mirror of
https://github.com/rocky-linux/peridot.git
synced 2024-11-16 10:41:25 +00:00
366 lines
12 KiB
Go
366 lines
12 KiB
Go
|
package xmlquery
|
||
|
|
||
|
import (
|
||
|
"bufio"
|
||
|
"encoding/xml"
|
||
|
"errors"
|
||
|
"fmt"
|
||
|
"io"
|
||
|
"net/http"
|
||
|
"regexp"
|
||
|
"strings"
|
||
|
|
||
|
"github.com/antchfx/xpath"
|
||
|
"golang.org/x/net/html/charset"
|
||
|
)
|
||
|
|
||
|
var xmlMIMERegex = regexp.MustCompile(`(?i)((application|image|message|model)/((\w|\.|-)+\+?)?|text/)(wb)?xml`)
|
||
|
|
||
|
// LoadURL loads the XML document from the specified URL.
|
||
|
func LoadURL(url string) (*Node, error) {
|
||
|
resp, err := http.Get(url)
|
||
|
if err != nil {
|
||
|
return nil, err
|
||
|
}
|
||
|
defer resp.Body.Close()
|
||
|
// Make sure the Content-Type has a valid XML MIME type
|
||
|
if xmlMIMERegex.MatchString(resp.Header.Get("Content-Type")) {
|
||
|
return Parse(resp.Body)
|
||
|
}
|
||
|
return nil, fmt.Errorf("invalid XML document(%s)", resp.Header.Get("Content-Type"))
|
||
|
}
|
||
|
|
||
|
// Parse returns the parse tree for the XML from the given Reader.
|
||
|
func Parse(r io.Reader) (*Node, error) {
|
||
|
return ParseWithOptions(r, ParserOptions{})
|
||
|
}
|
||
|
|
||
|
// ParseWithOptions is like parse, but with custom options
|
||
|
func ParseWithOptions(r io.Reader, options ParserOptions) (*Node, error) {
|
||
|
p := createParser(r)
|
||
|
options.apply(p)
|
||
|
for {
|
||
|
_, err := p.parse()
|
||
|
if err == io.EOF {
|
||
|
return p.doc, nil
|
||
|
}
|
||
|
if err != nil {
|
||
|
return nil, err
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
|
||
|
type parser struct {
|
||
|
decoder *xml.Decoder
|
||
|
doc *Node
|
||
|
space2prefix map[string]string
|
||
|
level int
|
||
|
prev *Node
|
||
|
streamElementXPath *xpath.Expr // Under streaming mode, this specifies the xpath to the target element node(s).
|
||
|
streamElementFilter *xpath.Expr // If specified, it provides further filtering on the target element.
|
||
|
streamNode *Node // Need to remember the last target node So we can clean it up upon next Read() call.
|
||
|
streamNodePrev *Node // Need to remember target node's prev so upon target node removal, we can restore correct prev.
|
||
|
reader *cachedReader // Need to maintain a reference to the reader, so we can determine whether a node contains CDATA.
|
||
|
}
|
||
|
|
||
|
func createParser(r io.Reader) *parser {
|
||
|
reader := newCachedReader(bufio.NewReader(r))
|
||
|
p := &parser{
|
||
|
decoder: xml.NewDecoder(reader),
|
||
|
doc: &Node{Type: DocumentNode},
|
||
|
space2prefix: make(map[string]string),
|
||
|
level: 0,
|
||
|
reader: reader,
|
||
|
}
|
||
|
// http://www.w3.org/XML/1998/namespace is bound by definition to the prefix xml.
