Switch to using libaegisub for character set detection. There are some bugs here but it seems to be more consistent than the last version. There are two remaining issues left: when the character set is unknown we need to provide a full list to choose from. The second: if the file is detected as US-ASCII we need to give a parse error to the user if we run into problems. Right now we'll load the file fine and disable a lot of options.. with no message to the user.

Originally committed to SVN as r4370.
This commit is contained in:
Amar Takhar 2010-05-29 02:25:19 +00:00
parent 9d854b69f3
commit e7b859b9f7
3 changed files with 32 additions and 172 deletions

View file

@ -1,4 +1,4 @@
// Copyright (c) 2007, Rodrigo Braz Monteiro
// Copyright (c) 2010, Amar Takhar
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
@ -39,7 +39,6 @@
// Headers
#include "config.h"
#ifdef WITH_UNIVCHARDET
#ifndef AGI_PRE
#include <fstream>
#include <list>
@ -49,132 +48,47 @@
#include <wx/intl.h>
#endif
#include "../universalchardet/nsCharSetProber.h"
#include <libaegisub/charset.h>
#include "charset_detect.h"
#include "text_file_reader.h"
#include "compat.h"
/// DOCME
struct CharDetResult {
/// DOCME
float confidence;
/// DOCME
wxString name;
/// @brief DOCME
/// @param par
/// @return
///
bool operator < (CharDetResult &par) { return confidence > par.confidence; }
};
/// @brief Get encoding
/// @param filename
/// @return
///
wxString CharSetDetect::GetEncoding(wxString filename) {
std::ifstream file;
#ifdef __WINDOWS__
file.open(filename.wc_str(),std::ios::in | std::ios::binary);
#else
file.open(wxFNCONV(filename),std::ios::in | std::ios::binary);
#endif
if (!file.is_open()) {
throw _T("Failed opening file for reading.");
wxLogDebug("Filename: %s", filename);
bool unknown = 0;
agi::charset::CharsetListDetected list;
agi::charset::CharsetListDetected::const_iterator i_lst;
try {
agi::charset::DetectAll(STD_STR(filename), list);
} catch (const agi::charset::UnknownCharset&) {
unknown = 1;
}
// Loop through it until it finds interesting lines
while (!file.eof() && !done()) {
char buffer[512];
file.read(buffer, 512);
size_t bytesRead = file.gcount();
HandleData(buffer, bytesRead);
}
// Flag as finished
DataEnd();
// Grab every result obtained
wxString local = wxLocale::GetSystemEncodingName();
std::list<CharDetResult> results;
bool gotLocal = false;
for (int i=0;i<NUM_OF_CHARSET_PROBERS;i++) {
if (mCharSetProbers[i]) {
float conf = mCharSetProbers[i]->GetConfidence();
// Only bother with those whose confidence is at least 1%
wxString curName = wxString(mCharSetProbers[i]->GetCharSetName(),wxConvUTF8);
if (conf > 0.01f || curName == local) {
results.push_back(CharDetResult());
results.back().name = curName;
results.back().confidence = conf;
}
}
}
// If you got more than one valid result, ask the user which he wants
if (results.size() > 1) {
// Add local
if (!gotLocal) {
results.push_back(CharDetResult());
results.back().name = local;
results.back().confidence = 0;
}
// Sort by confidence
results.sort();
/// @todo If the charset is unknown we need to display a complete list of character sets.
if (list.size() > 1) {
// Get choice from user
wxArrayString choices;
wxArrayString picked;
int i = 0;
for (std::list<CharDetResult>::iterator cur=results.begin();cur!=results.end();cur++) {
wxString name = (*cur).name;
if (picked.Index(name) == wxNOT_FOUND) {
picked.Add(name);
// Generate name
wxString choiceStr;
if ((*cur).confidence > 0.0f) choiceStr = wxString::Format(_T("%f%% - "),(*cur).confidence*100.0f);
else choiceStr = _T("Unknown - ");
choiceStr += name;
if (name == local) choiceStr += _T(" (local)");
for (i_lst = list.begin(); i_lst != list.end(); ++i_lst) {
choices.Add(lagi_wxString(i_lst->second));
}
// Insert
choices.Add(choiceStr);
i++;
if (i == 20) break;
}
}
int choice = wxGetSingleChoiceIndex(_("Aegisub could not narrow down the character set to a single one.\nPlease pick one below:"),_("Choose character set"),choices);
if (choice == -1) throw _T("Canceled");
// Retrieve name
i = 0;
for (std::list<CharDetResult>::iterator cur=results.begin();cur!=results.end();cur++,i++) {
if (i == choice) result = (*cur).name;
}
return choices.Item(choice);
}
// Return whatever it got
return result;
i_lst = list.begin();
return i_lst->second;
}
/// @brief Report
/// @param aCharset
///
void CharSetDetect::Report(const char* aCharset) {
// Store the result reported
result = wxString(aCharset,wxConvUTF8);
}
#endif // WITH_UNIVCHARDET

