private void ParseSizes()
{
double glyphSpaceToTextSpace = 1000.0; // TODO: PdfFont.ParseSizes: SubType:Type3 Uses a FontMatrix that may not correspond to 1/1000th
_widths = new Dictionary <char, double>();
char firstChar = (char)_baseData.GetParamAsInt("FirstChar");
char lastChar = (char)_baseData.GetParamAsInt("LastChar");
PdfArray widths = _baseData.Values["Widths"] as PdfArray;
char actualChar = firstChar;
foreach (IPdfElement elem in widths.Values)
{
double width = PdfElementUtils.GetReal(elem, 500);
if (width < 0.0001f && width > -0.0001f)
{
width = 500;
}
_widths.Add(actualChar, width / glyphSpaceToTextSpace);
actualChar++;
}
// FIMXE: Calculate real height
}
private void OpTextPutMultiple(PdfArray array)
{
if (inText == false)
{
return;
}
foreach (IPdfElement elem in array.Values)
{
if (elem is PdfString)
{
OpTextPut(((PdfString)elem).Value);
}
else if (elem is PdfInteger || elem is PdfReal)
{
double spacing = PdfElementUtils.GetReal(elem, 0);
_textWidth -= (spacing / 1000) * _fontSize;
}
else if (elem is PdfArray)
{
OpTextPutMultiple(((PdfArray)elem));
}
}
}
public PdfObject ParseObject(List <PdfObject> knownObjects)
{
PdfObject obj = null;
long startPosition = _streamPosition;
do
{
SkipWhitespace();
byte character = PeekChar();
if (character == '%')
{
SkipComment();
}
else if (IsDigit(character))
{
IPdfElement objectID = ParseNumber();
SkipWhitespace();
IPdfElement objectGeneration = ParseNumber();
SkipWhitespace();
string token = ParseToken();
if (token == "obj")
{
SkipWhitespace();
IPdfElement element = ParseElement();
string endToken = ParseToken();
// Intercept streams
if (endToken == "stream")
{
PdfDictionary streamDict = element as PdfDictionary;
if (streamDict == null)
{
throw new Exception(string.Format("Stream after a not dictionary element at: {0}", _streamPosition));
}
SkipEndOfLine();
// Find the length of the stream
long length = -1;
if (streamDict.Values.ContainsKey("Length"))
{
length = PdfElementUtils.GetInt(streamDict.Values["Length"], -1);
if (length == -1 && streamDict.Values["Length"] is PdfObjectReference)
{
IPdfElement lenghtObj = SearchObjectID(knownObjects, ((PdfObjectReference)streamDict.Values["Length"]).ObjectID);
length = PdfElementUtils.GetInt(lenghtObj, -1);
}
}
if (length == -1)
{
byte lineFeed = 0x0A;
byte carriageReturn = 0x0D;
length = MeasureToMarkers(new char[][] {
new char[] { (char)carriageReturn, (char)lineFeed, 'e', 'n', 'd', 's', 't', 'r', 'e', 'a', 'm' },
new char[] { (char)lineFeed, 'e', 'n', 'd', 's', 't', 'r', 'e', 'a', 'm' },
new char[] { 'e', 'n', 'd', 's', 't', 'r', 'e', 'a', 'm', (char)lineFeed },
new char[] { 'e', 'n', 'd', 's', 't', 'r', 'e', 'a', 'm', (char)carriageReturn, (char)lineFeed },
});
}
// Get the stream
byte[] streamBody = GetRawData(length);
SkipEndOfLine();
endToken = ParseToken();
if (endToken != "endstream")
{
throw new Exception(string.Format("Expected \"endstream\" token, \"{0}\" found at: {1}", token, _streamPosition));
}
SkipWhitespace();
endToken = ParseToken();
PdfStream stream = new PdfStream();
stream.Dictionary = streamDict;
stream.Data = streamBody;
element = stream;
}
if (endToken == "endobj")
{
obj = new PdfObject();
obj.ObjectID = (int)((PdfInteger)objectID).Value;
obj.ObjectGeneration = (int)((PdfInteger)objectGeneration).Value;
obj.Data = element;
break;
}
}
}
else
{
long streamPosition = _streamPosition;
string token = ParseToken();
if (token == "startxref")
{
// TODO: PdfParser: Ignoring startxref for now
SkipEndOfLine();
SkipToEndOfLine();
SkipEndOfLine();
SkipToEndOfLine();
SkipEndOfLine();
SkipWhitespace();
continue;
}
if (token == "xref")
{
// TODO: PdfParser: Ignoring xref for now
SkipToEndOfLine();
SkipEndOfLine();
do
{
SkipWhitespace();
IPdfElement objNumber = ParseNumber();
SkipWhitespace();
objNumber = ParseNumber();
SkipEndOfLine();
PdfInteger refNumber = objNumber as PdfInteger;
for (int i = 0; i < refNumber.Value; i++)
{
SkipToEndOfLine();
SkipEndOfLine();
}
long currentPosition = _streamPosition;
IPdfElement testElem = ParseElement();
_streamPosition = currentPosition;
if ((testElem is PdfInteger) == false)
{
break;
}
} while (IsEndOfStream() == false);
continue;
}
if (token == "trailer")
{
// TODO: PdfParser: Ignoring trailer for now
SkipEndOfLine();
ParseElement();
