SharedBitmap::SharedBitmap(BPositionIO& data)
:
BReferenceable(),
fResourceID(-1),
fBuffer(NULL),
fSize(0),
fMimeType()
{
status_t status = data.GetSize(&fSize);
const off_t kMaxSize = 1024 * 1024;
if (status == B_OK && fSize > 0 && fSize <= kMaxSize) {
fBuffer = new(std::nothrow) uint8[fSize];
if (fBuffer != NULL) {
data.Seek(0, SEEK_SET);
off_t bytesRead = 0;
size_t chunkSize = std::min((off_t)4096, fSize);
while (bytesRead < fSize) {
ssize_t read = data.Read(fBuffer + bytesRead, chunkSize);
if (read > 0)
bytesRead += read;
else
break;
}
if (bytesRead != fSize) {
delete[] fBuffer;
fBuffer = NULL;
fSize = 0;
}
} else
fSize = 0;
} else {
fprintf(stderr, "SharedBitmap(): Stream too large: %" B_PRIi64
", max: %" B_PRIi64 "\n", fSize, kMaxSize);
}
fBitmap[0] = NULL;
fBitmap[1] = NULL;
fBitmap[2] = NULL;
}
_EXPORT ssize_t readfoldedline(BPositionIO &in, char **buffer, size_t *buflen)
{
ssize_t len = buflen && *buflen ? *buflen : 0;
char * buf = buffer && *buffer ? *buffer : NULL;
ssize_t cnt = 0; // Number of characters currently in the buffer.
char c;
status_t errorCode;
while (true)
{
// Make sure there is space in the buffer for two more characters (one
// for the next character, and one for the end of string NUL byte).
if (buf == NULL || cnt + 2 >= len)
{
char *temp = (char *)realloc(buf, len + 64);
if (temp == NULL) {
// Out of memory, however existing buffer remains allocated.
cnt = ENOMEM;
break;
}
len += 64;
buf = temp;
}
errorCode = in.Read (&c,1); // A really slow way of reading - unbuffered.
if (errorCode != 1) {
if (errorCode < 0) {
cnt = errorCode; // IO error encountered, just return the code.
} else {
// Really is end of file. Also make it end of line if there is
// some text already read in. If the first thing read was EOF,
// just return an empty string.
if (cnt > 0) {
buf[cnt++] = '\n';
if (buf[cnt-2] == '\r') {
buf[cnt-2] = '\n';
--cnt;
}
}
}
break;
}
buf[cnt++] = c;
if (c == '\n') {
// Convert CRLF end of line to just a LF. Do it before folding, in
// case we don't need to fold.
if (cnt >= 2 && buf[cnt-2] == '\r') {
buf[cnt-2] = '\n';
--cnt;
}
// If the current line is empty then return it (so that empty lines
// don't disappear if the next line starts with a space).
if (cnt <= 1)
break;
// if first character on the next line is whitespace, fold lines
errorCode = in.Read(&c,1);
if (errorCode == 1) {
if (c == ' ' || c == '\t')
buf[cnt-1] = c; // Replace \n with the white space character.
else {
// Not folding, we finished reading a whole line.
in.Seek(-1,SEEK_CUR); // Undo the look-ahead character read.
break;
}
} else if (errorCode < 0) {
cnt = errorCode;
break;
} else // No next line; at the end of the file. Return the line.
