Description of Parser problem

Parsing good HTML code is difficult. Parsing missformatted HTML is harder. Code MUST not crash. It MUST repair this HTML page. Often it must guess what the HTML writer want. It is complex to be compatible with all release of the HTML standard.
Analyzing some attribut of some tag, is a nightmare. Take care of '"' and "'" . A good test is to analyze : "\"" .

Partial Solve

• Most of tag is knowed. Be not confused: MAP tag is knowed but not acting.

Functions

• struct mark_up * HTMLLexem( char *str)
  Input a HTML string, return a list of entity. An entity is a TAG or text. In most case you don't need to call it.
• HtmlTextInfo * HTMLParseRepair( char *str)
  

  typedef struct _HtmlTextInfo {
          char * title;
          char * base_url;
          char * base_target;
          struct mark_up *mlist;
  } HtmlTextInfo;
  

  Return a repair markup list in mlist. Return global info on page, title, base_url and base_target from the HEAD section. In most case you MUST call it, analyze the mlist, and put it in the HTML widget for acting (displaying). All part of structure is allocated and must be free when not needed.
• char* ParseMarkTag(char *text, char *mtext, char *mtag) Return the value of a tag attribute. Exemple:

  	value = ParseMarkTag(mlist->start, "TABLE", "WIDTH");
  	if (value) width = atoi(value);
  

  If attribut is not specified, it return NULL. If attribut is specified, but has no value, return "". Else return the value associated with the passed mark tag.

TODO

• File : libhtmlw/HTMLparse.c
• Problem when scanning tag with attribute like "\""
• Find something for Cascaded Style Sheet.
• Extract more information from the HEAD section.
• In HTMLParseRepair, it is possible to process more check for TABLE and FRAMESET.
• Provide processing for MAP tag.