|
||
|
p.space2prefix["http://www.w3.org/XML/1998/namespace"] = "xml"
|
||
|
p.decoder.CharsetReader = charset.NewReaderLabel
|
||
|
p.prev = p.doc
|
||
|
return p
|
||
|
}
|
||
|
|
||
|
func (p *parser) parse() (*Node, error) {
|
||
|
var streamElementNodeCounter int
|
||
|
|
||
|
for {
|
||
|
tok, err := p.decoder.Token()
|
||
|
if err != nil {
|
||
|
return nil, err
|
||
|
}
|
||
|
|
||
|
switch tok := tok.(type) {
|
||
|
case xml.StartElement:
|
||
|
if p.level == 0 {
|
||
|
// mising XML declaration
|
||
|
node := &Node{Type: DeclarationNode, Data: "xml", level: 1}
|
||
|
AddChild(p.prev, node)
|
||
|
p.level = 1
|
||
|
p.prev = node
|
||
|
}
|
||
|
// https://www.w3.org/TR/xml-names/#scoping-defaulting
|
||
|
for _, att := range tok.Attr {
|
||
|
if att.Name.Local == "xmlns" {
|
||
|
p.space2prefix[att.Value] = ""
|
||
|
} else if att.Name.Space == "xmlns" {
|
||
|
p.space2prefix[att.Value] = att.Name.Local
|
||
|
}
|
||
|
}
|
||
|
|
||
|
if tok.Name.Space != "" {
|
||
|
if _, found := p.space2prefix[tok.Name.Space]; !found {
|
||
|
return nil, errors.New("xmlquery: invalid XML document, namespace is missing")
|
||
|
}
|
||
|
}
|
||
|
|
||
|
attributes := make([]Attr, len(tok.Attr))
|
||
|
for i, att := range tok.Attr {
|
||
|
name := att.Name
|
||
|
if prefix, ok := p.space2prefix[name.Space]; ok {
|
||
|
name.Space = prefix
|
||
|
}
|
||
|
attributes[i] = Attr{
|
||
|
Name: name,
|
||
|
Value: att.Value,
|
||
|
NamespaceURI: att.Name.Space,
|
||
|
}
|
||
|
}
|
||
|
|
||
|
node := &Node{
|
||
|
Type: ElementNode,
|
||
|
Data: tok.Name.Local,
|
||
|
Prefix: p.space2prefix[tok.Name.Space],
|
||
|
NamespaceURI: tok.Name.Space,
|
||
|
Attr: attributes,
|
||
|
level: p.level,
|
||
|
}
|
||
|
|
||
|
if p.level == p.prev.level {
|
||
|
AddSibling(p.prev, node)
|
||
|
} else if p.level > p.prev.level {
|
||
|
AddChild(p.prev, node)
|
||
|
} else if p.level < p.prev.level {
|
||
|
for i := p.prev.level - p.level; i > 1; i-- {
|
||
|
p.prev = p.prev.Parent
|
||
|
}
|
||
|
AddSibling(p.prev.Parent, node)
|
||
|
}
|
||
|
// If we're in the streaming mode, we need to remember the node if it is the target node
|
||
|
// so that when we finish processing the node's EndElement, we know how/what to return to
|
||
|
// caller. Also we need to remove the target node from the tree upon next Read() call so
|
||
|
// memory doesn't grow unbounded.
|
||
|
if p.streamElementXPath != nil {
|
||
|
if p.streamNode == nil {
|
||
|
if QuerySelector(p.doc, p.streamElementXPath) != nil {
|
||
|
p.streamNode = node
|
||
|
p.streamNodePrev = p.prev
|
||
|
streamElementNodeCounter = 1
|
||
|
}
|
||
|
} else {
|
||
|
streamElementNodeCounter++
|
||
|
}
|
||
|
}
|
||
|
p.prev = node
|
||
|
p.level++
|
||
|
p.reader.StartCaching()
|
||
|
case xml.EndElement:
|
||
|
p.level--
|
||
|
// If we're in streaming mode, and we already have a potential streaming
|
||
|
// target node identified (p.streamNode != nil) then we need to check if
|
||
|
// this is the real one we want to return to caller.
|
||
|
if p.streamNode != nil {
|
||
|
streamElementNodeCounter--
|
||
|
if streamElementNodeCounter == 0 {
|
||
|
// Now we know this element node is the at least passing the initial
|
||
|
// p.streamElementXPath check and is a potential target node candidate.
|
||
|
// We need to have 1 more check with p.streamElementFilter (if given) to
|
||
|
// ensure it is really the element node we want.
|
||
|
// The reason we need a two-step check process is because the following
|
||
|
// situation:
|
||
|
// <AAA><BBB>b1</BBB></AAA>
|
||
|
// And say the p.streamElementXPath = "/AAA/BBB[. != 'b1']". Now during
|
||
|
// xml.StartElement time, the <BBB> node is still empty, so it will pass
|
||
|
// the p.streamElementXPath check. However, eventually we know this <BBB>
|
||
|
// shouldn't be returned to the caller. Having a second more fine-grained
|
||
|
// filter check ensures that. So in this case, the caller should really
|
||
|
// setup the stream parser with:
|
||
|
// streamElementXPath = "/AAA/BBB["
|
||
|
// streamElementFilter = "/AAA/BBB[. != 'b1']"
|
||
|
if p.streamElementFilter == nil || QuerySelector(p.doc, p.streamElementFilter) != nil {
|
||
|
return p.streamNode, nil
|
||
|
}
|
||
|
// otherwise, this isn't our target node, clean things up.
