Documentation for the WHIRL Query Interpreter

STIR File Format

There are two syntaxes for STIR files. The first is useful for marking up HTML documents, are the control characters are of the form %xxx{ %xxx: %}. The syntax for a tuple is roughly as follows:

	%rel{ %field1: ... %field2: ... %fieldn: ... %}

eg, for the relation person(name:doc address:doc email:string)

	%person{
	   %name: William Cohen 
	   %address: AT&T Labs--Research 
	   %email: "wcohen@research.att.com"	
	%}

The second syntax is XML-style. That's what's used in the snapshot data. A tuple is encoded as follows:

	<rel> <field1/> ... <field/> ... <field/> ... </rel>

	<person>
	   <name/> William Cohen 
	   <address/> AT&T Labs--Research 
	   <email/> "wcohen@research.att.com"	
	</person>

Some notes:

In the XML style, empty elements are taken as column separators; everything immediately after the separator is put into the field named "foo". Element attributes (like name="bar") will not be recognized. Case is important (unlike HTML).
Everything is tokenized and stemmed, except for stuff in double quotes. In particular numerical fields won't be interpreted properly unless they are quoted (!)
Tuples can be nested. In this case the first field of of the nested tuple is automatically filled with the id of the document/string/number in which the tuple is nested.
The field names are optional; the default is to start with the first field (field 0), and move to the next when a '%:' token is encountered. Hence an alternate syntax is
```
	%rel{ ... %: ... %: ... %}
```
For example:
```
	%person{ William Cohen %: AT&T Labs %: "wcohen@research.att.com" %}
```
The equivalent for the XML-style syntax uses <NEXT/> as a column separator
```
	<person>
	   William Cohen <NEXT/> AT&T Labs--Research <NEXT/> "wcohen@research.att.com"	
	</person>
```