break;
}
}
if (buf != NULL && cnt >= 0)
buf[cnt] = '\0';
if (buffer)
*buffer = buf;
else if (buf)
free(buf);
if (buflen)
*buflen = len;
return cnt;
}
void CCellView::Read(BPositionIO& stream)
{
CSwapStream str(stream);
scChunk chunk;
long offset;
int *funcList, *styleList, *fontList, *formatList;
int funcCount, styleCount, fontCount, formatCount, borderFont = fBorderFontID;
bool warnForIncorrectFormula = true;
scCell cl;
styleCount = 0;
styleList = (int *)MALLOC(0);
fontCount = 0;
fontList = (int *)MALLOC(0);
formatCount = 0;
formatList = (int *)MALLOC(0);
funcList = NULL;
scCSElement *colStyles = NULL;
int colStyleCount = 0;
offset = 0;
StProgress progress(this, 1, pColorYellow, false);
StWriteLock lock(fContainer);
try
{
stream.Seek(offset, SEEK_SET);
str >> chunk;
if (chunk.type == kscVersion)
{
scVersion vers;
str >> vers;
if (vers.major != 3)
THROW((errTooNewFileFormat));
}
else
THROW((errUnknownFileFormat, ((CCellWindow *)Window())->EntryRef()->name));
do
{
if (stream.Seek(offset, SEEK_SET) < offset)
{
MStopAlert("File is too short").Go();
break;
}
str >> chunk;
offset += 4 + chunk.size;
switch (chunk.type)
{
case kscVersion:
break;
case kscHeader:
{
scHeader head;
str >> head;
funcList = (int *)CALLOC(head.functionCount, sizeof(int));
FailNil(funcList);
funcCount = head.functionCount;
progress.NewMax(head.cellCount);
break;
}
case kscView:
{
scView view;
str >> view;
if (view.windowRect.left > 0 && view.windowRect.top > 0)
{
BRect r;
{
r = BScreen().Frame();
}
r.InsetBy(4, 4);
r = r & view.windowRect;
if (r.IsValid() && r.Width() >= 300 && r.Height() >= 100)
{
Window()->MoveTo(r.left, r.top);
Window()->ResizeTo(r.Width(), r.Height());
}
}
if (view.position.h > 0 && view.position.v > 0)
fPosition = view.position;
fFrozen = view.frozen;
if (view.selection.IsValid())
fSelection = view.selection;
if (fSelection.Contains(view.curCell))
fCurCell = view.curCell;
borderFont = view.headingFont;
fShowGrid = (view.flags & kscShowGrid) != 0;
fShowBorders = (view.flags & kscShowHeadings) != 0;
if (!fShowBorders)
{
fCellBounds.left -= fBorderWidth;
fCellBounds.top -= fBorderHeight;
}
Window()->Show();
Window()->UpdateIfNeeded();
break;
}
case kscPrint:
break;
case kscWidths:
case kscHeights:
{
int count = chunk.size/(sizeof(short)*2);
scWidthElement *elems = (scWidthElement *)MALLOC(chunk.size);
FailNil(elems);
int k = count * sizeof(scWidthElement);
stream.Read(elems, k);
if (chunk.type == kscWidths)
fCellWidths.Read(count, elems);
else
fCellHeights.Read(count, elems);
FREE(elems);
break;
}
case kscColStyles:
{
colStyleCount = chunk.size/(sizeof(short)*2);
colStyles = (scCSElement *)MALLOC(chunk.size);
FailNil(colStyles);
for (int i = 0; i < colStyleCount; i++)
str >> colStyles[i];
break;
}
case kscName:
{
char buf[50];
scName *name = (scName *)buf;
int l = std::min((int)chunk.size, 50);
stream.Read(name, l);
range r;
memcpy(&r, name->reference+1, sizeof(cell));
if (name->reference[0] == valRange)
memcpy(&r.bottom, name->reference+5, sizeof(cell));
else
r.BotRight() = r.TopLeft();
swap_order(r);
if (fNames->count(name->name) == 0)
(*fNames)[CName(name->name)] = r;
break;
}
case kscFunc:
{
scFunc func;
str >> func;
if (!funcList) THROW((errCorruptedFile));
if (func.funcNr >= funcCount)
THROW((errCorruptedFile));
funcList[func.funcNr] = GetFunctionNr(func.name);
if (funcList[func.funcNr] == -1)
WarnForMissingFunction(func.name);
break;
}
case kscFont:
{
scFont font;
str >> font;
font_family fam;
font_style sty;
ReadCString(stream, sizeof(font_style), sty);
ReadCString(stream, sizeof(font_family), fam);
int *t = (int *)REALLOC(fontList, (fontCount+1)*sizeof(int));
FailNil(t);
fontList = t;
fontList[fontCount] = gFontSizeTable.GetFontID(
fam, sty, font.size, font.color);
fontCount++;
break;
}
case kscFormat:
{
scFormat format;
str >> format;
int *t = (int *)REALLOC(formatList, (formatCount+1)*sizeof(int));
FailNil(t);
formatList = t;
if (format.nr < eFirstNewFormat)
formatList[formatCount] = format.nr;
else
{
char fs[256];
ReadCString(stream, 255, fs);
formatList[formatCount] = gFormatTable.GetFormatID(fs);
}
formatCount++;
break;
}
case kscStyle:
{
scStyle style;
str >> style;
int *t = (int *)REALLOC(styleList, (styleCount+1)*sizeof(int));
FailNil(t);
styleList = t;
CellStyle cs;
if (style.font >= fontCount)
THROW((errCorruptedFile));
cs.fFont = fontList[style.font];
if (style.format >= formatCount)
THROW((errCorruptedFile));
cs.fFormat = formatList[style.format];
cs.fAlignment = style.align;
cs.fLowColor = style.lowColor;
cs.fLocked = (style.flags & kscLocked) != 0;
cs.fHidden = (style.flags & kscHidden) != 0;
styleList[styleCount] = gStyleTable.GetStyleID(cs);
styleCount++;
break;
}
case kscCellEmpty:
case kscCellBool:
case kscCellNumber:
case kscCellDateTime:
case kscCellText:
{
str >> cl;
Value val;
switch (chunk.type)
{
case kscCellBool:
{
bool b;
str >> b;
val = b;
break;
}
case kscCellNumber:
{
double d;
str >> d;
val = d;
break;
}
case kscCellDateTime:
{
time_t t;
str >> t;
val = t;
break;
}
case kscCellText:
{
// do nothing yet...