View file

@ -1,4 +1,4 @@
// Copyright (c) 2007, Rodrigo Braz Monteiro
// Copyright (c) 2010, Amar Takhar
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
@ -34,34 +34,17 @@
/// @ingroup utility
///
///////////
// Headers
#include "../universalchardet/nscore.h"
#include "../universalchardet/nsUniversalDetector.h"
#include "../universalchardet/nsMBCSGroupProber.h"
/// DOCME
/// @class CharSetDetect
/// @brief DOCME
///
/// DOCME
class CharSetDetect : public nsUniversalDetector {
/// @brief Detect character set of a file
class CharSetDetect {
private:
/// DOCME
/// Character set
wxString result;
void Report(const char* aCharset);
public:
CharSetDetect() : nsUniversalDetector(NS_FILTER_ALL) { };
/// @brief Get character set name.
/// @param filename File to check
/// @return Character set name
wxString GetEncoding(wxString filename);
/// @brief DOCME
///
PRBool done() const { return mDone; }
};

View file

@ -74,49 +74,12 @@ TextFileReader::~TextFileReader() {
}
wxString TextFileReader::GetEncoding(wxString const& filename) {
// Prepare
unsigned char b[4];
memset(b, 0, sizeof(b));
// Read four bytes from file
std::ifstream ifile;
#ifdef __WINDOWS__
ifile.open(filename.wc_str());
#else
ifile.open(wxFNCONV(filename));
#endif
if (!ifile.is_open()) {
return L"unknown";
}
ifile.read(reinterpret_cast<char *>(b),4);
ifile.close();
// Try to get the byte order mark from them
if (b[0] == 0xEF && b[1] == 0xBB && b[2] == 0xBF) return L"UTF-8";
else if (b[0] == 0xFF && b[1] == 0xFE && b[2] == 0x00 && b[3] == 0x00) return L"UTF-32LE";
else if (b[0] == 0x00 && b[1] == 0x00 && b[2] == 0xFE && b[3] == 0xFF) return L"UTF-32BE";
else if (b[0] == 0xFF && b[1] == 0xFE) return L"UTF-16LE";
else if (b[0] == 0xFE && b[1] == 0xFF) return L"UTF-16BE";
else if (b[0] == 0x2B && b[1] == 0x2F && b[2] == 0x76) return L"UTF-7";
// Try to guess UTF-16
else if (b[0] == 0 && b[1] >= 32 && b[2] == 0 && b[3] >= 32) return L"UTF-16BE";
else if (b[0] >= 32 && b[1] == 0 && b[2] >= 32 && b[3] == 0) return L"UTF-16LE";
// If any of the first four bytes are under 0x20 (the first printable character),
// except for 9-13 range, assume binary
for (int i=0;i<4;i++) {
if (b[i] < 9 || (b[i] > 13 && b[i] < 32)) return L"binary";
}
#ifdef WITH_UNIVCHARDET
// Use universalchardet library to detect charset
CharSetDetect det;
return det.GetEncoding(filename);
#else
// Fall back to local
return L"local";
#endif
wxString str(det.GetEncoding(filename));
wxLogDebug("Encoding: %s", str);
return str;
}
wchar_t TextFileReader::GetWChar() {