The basis has two parts:

The syntax for the basis is the name of the axis followed by a double colon followed by the node test. For example, the basis descendant::para is a list of the para element descendants of the context node: descendant specifies that each node in the basis must be a descendant of the context; para specifies that each node in the basis must be an element named para.

The order of the nodes in the basis is determined by the axis. The general principal is that nodes in an axis are ordered going away from the context node. An axis that can never contain a node that is after the context node in document order is ordered in reverse document order; otherwise, an axis is ordered in document order.

A predicate filters a list of nodes to produce a new list of nodes. For each node in the list to be filtered, the PredicateExpr is evaluated with that node as the context node and with the complete list of nodes to be filtered as the context node list; if PredicateExpr evaluates to true for that node, the node is included in the new list; otherwise, it is not included.

A PredicateExpr is evaluated by evaluating the Expr and converting the result to a boolean. If the result is a number, the result will be converted to true if the number is equal to the position of the context node in the context node list (as returned by the position function) and will be converted to false otherwise; if the result is not a number, then the result will be converted as if by a call to the boolean function. Thus a location path para[3] is equivalent to para[position()=3].

Here are some examples of location paths using abbreviated syntax:

The most important abbreviation is that child:: can be omitted from a location step. In effect child is the default axis. For example, a location path div/para is short for child::div/child::para.

There is also an abbreviation for attributes: attribute:: can be abbreviated to @. For example, a location path para[@type="warning"] is short for child::para[attribute::type="warning"] and so selects para children with a type attribute with value equal to warning.

// is short for /descendant-or-self::node()/. For example, //para is short for /descendant-or-self::node()/child::para and so will select any para element in the document (even a para element that is a document element will be selected by //para since the document element node is a child of the root node); div//para is short for div/descendant-or-self::node()/child::para and so will select all para descendants of div children.

A location step of . is short for self::node(). This is particularly useful in conjunction with //. For example, the location path .//para is short for

and so will select all para descendant elements of the context node.

Similarly, a location step of .. is short for parent::node(). For example, ../title is short for parent::node()/child::title and so will select the title children of the parent of the context node.

A VariableReference evaluates to the value to which the variable name is bound in the set of variable bindings in the context. It is an error if the variable is not bound to any value in the set of variable bindings in the expression context.

Parentheses may be used for grouping.

A FunctionCall expression is evaluated by evaluating each of the Arguments, converting each argument to the type required by the function, calling the named function from the function library that is part of the expression evaluation context passing it the converted arguments. The result of the FunctionCall expression is the result returned by the function.

An argument is converted to type string as if by calling the string function. An argument is converted to type number as if by calling the number function. An argument is converted to type boolean as if by calling the boolean function. An argument that is not of type node-set cannot be converted to a node-set. It is an error if the number or type of arguments is wrong.

A location path can be used as an expression. The expression returns the set of nodes selected by the path.

The | operator computes the union of its operands, which must be node-sets.

Square brackets are used to filter expressions in the same way that they are used in location paths. It is an error if the expression to be filtered does not evaluate to a node-set. The context node list used for evaluating the expression in square brackets is the node-set to be filtered listed in document order.

The / operator and // operators combine an arbitrary expression and a relative location path. It is an error if the expression does not evaluate to a node-set. The / operator does composition in the same way as when / is used in a location path. As in location paths, // is short for /descendant-or-self::node()/.

There are no types of objects that can be converted to node-sets.

An object of type boolean can have two values, true and false.

An or expression is evaluated by evaluating each operand and converting its value to a boolean. The result is true if either value is true and false otherwise.

An and expression is evaluated by evaluating each operand and converting its value to a boolean. The result is true if both values are true and false otherwise.

A RelationalExpr or an EqualityExpr is evaluated by comparing the objects that result from evaluating the operands. If both operands are node-sets, then the comparison will be true if and only if there is a node in the first node-set operand and a node in the second node-set operand such that the result of performing the comparison on the string values of the two nodes is true. If one operand is a node-set and the other is a number, then the comparison will be true if and only if there is a node in the node-set operand such that the result of performing the comparison on the number operand and on the result of converting the value of that node to a number using the number function is true. If one operand is a node-set and the other is a string, then the comparison will be true if and only if there is a node in the node-set operand such that the result of performing the comparison on the string value of the node and the other operand is true. If one operand is a node-set and the other is a boolean, then the comparison will be true if and only if the result of performing the comparison on the boolean operand and on the result of converting the node-set to a boolean using the boolean function is true.

When neither operand is a node-set, then the operands of an EqualityExpr are converted to a common type as follows and then compared. If at least one operand is a boolean, then each operand is converted to a boolean as if by applying the boolean function. Otherwise, if at least one operand is a number, then each operand is converted to a number as if by applying the number function. Otherwise, both operands are converted to strings as if by applying the string function. The = comparison will be true if and only if both operands are equal; the != comparison will be true if and only if both operands are not equal. Numbers are compared for equality according to IEEE 754. Two booleans are true if either both are true or both are false. Two strings are equal if and only if they both consist of the same sequence of UCS characters.

