
  fricasie, or a ragoust.

There are types of writing that do not revolve around literally and bluntly saying what you mean. I rather doubt that Mr. Naggum actually believes that XML is morally comparable to rape, in much the same way that I rather doubt Mr. Swift was seriously arguing that the Irish should begin eating their children.

To compare one of the greatest works of satire with Eric's vitriol belies a fundamental ignorance or insincerity.

What Eric wrote is puerile condescending crap.

But please, quote me an ironic passage of Shakespeare to reiterate Mr. Naggum's subtle genius.

To compare one of the greatest works of satire with Eric's vitriol belies a fundamental ignorance or insincerity.

I would hope neither, obviously, though it's entirely possible I'm wrong. I would rather just discuss the writer's actual point rather than his stylistic decisions or personality flaws.

It distresses me that so many people can't appreciate Naggum's genius :(

Many people can appreciate any genius. They can't appreciate incoherent rants that wander from technical points to politics to cultural criticism to analogies doomed to distract any reader from the technical points.

You will be recognized as a genius when you make insightful arguments others can follow and which teach them something; not when you require them to wade through a pile of irrelevant drivel, as if it was a high school exercise in summarising.

Erik doesn't even make arguments. He just states things, like

   Remove the syntactic mess that is attributes. 
   (You will then find that you do not need them at all.)

Yeah, that's a really solid argument, that immediately convinced me that seperating syntactically between metadata elements and child data elements is not necessary...

Can you easily filter raw text out of a Sexp stream? I don't think so. It would require a full Sexp machine to read and "understand" the tree, whereas with markup-style data streams it takes a trivial regexp to extract text. (Why would you need this? Search/indexing, for example.)

An "sexp machine" is fairly simple.

And yes, you can use regexs on s-expressions in exactly the same way that you'd use them on xml/html/markup. And you'd run into the same errors.

Note that reading s-expressions and extracting things from the result is significantly easier and faster than using regexs.... (The same is true of xml.)

Regular expressions are a great hammer, but many things are not nails.

With XML or any markup you can ignore the hierarchy and easily strip out everything that's not text, while with Sexp's I'm afraid you can't. In order to extract raw text (or actually strings) from SXML you'd need to read and understand the tree structure, which is not possible with a single-regex run.

Oh really?

How do you know that the sequence of bytes http://foo.bar.com/a/c is text or an attribute value with a regexp? How about foo.bar? How about Times Roman or 16px? How about the four bytes html ?

I don't understand. How would sexps be any worse than xml in terms of regexes? They're not that different from a parsing point of view (are they?) Besides which, regexes can't parse arbitrarily nested structures anyway.

If by "raw text" you mean string values (i.e. the equivalent of text nodes in XML) I don't see this as being particularly difficult, I don't think I'd use a Regex for it but then I wouldn't use a Regex to extract text from an XML document/stream either.

Regex are practical, you can use them from the command line, in your SQL statements, with your browser's JS engine, or on the server side - everywhere. A full Sexpr parser, on the other hand, is a more complicated machine not readily available everywhere. You don't have it in the UNIX userland by default, in SQL implementations, etc.

Strings that look like text or markup may or may not be, depending on whether they are surrounded by comments, CDATA marked sections, or processing instructions. Most people who try this get it wrong, and end up with a tool that randomly fails on some documents. Getting it right requires a regex big enough to actually be an XML parser, and why write that from scratch rather than use one that's known to work (and understand well-formedness and namespaces as well)?

Reply on news.ycombinator.com