katydid / relaxng

Translates RelaxNG to Relapse

Katydid

Katydid is a toolkit for trees.

Currently there are three tools in the katydid toolkit:

• Relapse: a regular expression type language for trees that matches up to 1000000s of records per second,
• A collection of parsers (protobuf, json, xml, reflected go structures, yaml) which are easily extendable, and
• A collection of encoders (protobuf, json, xml, reflected go structures) which are useful for dynamic transcoding.

more...

relaxng's Issues

namespaces are not supported

Adding namespaces has three problems.

1. Currently the xml parser does not return the namespaces. This can be fixed with this diff

diff --git a/parser/xml/parser.go b/parser/xml/parser.go
index d63e9bb..4e40e18 100644
--- a/parser/xml/parser.go
+++ b/parser/xml/parser.go
@@ -196,14 +196,21 @@ func (p *xmlParser) Bool() (bool, error) {

 func (p *xmlParser) String() (string, error) {
        if p.tok == nil && p.attrIndex < len(p.attrs) {
+               a := p.attrs[p.attrIndex]
                if p.attrValue {
-                       return p.textPrefix + p.attrs[p.attrIndex].Value, nil
+                       return p.textPrefix + a.Value, nil
                } else {
-                       return p.attrPrefix + p.attrs[p.attrIndex].Name.Local, nil
+                       if len(a.Name.Space) == 0 {
+                               return p.attrPrefix + a.Name.Local, nil
+                       }
+                       return p.attrPrefix + a.Name.Space + ":" + a.Name.Local, nil
                }
        }
        if s, ok := p.tok.(StartElement); ok {
-               return p.elemPrefix + s.Name.Local, nil
+               if len(s.Name.Space) == 0 {
+                       return p.elemPrefix + s.Name.Local, nil
+               }
+               return p.elemPrefix + s.Name.Space + ":" + s.Name.Local, nil
        }
        if c, ok := p.tok.(CharData); ok {
                return p.textPrefix + string(c), nil

This is so easy because the golang xml parsing library already resolves the namespaces for us. Probably with a preparse phase, which is technically some sort of cheating, because now we don't need to keep any state in the validator, this state is kept in the parser for us by the golang xml parser.

2. nsName in relaxNG only matches the namespace and not the element name, so in this case relapse will need to be able match prefixes on names. The hardest part about this would be to design the syntax and thinking about whether names should allow all types of functions and not just a very constrained subset.

3. Also xmlns attributes would need to be matched which is currently conveniently done as an optional attribute with any value. When namespaces are returned by the parser their name will also require a more complex and specific pattern.

No build with fresh katydid/katydid

With fresh katydid/katydid/relapse/* have next errors:

go test -v .
./funcs.go:35:74: cannot use token literal (type *token) as type *ast.Expr in argument to combinator.Value
./funcs.go:135:53: cannot use whitespace literal (type *whitespace) as type *ast.Expr in argument to combinator.Value
./funcs.go:139:53: cannot use whitespace literal (type *whitespace) as type *ast.Expr in argument to combinator.Value
./funcs.go:143:50: cannot use anytext literal (type *anytext) as type *ast.Expr in argument to combinator.Value
./funcs.go:147:77: cannot use text literal (type *text) as type *ast.Expr in argument to combinator.Value
./funcs.go:223:73: cannot use list literal (type *list) as type *ast.Expr in argument to combinator.Value`