commit
bae1c31406
File diff suppressed because it is too large
Load Diff
@ -1 +1 @@
|
||||
Subproject commit 7d01cb01cb1a926ecb4c9c98b107ef3c26f59dfb
|
||||
Subproject commit 8289d0d07de6553bf4b900bf60e808ea3f7f59da
|
@ -1,8 +0,0 @@
|
||||
add_library(lurlparser
|
||||
LUrlParser.cpp
|
||||
LUrlParser.h
|
||||
)
|
||||
|
||||
create_target_directory_groups(lurlparser)
|
||||
|
||||
target_include_directories(lurlparser INTERFACE ${CMAKE_CURRENT_SOURCE_DIR})
|
@ -1,265 +0,0 @@
|
||||
/*
|
||||
* Lightweight URL & URI parser (RFC 1738, RFC 3986)
|
||||
* https://github.com/corporateshark/LUrlParser
|
||||
*
|
||||
* The MIT License (MIT)
|
||||
*
|
||||
* Copyright (C) 2015 Sergey Kosarevsky (sk@linderdaum.com)
|
||||
*
|
||||
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
* of this software and associated documentation files (the "Software"), to deal
|
||||
* in the Software without restriction, including without limitation the rights
|
||||
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
* copies of the Software, and to permit persons to whom the Software is
|
||||
* furnished to do so, subject to the following conditions:
|
||||
*
|
||||
* The above copyright notice and this permission notice shall be included in all
|
||||
* copies or substantial portions of the Software.
|
||||
*
|
||||
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
* SOFTWARE.
|
||||
*/
|
||||
|
||||
#include "LUrlParser.h"
|
||||
|
||||
#include <algorithm>
|
||||
#include <cstring>
|
||||
#include <stdlib.h>
|
||||
|
||||
// check if the scheme name is valid
|
||||
static bool IsSchemeValid( const std::string& SchemeName )
|
||||
{
|
||||
for ( auto c : SchemeName )
|
||||
{
|
||||
if ( !isalpha( c ) && c != '+' && c != '-' && c != '.' ) return false;
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
bool LUrlParser::clParseURL::GetPort( int* OutPort ) const
|
||||
{
|
||||
if ( !IsValid() ) { return false; }
|
||||
|
||||
int Port = atoi( m_Port.c_str() );
|
||||
|
||||
if ( Port <= 0 || Port > 65535 ) { return false; }
|
||||
|
||||
if ( OutPort ) { *OutPort = Port; }
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
// based on RFC 1738 and RFC 3986
|
||||
LUrlParser::clParseURL LUrlParser::clParseURL::ParseURL( const std::string& URL )
|
||||
{
|
||||
LUrlParser::clParseURL Result;
|
||||
|
||||
const char* CurrentString = URL.c_str();
|
||||
|
||||
/*
|
||||
* <scheme>:<scheme-specific-part>
|
||||
* <scheme> := [a-z\+\-\.]+
|
||||
* For resiliency, programs interpreting URLs should treat upper case letters as equivalent to lower case in scheme names
|
||||
*/
|
||||
|
||||
// try to read scheme
|
||||
{
|
||||
const char* LocalString = strchr( CurrentString, ':' );
|
||||
|
||||
if ( !LocalString )
|
||||
{
|
||||
return clParseURL( LUrlParserError_NoUrlCharacter );
|
||||
}
|
||||
|
||||
// save the scheme name
|
||||
Result.m_Scheme = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
if ( !IsSchemeValid( Result.m_Scheme ) )
|
||||
{
|
||||
return clParseURL( LUrlParserError_InvalidSchemeName );
|
||||
}
|
||||
|
||||
// scheme should be lowercase
|
||||
std::transform( Result.m_Scheme.begin(), Result.m_Scheme.end(), Result.m_Scheme.begin(), ::tolower );
|
||||
|
||||
// skip ':'
|
||||
CurrentString = LocalString+1;
|
||||
}
|
||||
|
||||
/*
|
||||
* //<user>:<password>@<host>:<port>/<url-path>
|
||||