SkipWhitespace();
SkipToEndOfLine();
SkipEndOfLine();
SkipToEndOfLine();
SkipEndOfLine();
SkipToEndOfLine();
SkipEndOfLine();
SkipWhitespace();
continue;
}
// Try to find an object marker
byte lineFeed = 0x0A;
byte carriageReturn = 0x0D;
long distToObject = MeasureToMarkers(new char[][] {
new char[] { ' ', 'o', 'b', 'j', (char)lineFeed },
new char[] { ' ', 'o', 'b', 'j', (char)carriageReturn, (char)lineFeed },
});
if (distToObject > 0)
{
// Object marker found, backtrack and retry
long originalPosition = _streamPosition;
_streamPosition += distToObject;
long marker = _streamPosition;
SkipWhitespaceBack();
if (_streamPosition == marker)
{
// Abort backtrack, skip garbage
_streamPosition = originalPosition + distToObject + 4;
continue;
}
marker = _streamPosition;
SkipDigitsBack();
if (_streamPosition == marker)
{
// Abort backtrack, skip garbage
_streamPosition = originalPosition + distToObject + 4;
continue;
}
marker = _streamPosition;
SkipWhitespaceBack();
if (_streamPosition == marker)
{
// Abort backtrack, skip garbage
_streamPosition = originalPosition + distToObject + 4;
continue;
}
marker = _streamPosition;
SkipDigitsBack();
if (_streamPosition == marker)
{
// Abort backtrack, skip garbage
_streamPosition = originalPosition + distToObject + 4;
continue;
}
NextChar();
}
else
{
// No more obj markers found, abort all.
_streamPosition = _stream.Length;
}
}
} while (IsEndOfStream() == false);
return(obj);
}
private void ProcessPageContent()
{
int unknowCount = 0;
int lineCount = 0;
int strokeCount = 0;
int pathCount = 0;
for (int i = 0; i < _page.ContentActions.Count; i++)
{
PdfContentAction action = _page.ContentActions[i];
// Special graphics state
if (action.Token == "q")
{
OpPushGraphState();
}
else if (action.Token == "Q")
{
OpPopGraphState();
}
else if (action.Token == "cm")
{
double a = PdfElementUtils.GetReal(action.Parameters[0], 0);
double b = PdfElementUtils.GetReal(action.Parameters[1], 0);
double c = PdfElementUtils.GetReal(action.Parameters[2], 0);
double d = PdfElementUtils.GetReal(action.Parameters[3], 0);
double e = PdfElementUtils.GetReal(action.Parameters[4], 0);
double f = PdfElementUtils.GetReal(action.Parameters[5], 0);
OpSetGraphMatrix(a, b, c, d, e, f);
}
// Text Operations
else if (action.Token == "BT")
{
OpBeginText();
}
else if (action.Token == "ET")
{
OpEndText();
}
else if (action.Token == "Tc")
{
double charSpacing = PdfElementUtils.GetReal(action.Parameters[0], 0);
OpTextCharSpacing(charSpacing);
}
else if (action.Token == "Tw")
{
double wordSpacing = PdfElementUtils.GetReal(action.Parameters[0], 0);
OpTextWordSpacing(wordSpacing);
}
else if (action.Token == "Tz")
{
// TODO: PdfTextExtractor: Horizontal Scale
}
else if (action.Token == "Tf")
{
string fontName = PdfElementUtils.GetString(action.Parameters[0], string.Empty);
double fontSize = PdfElementUtils.GetReal(action.Parameters[1], 0);
OpTextFont(fontName, fontSize);
}
else if (action.Token == "TL")
{
double leading = PdfElementUtils.GetReal(action.Parameters[0], 0);
OpTextLeading(leading);
}
else if (action.Token == "Tr")
{
// TODO: PdfTextExtractor: Rendering mode
}
else if (action.Token == "Ts")
{
// TODO: PdfTextExtractor: Text rise
}
else if (action.Token == "Td")
{
double x = PdfElementUtils.GetReal(action.Parameters[0], 0);
double y = PdfElementUtils.GetReal(action.Parameters[1], 0);
OpTextDisplace(x, y);
}
else if (action.Token == "TD")
{
double x = PdfElementUtils.GetReal(action.Parameters[0], 0);
double y = PdfElementUtils.GetReal(action.Parameters[1], 0);
OpTextLeading(-y);
OpTextDisplace(x, y);
}
else if (action.Token == "Tm")
{
double a = PdfElementUtils.GetReal(action.Parameters[0], 0);
double b = PdfElementUtils.GetReal(action.Parameters[1], 0);
double c = PdfElementUtils.GetReal(action.Parameters[2], 0);
double d = PdfElementUtils.GetReal(action.Parameters[3], 0);
double e = PdfElementUtils.GetReal(action.Parameters[4], 0);
double f = PdfElementUtils.GetReal(action.Parameters[5], 0);
OpSetTextMatrix(a, b, c, d, e, f);
}
else if (action.Token == "T*")
{
OpTextLineFeed();
}
else if (action.Token == "Tj")
{
string text = PdfElementUtils.GetString(action.Parameters[0], string.Empty);
OpTextPut(text);
}
else if (action.Token == "'")
{