When neither operand is a node-set, then the operands of a RelationalExpr are each converted to a number and compared according to IEEE 754. The < comparison will be true if and only if the first number is less than the second number. The <= comparison will be true if and only if the first number is less than or equal to the second number. The > comparison will be true if and only if the first number is greater than the second number. The >= comparison will be true if and only if the first number is greater than or equal to the second number.

A number represents a floating-point number. A number can have any double-precision 64-bit format IEEE 754 value. These include a special Not-a-Number (NaN) value, positive and negative infinity, and positive and negative zero.

The numeric operators convert their operands to numbers as if by calling the number function.

The div operator performs floating-point division according to IEEE 754.

The mod operator returns the remainder from a truncating division. For example,

Strings consist of a sequence of zero or more characters, where a character is defined as in the XML Recommendation . A single character in XPath thus corresponds to a single Unicode 2.0 abstract character with a single corresponding Unicode scalar value; this is not the same thing as a 16-bit Unicode code value: the Unicode coded character representation for an abstract character with Unicode scalar value greater that U+FFFF is a pair of 16-bit Unicode code values (a surrogate pair). In many programming languages, a string is represented by a sequence of 16-bit Unicode code values; implementations of XPath in such languages must take care to ensure that a surrogate pair is correctly treated as a single XPath character.

When tokenizing, the longest possible token is always returned.

For readability, whitespace may be used in patterns even though not explicitly allowed by the grammar: ExprWhitespace may be freely added within patterns before or after any ExprToken.

A NodeType or FunctionName token is recognized only when the following token is (. An AxisName token is recognized only when the following token is ::. An OperatorName token or MultiplyOperator token is recognized as such only when there is a preceding token and the preceding token is not one of @, ::, (, [, , or an Operator.

The last function returns the number of nodes in the context node list.

The position function returns the position of the context node in the context node list. The first position is 1, and so the last position will be equal to last().

The count function returns the number of nodes in the argument node-set.

The id function selects elements by their unique ID (see ). When the argument to id is of type node-set, then the result is the union of the result of applying id to the string value of each of the nodes in the argument node-set. When the argument to id is of any other type, the argument is converted to a string as if by a call to the string function; the string is split into a whitespace-separated list of tokens (whitespace is any sequence of characters matching the production S); the result is a node-set containing the elements in the same document as the context node that have a unique ID equal to any of the tokens in the list.

The local-part function returns a string containing the local part of the name of the node in the argument node-set that is first in document order. If the node-set is empty or the first node has no name, an empty string is returned. If the argument is omitted, it defaults to a node-set with the context node as its only member.

The namespace function returns a string containing the namespace URI of the expanded name of the node in the argument node-set that is first in document order. If the node-set is empty, the first node has no name, or the expanded name has no namespace URI, an empty string is returned. If the argument is omitted, it defaults to a node-set with the context node as its only member.

The name function returns a string containing a QName representing the name of the node in the argument node-set that is first in document order. The QName must represent the name with respect to the namespace declarations in effect on the node whose name is being represented. Typically, this will be the form in which the name occurred in the XML source. This need not be the case if there are namespace declarations in effect on the node that associate multiple prefixes with the same namespace. However, an implementation may include information about the original prefix in its representation of nodes; in this case, an implementation can ensure that the returned string is always the same as the QName used in the XML source. If the argument it omitted, it defaults to a node-set with the context node as its only member.

The string function converts an object to a string as follows:

If the argument is omitted, it defaults to a node-set with the context node as its only member.

The concat function returns the concatenation of its arguments.

The starts-with function returns true if the first argument string starts with the second argument string, and otherwise returns false.

The contains function returns true if the first argument string contains the second argument string, and otherwise returns false.

The substring-before function returns the substring of the first argument string that precedes the first occurrence of the second argument string in the first argument string, or the empty string if the first argument string does not contain the second argument string. For example, substring-before("1999/04/01","/") returns 1999.

The substring-after function returns the substring of the first argument string that follows the first occurrence of the second argument string in the first argument string, or the empty string if the first argument string does not contain the second argument string. For example, substring-after("1999/04/01","/") returns 04/01, and substring-after("1999/04/01","19") returns 99/04/01.

The substring function returns the substring of the first argument starting at the position specified in the second argument with length specified in the third argument. For example, substring("12345",2,3) returns "234". If the third argument is not specified, it returns the substring starting at the position specified in the second argument and continuing to the end of the string. For example, substring("12345",2) returns "2345".

More precisely, each character in the string (see ) is considered to have a numeric position: the position of the first character is 1, the position of the second character is 2 and so on. The returned substring contains those characters for which the position of the character is greater than or equal to the second argument and, if the third argument is specified, less than the sum of the second and third arguments; the comparisons and addition used for the above follow the standard IEEE 754 rules. Thus:

The string-length returns the number of characters in the string (see ). If the argument is omitted, it defaults to the context node converted to a string, in other words the value of the context node.