* any ":", "@" and "/" must be normalized
|
||||
*/
|
||||
|
||||
// skip "//"
|
||||
if ( *CurrentString++ != '/' ) return clParseURL( LUrlParserError_NoDoubleSlash );
|
||||
if ( *CurrentString++ != '/' ) return clParseURL( LUrlParserError_NoDoubleSlash );
|
||||
|
||||
// check if the user name and password are specified
|
||||
bool bHasUserName = false;
|
||||
|
||||
const char* LocalString = CurrentString;
|
||||
|
||||
while ( *LocalString )
|
||||
{
|
||||
if ( *LocalString == '@' )
|
||||
{
|
||||
// user name and password are specified
|
||||
bHasUserName = true;
|
||||
break;
|
||||
}
|
||||
else if ( *LocalString == '/' )
|
||||
{
|
||||
// end of <host>:<port> specification
|
||||
bHasUserName = false;
|
||||
break;
|
||||
}
|
||||
|
||||
LocalString++;
|
||||
}
|
||||
|
||||
// user name and password
|
||||
LocalString = CurrentString;
|
||||
|
||||
if ( bHasUserName )
|
||||
{
|
||||
// read user name
|
||||
while ( *LocalString && *LocalString != ':' && *LocalString != '@' ) LocalString++;
|
||||
|
||||
Result.m_UserName = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
// proceed with the current pointer
|
||||
CurrentString = LocalString;
|
||||
|
||||
if ( *CurrentString == ':' )
|
||||
{
|
||||
// skip ':'
|
||||
CurrentString++;
|
||||
|
||||
// read password
|
||||
LocalString = CurrentString;
|
||||
|
||||
while ( *LocalString && *LocalString != '@' ) LocalString++;
|
||||
|
||||
Result.m_Password = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
CurrentString = LocalString;
|
||||
}
|
||||
|
||||
// skip '@'
|
||||
if ( *CurrentString != '@' )
|
||||
{
|
||||
return clParseURL( LUrlParserError_NoAtSign );
|
||||
}
|
||||
|
||||
CurrentString++;
|
||||
}
|
||||
|
||||
bool bHasBracket = ( *CurrentString == '[' );
|
||||
|
||||
// go ahead, read the host name
|
||||
LocalString = CurrentString;
|
||||
|
||||
while ( *LocalString )
|
||||
{
|
||||
if ( bHasBracket && *LocalString == ']' )
|
||||
{
|
||||
// end of IPv6 address
|
||||
LocalString++;
|
||||
break;
|
||||
}
|
||||
else if ( !bHasBracket && ( *LocalString == ':' || *LocalString == '/' ) )
|
||||
{
|
||||
// port number is specified
|
||||
break;
|
||||
}
|
||||
|
||||
LocalString++;
|
||||
}
|
||||
|
||||
Result.m_Host = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
CurrentString = LocalString;
|
||||
|
||||
// is port number specified?
|
||||
if ( *CurrentString == ':' )
|
||||
{
|
||||
CurrentString++;
|
||||
|
||||
// read port number
|
||||
LocalString = CurrentString;
|
||||
|
||||
while ( *LocalString && *LocalString != '/' ) LocalString++;
|
||||
|
||||
Result.m_Port = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
CurrentString = LocalString;
|
||||
}
|
||||
|
||||
// end of string
|
||||
if ( !*CurrentString )
|
||||
{
|
||||
Result.m_ErrorCode = LUrlParserError_Ok;
|
||||
|
||||
return Result;
|
||||
}
|
||||
|
||||
// skip '/'
|
||||
if ( *CurrentString != '/' )
|
||||
{
|
||||
return clParseURL( LUrlParserError_NoSlash );
|
||||
}
|
||||
|
||||
CurrentString++;
|
||||
|
||||
// parse the path
|
||||
LocalString = CurrentString;
|
||||
|
||||
while ( *LocalString && *LocalString != '#' && *LocalString != '?' ) LocalString++;
|
||||
|
||||
Result.m_Path = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
CurrentString = LocalString;
|
||||
|
||||
// check for query
|
||||
if ( *CurrentString == '?' )
|
||||
{
|
||||
// skip '?'