|
||
|
// note we also remove the underlying *Node from the node tree, to prevent
|
||
|
// future stream node candidate selection error.
|
||
|
RemoveFromTree(p.streamNode)
|
||
|
p.prev = p.streamNodePrev
|
||
|
p.streamNode = nil
|
||
|
p.streamNodePrev = nil
|
||
|
}
|
||
|
}
|
||
|
case xml.CharData:
|
||
|
p.reader.StopCaching()
|
||
|
// First, normalize the cache...
|
||
|
cached := strings.ToUpper(string(p.reader.Cache()))
|
||
|
nodeType := TextNode
|
||
|
if strings.HasPrefix(cached, "<![CDATA[") {
|
||
|
nodeType = CharDataNode
|
||
|
}
|
||
|
|
||
|
node := &Node{Type: nodeType, Data: string(tok), level: p.level}
|
||
|
if p.level == p.prev.level {
|
||
|
AddSibling(p.prev, node)
|
||
|
} else if p.level > p.prev.level {
|
||
|
AddChild(p.prev, node)
|
||
|
} else if p.level < p.prev.level {
|
||
|
for i := p.prev.level - p.level; i > 1; i-- {
|
||
|
p.prev = p.prev.Parent
|
||
|
}
|
||
|
AddSibling(p.prev.Parent, node)
|
||
|
}
|
||
|
p.reader.StartCaching()
|
||
|
case xml.Comment:
|
||
|
node := &Node{Type: CommentNode, Data: string(tok), level: p.level}
|
||
|
if p.level == p.prev.level {
|
||
|
AddSibling(p.prev, node)
|
||
|
} else if p.level > p.prev.level {
|
||
|
AddChild(p.prev, node)
|
||
|
} else if p.level < p.prev.level {
|
||
|
for i := p.prev.level - p.level; i > 1; i-- {
|
||
|
p.prev = p.prev.Parent
|
||
|
}
|
||
|
AddSibling(p.prev.Parent, node)
|
||
|
}
|
||
|
case xml.ProcInst: // Processing Instruction
|
||
|
if p.prev.Type != DeclarationNode {
|
||
|
p.level++
|
||
|
}
|
||
|
node := &Node{Type: DeclarationNode, Data: tok.Target, level: p.level}
|
||
|
pairs := strings.Split(string(tok.Inst), " ")
|
||
|
for _, pair := range pairs {
|
||
|
pair = strings.TrimSpace(pair)
|
||
|
if i := strings.Index(pair, "="); i > 0 {
|
||
|
AddAttr(node, pair[:i], strings.Trim(pair[i+1:], `"`))
|
||
|
}
|
||
|
}
|
||
|
if p.level == p.prev.level {
|
||
|
AddSibling(p.prev, node)
|
||
|
} else if p.level > p.prev.level {
|
||
|
AddChild(p.prev, node)
|
||
|
}
|
||
|
p.prev = node
|
||
|
case xml.Directive:
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// StreamParser enables loading and parsing an XML document in a streaming
|
||
|
// fashion.
|
||
|
type StreamParser struct {
|
||
|
p *parser
|
||
|
}
|
||
|
|
||
|
// CreateStreamParser creates a StreamParser. Argument streamElementXPath is
|
||
|
// required.
|
||
|
// Argument streamElementFilter is optional and should only be used in advanced
|
||
|
// scenarios.