The normalize function returns the argument string with white space normalized by stripping leading and trailing whitespace and replacing sequences of whitespace characters by a single space. Whitespace characters are the same allowed by the S production in XML. If the argument is omitted, it defaults to the context node converted to a string, in other words the value of the context node.

The translate function returns the first argument string with occurrences of characters in the second argument string replaced by the character at the corresponding position in the third argument string. For example, translate("bar","abc","ABC") returns the string BAr. If there is a character in the second argument string with no character at a corresponding position in the third argument string (because the second argument string is longer than the third argument string), then occurrences of that character in the first argument string are removed. For example, translate("--aaa--","abc-","ABC") returns "AAA". If a character occurs more than once in second argument string, then the first occurrence determines the replacement character. If the third argument string is longer than the second argument string, then excess characters are ignored.

The boolean function converts its argument to a boolean as follows:

The not function returns true if its argument is false, and false otherwise.

The true function returns true.

The false function returns false.

The lang function returns true or false depending on whether the language of the context node as specified by xml:lang attributes is the same as or is a sublanguage of the language specified by the argument string. The language of the context node is determined by the value of the xml:lang attribute on the context node, or, if the context node has no xml:lang attribute, by the value of the xml:lang attribute on the nearest ancestor of the context node that has an xml:lang attribute. If there is no such attribute, then lang returns false. If there is such an attribute, then lang returns true if the attribute value is equal to the argument ignoring case, or if there is some suffix starting with - such that the attribute value is equal to the argument ignoring that suffix of the attribute value and ignoring case. For example, lang("en") would return true if the context node is any of these five elements:

The number function converts its argument to a number as follows:

If the argument is omitted, it defaults to a node-set with the context node as its only member.

The sum function returns the sum of the values of the nodes in the argument node-set.

The floor function returns the largest (closest to positive infinity) number that is not greater than the argument and that is an integer.

The ceiling function returns the smallest (closest to negative infinity) number that is not less than the argument and that is an integer.

The round function returns the number that is closest to the argument and that is an integer. If there are two such numbers, then the one that is even is returned.

The root node is the root of the tree. It does not occur anywhere else in the tree. The element node for the document element is a child of the root node. The root node also has as children processing instruction and comment nodes for processing instructions and comments that occur in the prolog and after the end of the document element.

The value of the root node is the value of the document element.

There is an element node for every element in the document. An element has an expanded name consisting of a local name and a possibly null URI reference (see ); the URI reference will be null if the element type name has no prefix and there is no default namespace in scope. A relative URI reference should be resolved into an absolute URI reference during namespace processing.

The children of an element node are the element nodes, comment nodes, processing instruction nodes and text nodes for its content. Entity references to both internal and external entities are expanded. Character references are resolved.

The descendants of an element node are the children of the element node and the descendants of the children that are element nodes.

The value of an element node is the string that results from concatenating all characters that are descendants of the element node in the order in which they occur in the document.

The set of all element nodes in a document can be ordered according to the order of the start-tags of the elements in the document; this is known as document order.

Each element node has an associated set of attribute nodes. A defaulted attribute is treated the same as a specified attribute. If an attribute was declared for the element type in the DTD, but the default was declared as #IMPLIED, and the attribute was not specified on the element, then the element's attribute set does not contain a node for the attribute.

An attribute node has an expanded name and has a string value. The expanded name consists of a local name and a possibly null URI reference (see ); the URI reference will be null if the specified attribute name did not have a prefix. The value is the normalized value as specified by the XML Recommendation . An attribute whose normalized value is a zero-length string is not treated specially: it results in an attribute node whose value is a zero-length string.

There are no attribute nodes corresponding to attributes that declare namespaces (see ).

Each element has an associated set of namespace nodes, one for each namespace prefix that is in scope for the element and one for the default namespace if one is in scope for the element. This means that an element will have a namespace node:

A namespace node has a name which is a string giving the prefix. This is empty if the namespace node is for the default namespace. A namespace node also has a value, which is the namespace URI. If the namespace declaration specifies a relative URI, then the resolved absolute URI is used as the value.

There is a processing instruction node for every processing instruction.

A processing instruction has a name. This is a string equal to the processing instruction's target. It also has a value. This is a string equal to the part of the processing instruction following the target and any whitespace. It does not include the terminating ?>.

There is a comment node for every comment.

A comment has a value. This is a string equal to the text of the comment not including the opening <!-- or the closing -->.

Character data is grouped into text nodes. As much character data as possible is grouped into each text node: a text node never has an immediately following or preceding sibling that is a text node. The value of a text node is the character data. A text node always has at least one character of data.

Each character within a CDATA section is treated as character data. Thus, <![CDATA[<]]> in the source document will treated the same as <. Both will result in a single < character in a text node in the tree. Thus, a CDATA section is treated as if the <![CDATA[ and ]]> were removed and every occurrence of < and & were replaced by < and & respectively.

Characters inside comments or processing instructions are not character data. Line-endings in external entities are normalized to #xA as specified in the XML Recommendation .