|
||||
CurrentString++;
|
||||
|
||||
// read query
|
||||
LocalString = CurrentString;
|
||||
|
||||
while ( *LocalString && *LocalString != '#' ) LocalString++;
|
||||
|
||||
Result.m_Query = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
CurrentString = LocalString;
|
||||
}
|
||||
|
||||
// check for fragment
|
||||
if ( *CurrentString == '#' )
|
||||
{
|
||||
// skip '#'
|
||||
CurrentString++;
|
||||
|
||||
// read fragment
|
||||
LocalString = CurrentString;
|
||||
|
||||
while ( *LocalString ) LocalString++;
|
||||
|
||||
Result.m_Fragment = std::string( CurrentString, LocalString - CurrentString );
|
||||
|
||||
CurrentString = LocalString;
|
||||
}
|
||||
|
||||
Result.m_ErrorCode = LUrlParserError_Ok;
|
||||
|
||||
return Result;
|
||||
}
|
@ -1,78 +0,0 @@
|
||||
/*
|
||||
* Lightweight URL & URI parser (RFC 1738, RFC 3986)
|
||||
* https://github.com/corporateshark/LUrlParser
|
||||
*
|
||||
* The MIT License (MIT)
|
||||
*
|
||||
* Copyright (C) 2015 Sergey Kosarevsky (sk@linderdaum.com)
|
||||
*
|
||||
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
* of this software and associated documentation files (the "Software"), to deal
|
||||
* in the Software without restriction, including without limitation the rights
|
||||
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
* copies of the Software, and to permit persons to whom the Software is
|
||||
* furnished to do so, subject to the following conditions:
|
||||
*
|
||||
* The above copyright notice and this permission notice shall be included in all
|
||||
* copies or substantial portions of the Software.
|
||||
*
|
||||
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
* SOFTWARE.
|
||||
*/
|
||||
|
||||
#pragma once
|
||||
|
||||
#include <string>
|
||||
|
||||
namespace LUrlParser
|
||||
{
|
||||
enum LUrlParserError
|
||||
{
|
||||
LUrlParserError_Ok = 0,
|
||||
LUrlParserError_Uninitialized = 1,
|
||||
LUrlParserError_NoUrlCharacter = 2,
|
||||
LUrlParserError_InvalidSchemeName = 3,
|
||||
LUrlParserError_NoDoubleSlash = 4,
|
||||
LUrlParserError_NoAtSign = 5,
|
||||
LUrlParserError_UnexpectedEndOfLine = 6,
|
||||
LUrlParserError_NoSlash = 7,
|
||||
};
|
||||
|
||||
class clParseURL
|
||||
{
|
||||
public:
|
||||
LUrlParserError m_ErrorCode;
|
||||
std::string m_Scheme;
|
||||
std::string m_Host;
|
||||
std::string m_Port;
|
||||
std::string m_Path;
|
||||
std::string m_Query;
|
||||
std::string m_Fragment;
|
||||
std::string m_UserName;
|
||||
std::string m_Password;
|
||||
|
||||
clParseURL()
|
||||
: m_ErrorCode( LUrlParserError_Uninitialized )
|
||||
{}
|
||||
|
||||
/// return 'true' if the parsing was successful
|
||||
bool IsValid() const { return m_ErrorCode == LUrlParserError_Ok; }
|
||||
|
||||
/// helper to convert the port number to int, return 'true' if the port is valid (within the 0..65535 range)
|
||||
bool GetPort( int* OutPort ) const;
|
||||
|
||||
/// parse the URL
|
||||
static clParseURL ParseURL( const std::string& URL );
|
||||
|
||||
private:
|
||||
explicit clParseURL( LUrlParserError ErrorCode )
|
||||
: m_ErrorCode( ErrorCode )
|
||||
{}
|
||||
};
|
||||
|
||||
} // namespace LUrlParser
|
@ -1,19 +0,0 @@
|
||||
From https://github.com/corporateshark/LUrlParser/commit/455d5e2d27e3946f11ad0328fee9ee2628e6a8e2
|
||||
|
||||
MIT License
|
||||
|
||||
===
|
||||
|
||||
Lightweight URL & URI parser (RFC 1738, RFC 3986)
|
||||
|
||||
(C) Sergey Kosarevsky, 2015
|
||||
|
||||
@corporateshark sk@linderdaum.com
|
||||
|
||||
http://www.linderdaum.com
|
||||
|
||||
http://blog.linderdaum.com
|
||||
|
||||
=============================
|
||||
|
||||
A tiny and lightweight URL & URI parser (RFC 1738, RFC 3986) written in C++.
|
Loading…
Reference in New Issue