}
}
fContainer->NewCell(cl.loc, val, NULL);
if (cl.style >= styleCount)
/*THROW((errCorruptedFile))*/;
else
fContainer->SetCellStyleNr(cl.loc, styleList[cl.style]);
progress.Step();
break;
}
case kscString:
{
char s[256];
Value val;
stream.Read(s, std::min((int)chunk.size, 255));
s[std::min((int)chunk.size, 255)] = 0;
val = s;
fContainer->SetValue(cl.loc, val);
break;
}
case kscFormula:
{
CFormula form;
try
{
if (!funcList) THROW((errCorruptedFile));
form.Read(stream, funcList);
Value v;
form.Calculate(cl.loc, v, fContainer);
fContainer->SetCellFormula(cl.loc, form.CopyString());
}
catch (CErr& e)
{
CATCHED;
char s[32];
cl.loc.GetName(s);
#if DEBUG
printf("%s in formula of cell %s\n", (char *)e, s);
#endif
if (warnForIncorrectFormula)
{
char m[256];
sprintf(m, GetMessage(msgIncorrectFormula), s);
MAlert *a = new MWarningAlert(m, GetMessage(msgOK),
GetMessage(msgNoMoreWarnings));
if (a->Go() == 2)
warnForIncorrectFormula = false;
}
}
form.Clear();
break;
}
case kscChart:
ReadChart(stream, chunk.size);
case kscEnd:
break;
default:
MStopAlert("File contains errors").Go();
chunk.type = kscEnd;
break;
}
}
while (chunk.type != kscEnd);
// adjust the fields that couldn't be adjusted before
if (fontCount && borderFont < fontCount)
fBorderFontID = fontList[borderFont];
if (colStyles && styleList)
{
for (int i = 0; i < colStyleCount; i++)
{
if (colStyles[i].style >= 0 && colStyles[i].style < styleCount)
colStyles[i].style = styleList[colStyles[i].style];
}
fContainer->GetColumnStyles().Read(colStyleCount, colStyles);
}
}
status_t
AGMSBayesianSpamFilter::ProcessMailMessage (
BPositionIO** io_message,
BEntry* io_entry,
BMessage* io_headers,
BPath* io_folder,
const char* io_uid)
{
ssize_t amountRead;
attr_info attributeInfo;
const char *classificationString;
off_t dataSize;
BPositionIO *dataStreamPntr = *io_message;
status_t errorCode = B_OK;
int32 headerLength;
BString headerString;
BString newSubjectString;
BNode nodeForOutputFile;
bool nodeForOutputFileInitialised = false;
const char *oldSubjectStringPntr;
char percentageString [30];
BMessage replyMessage;
BMessage scriptingMessage;
team_id serverTeam;
float spamRatio;
char *stringBuffer = NULL;
char tempChar;
status_t tempErrorCode;
const char *tokenizeModeStringPntr;
// Set up a BNode to the final output file so that we can write custom
// attributes to it. Non-custom attributes are stored separately in
// io_headers.
if (io_entry != NULL && B_OK == nodeForOutputFile.SetTo (io_entry))
nodeForOutputFileInitialised = true;
// Get a connection to the spam database server. Launch if needed, should
// only need it once, unless another e-mail thread shuts down the server
// inbetween messages. This code used to be in InitCheck, but apparently
// that isn't called.
printf("Checking for Spam Server.\n");
if (fLaunchAttemptCount == 0 || !fMessengerToServer.IsValid ()) {
if (fLaunchAttemptCount > 3)
goto ErrorExit; // Don't try to start the server too many times.
fLaunchAttemptCount++;
// Make sure the server is running.
if (!be_roster->IsRunning (kServerSignature)) {
errorCode = be_roster->Launch (kServerSignature);
if (errorCode != B_OK) {
BPath path;
entry_ref ref;
directory_which places[] = {B_COMMON_BIN_DIRECTORY,B_BEOS_BIN_DIRECTORY};
for (int32 i = 0; i < 2; i++) {
find_directory(places[i],&path);
path.Append("spamdbm");
if (!BEntry(path.Path()).Exists())
continue;
get_ref_for_path(path.Path(),&ref);
if ((errorCode = be_roster->Launch (&ref)) == B_OK)
break;
}
if (errorCode != B_OK)
goto ErrorExit;
}
}
// Set up the messenger to the database server.
