SVG-grammar.txt

XML GRAMMAR (with shortcuts; based on W3C XML 1.1 spec)

document = prolog element misc*
  prolog = xmldecl misc*
  misc = comment | pi
  comment = '<!--' ([^-] | '-' . [^-])* '-->'     IGNORE
  pi = '<?' name attr* '?>'          IGNORE ("processing instruction")
    xmldecl = '<?xml' attr* '?>'     IGNORE (treat as a PI)
element = emptytag | stag content etag
  emptytag = '<' name (attr)* '/>'
  stag = '<' name (attr)* '>'
  etag = '</' name '>'
  content = text? ((element | ref | cdsect | pi | comment) text?)*
    text = [^<&]*
    ref = '&#x' hexdigit+ ';'
        | '&' name ';'
    cdsect = '<![CDATA[' cdata ']]>'
    cdata = .*   (ENDING WITH ']]>')
name = [a-zA-Z_:] [a-zA-Z0-9.-]*
attr = name '=' value
  value = '"' ([^<&"] | ref)* '"'
        | "'" ([^<&'] | ref)* "'"

NOTES
- Whitespace can occur anywhere in a tag after the tag name
- Handle at least these entity refs:
    &lt; &gt; &amp;  (< > &)
    &apos; &quot;  (' ")

PARSING SVG+XML
- Skip over prolog
- Parse elements...  <svg> should be the outer element
- Generic element parsing:
  - Parse element name
  - Lookup in hash map, dispatch to parser function (including <svg>)
- Element parser functions
  - Parse each attribute
    - Lookup via hash map or conditional (as appropriate)
    - Typically parse as a number, store as a string, or ignore
    - Always parse "style", as child elements inherit
    - Always store "id" as a string