URL
https://opencores.org/ocsvn/openrisc/openrisc/trunk
Subversion Repositories openrisc
[/] [openrisc/] [trunk/] [rtos/] [ecos-2.0/] [tools/] [src/] [tools/] [configtool/] [standalone/] [wxwin/] [htmlparser.cpp] - Rev 403
Go to most recent revision | Compare with Previous | Blame | View Log
//####COPYRIGHTBEGIN#### // // ---------------------------------------------------------------------------- // Copyright (C) 1998, 1999, 2000 Red Hat, Inc. // // This program is part of the eCos host tools. // // This program is free software; you can redistribute it and/or modify it // under the terms of the GNU General Public License as published by the Free // Software Foundation; either version 2 of the License, or (at your option) // any later version. // // This program is distributed in the hope that it will be useful, but WITHOUT // ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or // FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for // more details. // // You should have received a copy of the GNU General Public License along with // this program; if not, write to the Free Software Foundation, Inc., // 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. // // ---------------------------------------------------------------------------- // //####COPYRIGHTEND#### // htmlparser.cpp : // //=========================================================================== //#####DESCRIPTIONBEGIN#### // // Author(s): julians // Contact(s): julians // Date: 2001/04/04 // Version: $Id: htmlparser.cpp,v 1.1.1.1 2004-02-14 13:28:46 phoenix Exp $ // Purpose: // Description: HTML parser/HTML Help file generator // Requires: // Provides: // See also: // Known bugs: // Usage: // //####DESCRIPTIONEND#### // //=========================================================================== // ============================================================================ // declarations // ============================================================================ // ---------------------------------------------------------------------------- // headers // ---------------------------------------------------------------------------- #ifdef __GNUG__ #pragma implementation "htmlparser.h" #endif // Includes other headers for precompiled compilation #include "ecpch.h" #ifdef __BORLANDC__ #pragma hdrstop #endif #include "wx/textfile.h" #include "wx/wfstream.h" #include "ecutils.h" #include "htmlparser.h" /* * wxSimpleHtmlAttribute * Representation of an attribute */ wxSimpleHtmlParser::wxSimpleHtmlParser() { m_topLevel = NULL; m_pos = 0; } wxSimpleHtmlParser::~wxSimpleHtmlParser() { Clear(); } bool wxSimpleHtmlParser::ParseFile(const wxString& filename) { wxTextFile textFile; if (textFile.Open(filename)) { wxString text; wxString line; int i; int count = textFile.GetLineCount(); for (i = 0; i < count; i++) { if (i == 0) line = textFile.GetFirstLine(); else line = textFile.GetNextLine(); text += line; if (i != (count - 1)) text += wxT("\n"); } #if 0 for ( line = textFile.GetFirstLine(); !textFile.Eof(); line = textFile.GetNextLine() ) { text += line; if (!textFile.Eof()) text += wxT("\n"); } #endif return ParseString(text); } else return FALSE; } bool wxSimpleHtmlParser::ParseString(const wxString& str) { Clear(); m_pos = 0; m_text = str; m_length = str.Length(); m_topLevel = new wxSimpleHtmlTag(wxT("TOPLEVEL"), wxSimpleHtmlTag_TopLevel); return ParseHtml(m_topLevel); } // Main recursive parsing function bool wxSimpleHtmlParser::ParseHtml(wxSimpleHtmlTag* parent) { while (!Eof()) { EatWhitespace(); if (IsComment()) { ParseComment(); } else if (IsDirective()) { wxSimpleHtmlTag* tag = ParseDirective(); if (tag) parent->AppendTag(tag); } else if (IsTagClose()) { wxSimpleHtmlTag* tag = ParseTagClose(); if (tag) parent->AppendTag(tag); } else if (IsTagStartBracket(GetChar(m_pos))) { wxSimpleHtmlTag* tag = ParseTagHeader(); if (tag) parent->AppendTag(tag); } else { // Just a text string wxString text; ParseText(text); wxSimpleHtmlTag* tag = new wxSimpleHtmlTag(wxT("TEXT"), wxSimpleHtmlTag_Text); tag->SetText(text); parent->AppendTag(tag); } } return TRUE; } // Plain text, up until an angled bracket bool wxSimpleHtmlParser::ParseText(wxString& text) { while (!Eof() && GetChar(m_pos) != wxT('<')) { text += GetChar(m_pos); m_pos ++; } return TRUE; } wxSimpleHtmlTag* wxSimpleHtmlParser::ParseTagHeader() { if (IsTagStartBracket(GetChar(m_pos))) { m_pos ++; EatWhitespace(); wxString word; ReadWord(word, TRUE); EatWhitespace(); wxSimpleHtmlTag* tag = new wxSimpleHtmlTag(word, wxSimpleHtmlTag_Open); ParseAttributes(tag); EatWhitespace(); if (IsTagEndBracket(GetChar(m_pos))) m_pos ++; return tag; } else return NULL; } wxSimpleHtmlTag* wxSimpleHtmlParser::ParseTagClose() { Matches(wxT("</"), TRUE); EatWhitespace(); wxString word; ReadWord(word, TRUE); EatWhitespace(); m_pos ++; wxSimpleHtmlTag* tag = new wxSimpleHtmlTag(word, wxSimpleHtmlTag_Close); return tag; } bool wxSimpleHtmlParser::ParseAttributes(wxSimpleHtmlTag* tag) { // Parse attributes of a tag header until we reach > while (!IsTagEndBracket(GetChar(m_pos)) && !Eof()) { EatWhitespace(); wxString attrName, attrValue; if (IsString()) { ReadString(attrName, TRUE); tag->AppendAttribute(attrName, wxEmptyString); } else if (IsNumeric(GetChar(m_pos))) { ReadNumber(attrName, TRUE); tag->AppendAttribute(attrName, wxEmptyString); } else { // Try to read an attribute name/value pair, or at least a name // without the value ReadLiteral(attrName, TRUE); EatWhitespace(); if (GetChar(m_pos) == wxT('=')) { m_pos ++; EatWhitespace(); if (IsString()) ReadString(attrValue, TRUE); else if (!Eof() && !IsTagEndBracket(GetChar(m_pos))) ReadLiteral(attrValue, TRUE); } if (!attrName.IsEmpty()) tag->AppendAttribute(attrName, attrValue); } } return TRUE; } // e.g. <!DOCTYPE ....> wxSimpleHtmlTag* wxSimpleHtmlParser::ParseDirective() { Matches(wxT("<!"), TRUE); EatWhitespace(); wxString word; ReadWord(word, TRUE); EatWhitespace(); wxSimpleHtmlTag* tag = new wxSimpleHtmlTag(word, wxSimpleHtmlTag_Directive); ParseAttributes(tag); EatWhitespace(); if (IsTagEndBracket(GetChar(m_pos))) m_pos ++; return tag; } bool wxSimpleHtmlParser::ParseComment() { // Eat the comment tag start Matches(wxT("<!--"), TRUE); while (!Eof() && !Matches(wxT("-->"), TRUE)) { m_pos ++; } return TRUE; } bool wxSimpleHtmlParser::EatWhitespace() { while (!Eof() && IsWhitespace(GetChar(m_pos))) m_pos ++; return TRUE; } bool wxSimpleHtmlParser::EatWhitespace(int& pos) { while (!Eof(pos) && IsWhitespace(GetChar(pos))) pos ++; return TRUE; } bool wxSimpleHtmlParser::ReadString(wxString& str, bool eatIt) { int pos = m_pos; if (GetChar(pos) == (int) '"') { pos ++; while (!Eof(pos) && GetChar(pos) != (int) '"') { // TODO: how are quotes escaped in HTML? str += (wxChar) GetChar(pos); pos ++; } if (GetChar(pos) == (int) '"') pos ++; if (eatIt) m_pos = pos; return TRUE; } else return FALSE; } bool wxSimpleHtmlParser::ReadWord(wxString& str, bool eatIt) { int pos = m_pos; if (!IsAlpha(GetChar(pos))) return FALSE; str += (wxChar) GetChar(pos) ; pos ++; while (!Eof(pos) && IsWordChar(GetChar(pos))) { str += (wxChar) GetChar(pos); pos ++; } if (eatIt) m_pos = pos; return TRUE; } bool wxSimpleHtmlParser::ReadNumber(wxString& str, bool eatIt) { int pos = m_pos; if (!IsNumeric(GetChar(pos))) return FALSE; str += (wxChar) GetChar(pos) ; pos ++; while (!Eof(pos) && IsNumeric(GetChar(pos))) { str += (wxChar) GetChar(pos); pos ++; } if (eatIt) m_pos = pos; return TRUE; } // Could be number, string, whatever, but read up until whitespace or end of tag (but not a quoted string) bool wxSimpleHtmlParser::ReadLiteral(wxString& str, bool eatIt) { int pos = m_pos; while (!Eof(pos) && !IsWhitespace(GetChar(pos)) && !IsTagEndBracket(GetChar(pos)) && GetChar(pos) != wxT('=')) { str += GetChar(pos); pos ++; } if (eatIt) m_pos = pos; return TRUE; } bool wxSimpleHtmlParser::IsTagClose() { return Matches(wxT("</")); } bool wxSimpleHtmlParser::IsComment() { return Matches(wxT("<!--")); } bool wxSimpleHtmlParser::IsDirective() { return Matches(wxT("<!")); } bool wxSimpleHtmlParser::IsString() { return (GetChar(m_pos) == (int) '"') ; } bool wxSimpleHtmlParser::IsWord() { return (IsAlpha(GetChar(m_pos))); } bool wxSimpleHtmlParser::IsTagStartBracket(int ch) { return (ch == wxT('<')); } bool wxSimpleHtmlParser::IsTagEndBracket(int ch) { return (ch == wxT('>')); } bool wxSimpleHtmlParser::IsWhitespace(int ch) { return ((ch == 13) || (ch == 10) || (ch == 32) || (ch == (int) '\t')) ; } bool wxSimpleHtmlParser::IsAlpha(int ch) { return (wxIsalpha((wxChar) ch) != 0); } bool wxSimpleHtmlParser::IsWordChar(int ch) { return (wxIsalpha((wxChar) ch) != 0 || ch == wxT('-') || ch == wxT('_') || IsNumeric(ch)); } bool wxSimpleHtmlParser::IsNumeric(int ch) { return (wxIsdigit((wxChar) ch) != 0 || ch == wxT('-') || ch == wxT('.')) ; } // Matches this string (case insensitive) bool wxSimpleHtmlParser::Matches(const wxString& tok, bool eatIt) { wxString text(m_text.Mid(m_pos, tok.Length())); bool success = (text.CmpNoCase(tok) == 0) ; if (success && eatIt) { m_pos += tok.Length(); } return success; } // Safe way of getting a character int wxSimpleHtmlParser::GetChar(size_t i) const { if (i >= m_length) return -1; return m_text[i]; } void wxSimpleHtmlParser::Clear() { if (m_topLevel) delete m_topLevel; m_topLevel = NULL; m_text = wxEmptyString; m_pos = 0; m_length = 0; } // Write this file void wxSimpleHtmlParser::Write(wxOutputStream& stream) { if (m_topLevel) m_topLevel->Write(stream); } bool wxSimpleHtmlParser::WriteFile(wxString& filename) { wxFileOutputStream fstream(filename); if (fstream.Ok()) { Write(fstream); return TRUE; } else return FALSE; } /* * wxSimpleHtmlTag * Representation of a tag or chunk of text */ wxSimpleHtmlTag::wxSimpleHtmlTag(const wxString& tagName, int tagType) { m_name = tagName; m_type = tagType; m_attributes = NULL; m_children = NULL; m_parent = NULL; m_next = NULL; } wxSimpleHtmlTag::~wxSimpleHtmlTag() { ClearAttributes(); ClearChildren(); } //// Operations void wxSimpleHtmlTag::ClearAttributes() { if (m_attributes) { wxSimpleHtmlAttribute* attr = m_attributes; while (attr) { wxSimpleHtmlAttribute* next = attr->m_next; attr->m_next = NULL; delete attr; attr = next; } m_attributes = NULL; } } wxSimpleHtmlAttribute* wxSimpleHtmlTag::FindAttribute(const wxString& name) const { wxSimpleHtmlAttribute* attr = m_attributes; while (attr) { if (attr->GetName().CmpNoCase(name) == 0) { return attr; } attr = attr->m_next; } return NULL; } void wxSimpleHtmlTag::AppendAttribute(const wxString& name, const wxString& value) { wxSimpleHtmlAttribute* attr = new wxSimpleHtmlAttribute(name, value); if (m_attributes) { // Find tail wxSimpleHtmlAttribute* last = m_attributes; while (last->m_next) last = last->m_next; last->m_next = attr; } else m_attributes = attr; } void wxSimpleHtmlTag::ClearChildren() { if (m_children) { wxSimpleHtmlTag* child = m_children; while (child) { wxSimpleHtmlTag* next = child->m_next; child->m_next = NULL; delete child; child = next; } m_children = NULL; } } void wxSimpleHtmlTag::AppendTag(wxSimpleHtmlTag* tag) { if (m_children) { // Find tail wxSimpleHtmlTag* last = m_children; while (last->m_next) last = last->m_next; last->m_next = tag; tag->m_parent = this; } else m_children = tag; } // Gets the text from this tag and its descendants wxString wxSimpleHtmlTag::GetTagText() { wxString text; if (m_children) { wxSimpleHtmlTag* tag = m_children; while (tag) { text += tag->GetTagText(); tag = tag->m_next; } return text; } else if (GetType() == wxSimpleHtmlTag_Text) return GetText(); else return wxEmptyString; } int wxSimpleHtmlTag::GetAttributeCount() const { int count = 0; wxSimpleHtmlAttribute* attr = m_attributes; while (attr) { count ++; attr = attr->m_next; } return count; } wxSimpleHtmlAttribute* wxSimpleHtmlTag::GetAttribute(int i) const { int count = 0; wxSimpleHtmlAttribute* attr = m_attributes; while (attr) { if (count == i) return attr; count ++; attr = attr->m_next; } return NULL; } int wxSimpleHtmlTag::GetChildCount() const { int count = 0; wxSimpleHtmlTag* tag = m_children; while (tag) { count ++; tag = tag->m_next; } return count; } bool wxSimpleHtmlTag::HasAttribute(const wxString& name, const wxString& value) const { wxSimpleHtmlAttribute* attr = FindAttribute(name); return (attr && (attr->GetValue().CmpNoCase(value) == 0)) ; } bool wxSimpleHtmlTag::HasAttribute(const wxString& name) const { return FindAttribute(name) != NULL ; } bool wxSimpleHtmlTag::GetAttributeValue(wxString& value, const wxString& attrName) { wxSimpleHtmlAttribute* attr = FindAttribute(attrName); if (attr) { value = attr->GetValue(); return TRUE; } else return FALSE; } // Search forward from this tag until we find a tag with this name & attribute wxSimpleHtmlTag* wxSimpleHtmlTag::FindTag(const wxString& tagName, const wxString& attrName) { wxSimpleHtmlTag* tag = m_next; while (tag) { if (tag->NameIs(tagName) && tag->FindAttribute(attrName)) return tag; tag = tag->m_next; } return NULL; } bool wxSimpleHtmlTag::FindTextUntilTagClose(wxString& text, const wxString& tagName) { wxSimpleHtmlTag* tag = this; while (tag) { if (tag->GetType() == wxSimpleHtmlTag_Close && tag->NameIs(tagName)) return TRUE; if (tag->GetType() == wxSimpleHtmlTag_Text) text += tag->GetText(); tag = tag->m_next; } return TRUE; } wxSimpleHtmlTag* wxSimpleHtmlTag::GetChild(int i) const { int count = 0; wxSimpleHtmlTag* tag = m_children; while (tag) { if (count == i) return tag; count ++; tag = tag->m_next; } return NULL; } void wxSimpleHtmlTag::Write(wxOutputStream& stream) { switch (GetType()) { case wxSimpleHtmlTag_Text: { stream << m_text; break; } case wxSimpleHtmlTag_Open: { stream << "<" << m_name; if (GetAttributeCount() > 0) stream << " "; int i; for (i = 0; i < GetAttributeCount(); i++) { wxSimpleHtmlAttribute* attr = GetAttribute(i); attr->Write(stream); if (i < GetAttributeCount() - 1) stream << " "; } stream << ">\n"; break; } case wxSimpleHtmlTag_Directive: { stream << "<!" << m_name << " "; int i; for (i = 0; i < GetAttributeCount(); i++) { wxSimpleHtmlAttribute* attr = GetAttribute(i); attr->Write(stream); if (i < GetAttributeCount() - 1) stream << " "; } stream << ">\n"; break; } case wxSimpleHtmlTag_Close: { stream << "</" << m_name << ">\n"; break; } default: { break; } } wxSimpleHtmlTag* tag = m_children; while (tag) { tag->Write(stream); tag = tag->m_next; } } void wxSimpleHtmlAttribute::Write(wxOutputStream& stream) { if (m_value.IsEmpty()) stream << m_name; else { stream << m_name; stream << "=\""; stream << m_value; stream << "\""; } }
Go to most recent revision | Compare with Previous | Blame | View Log