|
||
|
//
|
||
|
// Scenario 1: simple case:
|
||
|
// xml := `<AAA><BBB>b1</BBB><BBB>b2</BBB></AAA>`
|
||
|
// sp, err := CreateStreamParser(strings.NewReader(xml), "/AAA/BBB")
|
||
|
// if err != nil {
|
||
|
// panic(err)
|
||
|
// }
|
||
|
// for {
|
||
|
// n, err := sp.Read()
|
||
|
// if err != nil {
|
||
|
// break
|
||
|
// }
|
||
|
// fmt.Println(n.OutputXML(true))
|
||
|
// }
|
||
|
// Output will be:
|
||
|
// <BBB>b1</BBB>
|
||
|
// <BBB>b2</BBB>
|
||
|
//
|
||
|
// Scenario 2: advanced case:
|
||
|
// xml := `<AAA><BBB>b1</BBB><BBB>b2</BBB></AAA>`
|
||
|
// sp, err := CreateStreamParser(strings.NewReader(xml), "/AAA/BBB", "/AAA/BBB[. != 'b1']")
|
||
|
// if err != nil {
|
||
|
// panic(err)
|
||
|
// }
|
||
|
// for {
|
||
|
// n, err := sp.Read()
|
||
|
// if err != nil {
|
||
|
// break
|
||
|
// }
|
||
|
// fmt.Println(n.OutputXML(true))
|
||
|
// }
|
||
|
// Output will be:
|
||
|
// <BBB>b2</BBB>
|
||
|
//
|
||
|
// As the argument names indicate, streamElementXPath should be used for
|
||
|
// providing xpath query pointing to the target element node only, no extra
|
||
|
// filtering on the element itself or its children; while streamElementFilter,
|
||
|
// if needed, can provide additional filtering on the target element and its
|
||
|
// children.
|
||
|
//
|
||
|
// CreateStreamParser returns an error if either streamElementXPath or
|
||
|
// streamElementFilter, if provided, cannot be successfully parsed and compiled
|
||
|
// into a valid xpath query.
|
||
|
func CreateStreamParser(r io.Reader, streamElementXPath string, streamElementFilter ...string) (*StreamParser, error) {
|
||
|
return CreateStreamParserWithOptions(r, ParserOptions{}, streamElementXPath, streamElementFilter...)
|
||
|
}
|
||
|
|
||
|
// CreateStreamParserWithOptions is like CreateStreamParser, but with custom options
|
||
|
func CreateStreamParserWithOptions(
|
||
|
r io.Reader,
|
||
|
options ParserOptions,
|
||
|
streamElementXPath string,
|
||
|
streamElementFilter ...string,
|
||
|
) (*StreamParser, error) {
|
||
|
elemXPath, err := getQuery(streamElementXPath)
|
||
|
if err != nil {
|
||
|
return nil, fmt.Errorf("invalid streamElementXPath '%s', err: %s", streamElementXPath, err.Error())
|
||
|
}
|
||
|
elemFilter := (*xpath.Expr)(nil)
|
||
|
if len(streamElementFilter) > 0 {
|
||
|
elemFilter, err = getQuery(streamElementFilter[0])
|
||
|
if err != nil {
|
||
|
return nil, fmt.Errorf("invalid streamElementFilter '%s', err: %s", streamElementFilter[0], err.Error())
|
||
|
}
|
||
|
}
|
||
|
parser := createParser(r)
|
||
|
options.apply(parser)
|
||
|
sp := &StreamParser{
|
||
|
p: parser,
|
||
|
}
|
||
|
sp.p.streamElementXPath = elemXPath
|
||
|
sp.p.streamElementFilter = elemFilter
|
||
|
return sp, nil
|
||
|
}
|
||
|
|
||
|
// Read returns a target node that satisfies the XPath specified by caller at
|
||
|
// StreamParser creation time. If there is no more satisfying target nodes after
|
||
|
// reading the rest of the XML document, io.EOF will be returned. At any time,
|
||
|
// any XML parsing error encountered will be returned, and the stream parsing
|
||
|
// stopped. Calling Read() after an error is returned (including io.EOF) results
|
||
|
// undefined behavior. Also note, due to the streaming nature, calling Read()
|
||
|
// will automatically remove any previous target node(s) from the document tree.
|
||
|
func (sp *StreamParser) Read() (*Node, error) {
|
||
|
// Because this is a streaming read, we need to release/remove last
|
||
|
// target node from the node tree to free up memory.
|
||
|
if sp.p.streamNode != nil {
|
||
|
// We need to remove all siblings before the current stream node,
|
||
|
// because the document may contain unwanted nodes between the target
|
||
|
// ones (for example new line text node), which would otherwise
|
||
|
// accumulate as first childs, and slow down the stream over time
|
||
|
for sp.p.streamNode.PrevSibling != nil {
|
||
|
RemoveFromTree(sp.p.streamNode.PrevSibling)
|
||
|
}
|
||
|
sp.p.prev = sp.p.streamNode.Parent
|
||
|
RemoveFromTree(sp.p.streamNode)
|
||
|
sp.p.streamNode = nil
|
||
|
sp.p.streamNodePrev = nil
|
||
|
}
|
||
|
return sp.p.parse()
|
||
|
}
|