serverTeam = be_roster->TeamFor (kServerSignature);
if (serverTeam < 0)
goto ErrorExit;
fMessengerToServer =
BMessenger (kServerSignature, serverTeam, &errorCode);
if (!fMessengerToServer.IsValid ())
goto ErrorExit;
// Check if the server is running in headers only mode. If so, we only
// need to download the header rather than the entire message.
scriptingMessage.MakeEmpty ();
scriptingMessage.what = B_GET_PROPERTY;
scriptingMessage.AddSpecifier ("TokenizeMode");
replyMessage.MakeEmpty ();
if ((errorCode = fMessengerToServer.SendMessage (&scriptingMessage,
&replyMessage)) != B_OK)
goto ErrorExit;
if ((errorCode = replyMessage.FindInt32 ("error", &tempErrorCode))
!= B_OK)
goto ErrorExit;
if ((errorCode = tempErrorCode) != B_OK)
goto ErrorExit;
if ((errorCode = replyMessage.FindString ("result",
&tokenizeModeStringPntr)) != B_OK)
goto ErrorExit;
fHeaderOnly = (tokenizeModeStringPntr != NULL
&& strcmp (tokenizeModeStringPntr, "JustHeader") == 0);
}
// See if the message has already been classified. Happens for messages
// which are partially downloaded when you have auto-training on. Could
// untrain the partial part before training on the complete message, but we
// don't know how big it was, so instead just ignore the message.
if (nodeForOutputFileInitialised) {
if (nodeForOutputFile.GetAttrInfo ("MAIL:classification",
&attributeInfo) == B_OK)
return B_OK;
}
// Copy the message to a string so that we can pass it to the spam database
// (the even messier alternative is a temporary file). Do it in a fashion
// which allows NUL bytes in the string. This method of course limits the
// message size to a few hundred megabytes. If we're using header mode,
// only read the header rather than the full message.
if (fHeaderOnly) {
// Read just the header, it ends with an empty CRLF line.
dataStreamPntr->Seek (0, SEEK_SET);
while ((errorCode = dataStreamPntr->Read (&tempChar, 1)) == 1) {
headerString.Append (tempChar, 1);
headerLength = headerString.Length();
if (headerLength >= 4 && strcmp (headerString.String() +
headerLength - 4, "\r\n\r\n") == 0)
break;
}
if (errorCode < 0)
goto ErrorExit;
dataSize = headerString.Length();
stringBuffer = new char [dataSize + 1];
memcpy (stringBuffer, headerString.String(), dataSize);
stringBuffer[dataSize] = 0;
} else {
// Read the whole file. The seek to the end may take a while since
// that triggers downloading of the entire message (and caching in a
// slave file - see the MessageIO class).
dataSize = dataStreamPntr->Seek (0, SEEK_END);
if (dataSize <= 0)
goto ErrorExit;
try {
stringBuffer = new char [dataSize + 1];
} catch (...) {
errorCode = ENOMEM;
goto ErrorExit;
}
dataStreamPntr->Seek (0, SEEK_SET);
amountRead = dataStreamPntr->Read (stringBuffer, dataSize);
if (amountRead != dataSize)
goto ErrorExit;
stringBuffer[dataSize] = 0; // Add an end of string NUL, just in case.
}
// Send off a scripting command to the database server, asking it to
// evaluate the string for spaminess. Note that it can return ENOMSG
// when there are no words (a good indicator of spam which is pure HTML
// if you are using plain text only tokenization), so we could use that
// as a spam marker too. Code copied for the reevaluate stuff below.
scriptingMessage.MakeEmpty ();
scriptingMessage.what = B_SET_PROPERTY;
scriptingMessage.AddSpecifier ("EvaluateString");
errorCode = scriptingMessage.AddData ("data", B_STRING_TYPE,
stringBuffer, dataSize + 1, false /* fixed size */);
if (errorCode != B_OK)
goto ErrorExit;
replyMessage.MakeEmpty ();
errorCode = fMessengerToServer.SendMessage (&scriptingMessage,
&replyMessage);
if (errorCode != B_OK
|| replyMessage.FindInt32 ("error", &errorCode) != B_OK)
goto ErrorExit; // Unable to read the return code.
if (errorCode == ENOMSG && fNoWordsMeansSpam)
spamRatio = fSpamCutoffRatio; // Yes, no words and that means spam.
else if (errorCode != B_OK
|| replyMessage.FindFloat ("result", &spamRatio) != B_OK)
goto ErrorExit; // Classification failed in one of many ways.
