Expand description
XPath 1.0 over a parsec-html Document, written from the W3C
recommendation: every axis, node test and function of the core
library, with the number, string and boolean rules of sections 3 and 4.
The tree is read as lxml reads an HTML page, which is what scrapers
expect: element and attribute names are the parser’s lowercase ones
(SVG’s clippath, viewbox) in no namespace, so //div and @href
match, //DIV and @HREF don’t, and a prefixed name (svg:rect)
matches nothing. An
XML document built into the same tree brings its namespaces through
Namespaces, and names with a prefix take theirs from the prefixes
a query is run with.
A query reads the document through an Index, which its owner
builds once per version of the tree. Functions beyond the core library
come from the caller through Extensions.
Shadow roots are left out, as document.evaluate leaves them out: the
XPath data model has no shadow trees, so the axes follow the node tree
and never cross into a host’s shadow root or out of one. A shadow tree
is queried on its own by indexing its root as one of Index::build’s
roots, where it’s a root node as a template’s contents are: / from
inside it is the shadow root, and its host isn’t an ancestor.
Structs§
- Args
- The arguments of a call to an extension function, worked out when asked for, as the core library’s are.
- Doc
- The document a query runs over: its tree, the tree’s index, and for an XML document its namespaces.
- Index
- The index of one version of a tree. Its owner builds it once and keeps it until the tree changes: every query reads it, none builds it.
- NoExtensions
- No functions beyond the core library.
- XPath
- A compiled XPath expression. It holds no state of any document, so one can be shared by many threads.
- XPath
Error
Enums§
- Item
- A node in XPath’s data model: a tree node, or an attribute (which the
tree keeps on its element rather than as a node of its own). Sixteen
bytes: a node-set can hold every node of a big page, so its name and
value are read from the document (
Doc::attr_at), not kept here. - Value
- What an expression evaluates to. A node-set is in document order, without duplicates.
- XValue
- A value a caller can bind to a
$variable.
Constants§
- CDATA
- COMMENT
- DOCTYPE
- DOCUMENT
- ELEMENT
- The node kinds, as the index stores them: the DOM’s nodeType numbers.
- PI
- TEXT
- WORK_
LIMIT - How much work one run may do before it gives up: nodes looked at on
axes and by
lang(), predicate tests, and nodes and every 16 bytes of text read for string-values. Some queries are quadratic on any engine (//p[following::p]), and a hostile page can make them run for hours. A unit takes one to twenty-odd nanoseconds, so this is a few seconds at most: still over ten times what the slowest query of the benchmark pages needs. - XML_NS
- The namespace the
xmlprefix is always bound to.
Traits§
- Extensions
- Functions beyond XPath 1.0’s core library, which a caller provides. A core function of the same name wins.
- Namespaces
- The namespaces of an XML document built into a parsec-html tree, which the tree leaves to its host. Ids are the host’s; 0 is no namespace.