|
|
As we use XMLQUERY as search engine for TUSNELDA, parts of the
following specifications have been adopted from the XMLQUERY
documentation.
A query is an expression built up from atomic constants and operators
as specified below.
Atomic expressions
| Type | Precedence | Operator |
|
| BRACKET | 0 | . |
. matches any XML element |
| BRACKET | 0 | # |
# matches any piece of PCDATA |
| BRACKET | 0 | quoted string |
a string valued constant |
| BRACKET | 0 | numeric constant |
a numeric constant |
| BRACKET | 0 | name |
the name of an XML element or attribute |
Operators allowed in query language
There are really two sublanguages in the query language, one which
describes the tree structure of the elements you want to find, the
other is a language for expressing boolean conditions on the attribute
values of these elements.
Tree structure
| Type | Precedence | Operator |
|
| BINOPR | 200 | / |
A/B if a B is a child of an A |
| BINOPR | 100 | , |
A,B if a B is right sibling of an A |
| BINOPR | 150 | & |
A & B if an A and a B in any order |
| MONPOST | 400 | * |
A* is multiple occurances of A. This is context dependant, if
before a /, then this means multiple nested occurances, if not before
a /, then this means multiple sequential occurances. |
| MONPOST | 400 | ? |
A? if there is an optional occurance of A |
| BINOPR | 250 | | |
A|B if an A or a B |
Relational operators
| Type | Precedence | Operator |
|
| BINOPL | 300 | = |
string equality |
| BINOPL | 300 | == |
numeric equality |
| BINOPL | 300 | > |
greater than |
| BINOPL | 300 | < |
less than |
| BINOPL | 300 | >= |
greater than or equal |
| BINOPL | 300 | <= |
less than or equal |
| BINOPL | 300 | ~ |
string regular expression match |
| BINOPL | 300 | != |
string inequality |
| BINOPL | 300 | !== |
numeric inequality |
| BINOPL | 300 | !~ |
string regular expression non-match |
Arithmetic and string operators
| Type | Precedence | Operator |
|
| BINOPL | 500 | + |
numeric plus |
| BINOPL | 500 | - |
numeric minus |
Other constants and operators
| Type | Precedence | Operator |
|
| MONPRE | 400 | % |
introduces variables |
| MONPRE | 400 | $ |
introduces variables (as %) |
| MONPRE | 400 | - |
numeric negation |
| MONPOST | 270 | ! |
marking something to be saved |
| MONPOST | 400 | [ ] |
A[B] if there is an A which satisfies boolean condition B |
| BRACKET | 0 | ( ) |
general purpose brackets |
| BRACKET | 0 | ^ |
Marks the beginning (i.e. before the first child) of an element |
| BRACKET | 0 | $ |
Marks the end (i.e. after the last child) of an element |
In more detail, the query language syntax is described below. Text
in green are literal operators, text in
italics are nonterminals in the grammar. The grammar below is
ambiguous, as usual ambiguities are resolved using the operator
precedences given above (operators with higher precedence bind more
tightly).
The syntax of an tree_expression is
| tree_expression |
:== tree_expression / tree_expression
|
|
:== tree_expression & tree_expression
|
|
:== tree_expression , tree_expression
|
|
:== tree_expression | tree_expression
|
|
:== ( tree_expression )
|
|
:== tree_expression *
|
|
:== tree_expression !
|
|
:== element_expression
|
|
:== ^
|
:== $
|
The syntax of an element_expression is
| element_expression |
:==
element_name_expression ( [ boolean_expression ] )?
|
|
:== # string_bool_op string_expression
|
| element_name_expression |
:== .
|
|
:== #
|
|
:== (element_name_expression
)
|
|
:== $ variable_name
|
|
:== % variable_name
|
|
:== element_name
|
|
:== element_name_expression
| element_name_expression
|
| boolean_expression |
:== boolean_expression , boolean_expression |
|
:== boolean_expression |
boolean_expression |
|
:== string_expression string_bool_op
string_expression |
|
:== number_expression number_bool_op
number_expression |
|
:== $ variable_name
|
|
:== % variable_name
|
|
:== number_expression
|
| number_expression |
:== number_expression arith_op
number_expression
|
|
:== number_literal
|
| string_expression |
:== number_literal
|
|
:== quoted_string_literal
|
|
:== attribute_name
|
|
:== #
|
|
:== $ variable_name
|
|
:== % variable_name
|
| number_bool_op |
:== == | !== |
< | > |
<= | >=
|
| string_bool_op |
:== = | != |
~ | !~
|
To be continued.
.*/sp!/.*/#~"Idefiks"
matches every <sp> element which contains the text string
"Idefiks" (matches in the B8 Comic Corpus).
.*/sp!/.*/reg
matches every <sp> element which contains a
<reg> element (matches in the B8 sub-corpora).
.*/sp/.*/reg
same as above, but does not display the matching <sp> elements
(which must be indicated by the ! operator) but only the
number of matches found.
.*/sp!/.*/reg!
same as above, but displays each <sp> element once
for each <reg> element in it, highlighting the
<reg> element.
.*/sp!/.*/marked[type="deic-loc"]
matches every <sp> element which contains a local deictic
(matches in the B8 and B9 sub-corpora)
.*/.!/marked[type="deic-loc"]
matches any element which contains a local deictic as
immediate child.
.*/figure!/.*/situation/keywords/term/#~"forefinger"
matches every <figure> element which contains a
situational characterization with the keyword "forefinger" (matches in
the B8 Comic Corpus).
.*/figure!/.*/situation/keywords/(term/#~"forefinger" & term/#~"bent")
matches every <figure> element which contains a
situational characterization with the keywords "forefinger" and "bent"
(matches in the B8 Comic Corpus).
.*/((sp/spokenpar!/ptr[target=%X]),.*,figure[id=%X]!)
matches every spoken paragraph which contains a pointer to a figure and
the corresponding figure (matches in the B9 text Brasil - soccer
match (tv).
Last modified 11 March 2009. |