// If we are auto-training, feed back the message to the server as a
// training example (don't train if it is uncertain). Also redo the
// evaluation after training.
if (fAutoTraining) {
if (spamRatio >= fSpamCutoffRatio || spamRatio < fGenuineCutoffRatio) {
scriptingMessage.MakeEmpty ();
scriptingMessage.what = B_SET_PROPERTY;
scriptingMessage.AddSpecifier ((spamRatio >= fSpamCutoffRatio)
? "SpamString" : "GenuineString");
errorCode = scriptingMessage.AddData ("data", B_STRING_TYPE,
stringBuffer, dataSize + 1, false /* fixed size */);
if (errorCode != B_OK)
goto ErrorExit;
replyMessage.MakeEmpty ();
errorCode = fMessengerToServer.SendMessage (&scriptingMessage,
&replyMessage);
if (errorCode != B_OK
|| replyMessage.FindInt32 ("error", &errorCode) != B_OK)
goto ErrorExit; // Unable to read the return code.
if (errorCode != B_OK)
goto ErrorExit; // Failed to set a good example.
}
// Note the kind of example made so that the user doesn't reclassify
// the message twice (the spam server looks for this attribute).
classificationString =
(spamRatio >= fSpamCutoffRatio)
? "Spam"
: ((spamRatio < fGenuineCutoffRatio) ? "Genuine" : "Uncertain");
if (nodeForOutputFileInitialised)
nodeForOutputFile.WriteAttr ("MAIL:classification", B_STRING_TYPE,
0 /* offset */, classificationString,
strlen (classificationString) + 1);
// Now that the database has changed due to training, recompute the
// spam ratio. Hopefully it will have become more extreme in the
// correct direction (not switched from being spam to being genuine).
// Code copied from above.
scriptingMessage.MakeEmpty ();
scriptingMessage.what = B_SET_PROPERTY;
scriptingMessage.AddSpecifier ("EvaluateString");
errorCode = scriptingMessage.AddData ("data", B_STRING_TYPE,
stringBuffer, dataSize + 1, false /* fixed size */);
if (errorCode != B_OK)
goto ErrorExit;
replyMessage.MakeEmpty ();
errorCode = fMessengerToServer.SendMessage (&scriptingMessage,
&replyMessage);
if (errorCode != B_OK
|| replyMessage.FindInt32 ("error", &errorCode) != B_OK)
goto ErrorExit; // Unable to read the return code.
if (errorCode == ENOMSG && fNoWordsMeansSpam)
spamRatio = fSpamCutoffRatio; // Yes, no words and that means spam.
else if (errorCode != B_OK
|| replyMessage.FindFloat ("result", &spamRatio) != B_OK)
goto ErrorExit; // Classification failed in one of many ways.
}
// Store the spam ratio in an attribute called MAIL:ratio_spam,
// attached to the eventual output file.
if (nodeForOutputFileInitialised)
nodeForOutputFile.WriteAttr ("MAIL:ratio_spam",
B_FLOAT_TYPE, 0 /* offset */, &spamRatio, sizeof (spamRatio));
// Also add it to the subject, if requested.
if (fAddSpamToSubject
&& spamRatio >= fSpamCutoffRatio
&& io_headers->FindString ("Subject", &oldSubjectStringPntr) == B_OK) {
newSubjectString.SetTo ("[Spam ");
sprintf (percentageString, "%05.2f", spamRatio * 100.0);
newSubjectString << percentageString << "%] ";
newSubjectString << oldSubjectStringPntr;
io_headers->ReplaceString ("Subject", newSubjectString);
}
// Beep using different sounds for spam and genuine, as Jeremy Friesner
// nudged me to get around to implementing. And add uncertain to that, as
// "BiPolar" suggested. If the user doesn't want to hear the sound, they
// can turn it off in the system sound preferences.
if (spamRatio >= fSpamCutoffRatio) {
system_beep (kAGMSBayesBeepSpamName);
} else if (spamRatio < fGenuineCutoffRatio) {
system_beep (kAGMSBayesBeepGenuineName);
} else {
system_beep (kAGMSBayesBeepUncertainName);
}
return B_OK;
ErrorExit:
fprintf (stderr, "Error exit from "
"SpamFilter::ProcessMailMessage, code maybe %ld (%s).\n",
errorCode, strerror (errorCode));
delete [] stringBuffer;
return B_OK; // Not MD_ERROR so the message doesn't get left on server.
}