string text = PdfElementUtils.GetString(action.Parameters[0], string.Empty);
OpTextLineFeed();
OpTextPut(text);
}
else if (action.Token == "\"")
{
double wordSpacing = PdfElementUtils.GetReal(action.Parameters[0], 0);
double charSpacing = PdfElementUtils.GetReal(action.Parameters[1], 0);
string text = PdfElementUtils.GetString(action.Parameters[0], string.Empty);
OpTextCharSpacing(charSpacing);
OpTextWordSpacing(wordSpacing);
OpTextPut(text);
}
else if (action.Token == "TJ")
{
OpTextPutMultiple(((PdfArray)action.Parameters[0]));
}
else if (action.Token == "re")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "f")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "g")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "rg")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "BI")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "ID")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "EI")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "W")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "n")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "Do")
{
// TODO: PdfTextExtractor: Interpret this
}
else if (action.Token == "m")
{
// TODO: PdfTextExtractor: Interpret this "moveto: Begin new subpath"
}
else if (action.Token == "l")
{
// TODO: PdfTextExtractor: Interpret this "lineto: Append straight line segment to path"
lineCount++;
}
else if (action.Token == "h")
{
// TODO: PdfTextExtractor: Interpret this "closepath: Close subpath"
pathCount++;
}
else if (action.Token == "W")
{
// TODO: PdfTextExtractor: Interpret this "clip: Set clipping path using nonzero winding number rule"
}
else if (action.Token == "W*")
{
// TODO: PdfTextExtractor: Interpret this "eoclip: Set clipping path using even-odd rule"
}
else if (action.Token == "w")
{
// TODO: PdfTextExtractor: Interpret this "setlinewidth: Set line width"
}
else if (action.Token == "G")
{
// TODO: PdfTextExtractor: Interpret this "setgray: Set gray level for stroking operations"
}
else if (action.Token == "S")
{
// TODO: PdfTextExtractor: Interpret this "stroke: Stroke path"
strokeCount++;
}
else if (action.Token == "M")
{
// TODO: PdfTextExtractor: Interpret this "setmiterlimit: Set miter limit"
}
else
{
unknowCount++;
}
}
FlushTextElement();
}
private void btnProcess_Click(object sender, EventArgs e)
{
if (System.IO.File.Exists(txtPdfPath.Text) == false)
{
MessageBox.Show("File does not exist");
return;
}
PdfDocument doc = PdfDocument.Load(txtPdfPath.Text);
int nObjects = doc.Objects.Count;
int nRootObject = doc.Objects.Where(obj => obj.UsageCount == 0).Count();
List <PdfStream> streams = doc.Objects
.Where(obj => obj.Data.Type == PdfElementTypes.Stream)
.Select(obj => (PdfStream)obj.Data)
.ToList();
int nStreams = streams.Count;
int nPages = doc.Pages.Count;
List <string> lines = new List <string>();
lines.Add(string.Format("Filename : {0}", System.IO.Path.GetFileNameWithoutExtension(txtPdfPath.Text)));
lines.Add(string.Format("Number of Objects : {0}", nObjects));
lines.Add(string.Format("Number of Roots : {0}", nRootObject));
lines.Add(string.Format("Number of Streams : {0}", nStreams));
lines.Add(string.Format("Number of Pages : {0}", nPages));
int pageNumber = 1;
foreach (PdfDocumentPage page in doc.Pages)
{
lines.Add("-----------------------------------------------------------------------------------------");
if (page.BaseData.Values.ContainsKey("CropBox"))
{
PdfArray cropBox = page.BaseData.Values["CropBox"] as PdfArray;
lines.Add(string.Format("Page({0} of {1}): {2} {3} {4} {5}", pageNumber, doc.Pages.Count,
PdfElementUtils.GetReal(cropBox.Values[0], 0),
PdfElementUtils.GetReal(cropBox.Values[1], 0),
PdfElementUtils.GetReal(cropBox.Values[2], 0),
PdfElementUtils.GetReal(cropBox.Values[3], 0)));
}
else
{
lines.Add(string.Format("Page({0} of {1}): ", pageNumber, doc.Pages.Count));
}
pageNumber++;
PdfTextExtractor extractor = new PdfTextExtractor(page);
foreach (PdfTextElement textElement in extractor.Elements)
{
string fontName = textElement.Font == null ? "#NULL#" : textElement.Font.Name;
if (fontName == "#NULL#" && textElement.Childs.Count > 0)
{
var fontNames = textElement.Childs.Select(c => c.Font == null ? "#NULL#" : c.Font.Name);
StringBuilder sbFontName = new StringBuilder();
foreach (string fontNameAux in fontNames)
{
if (sbFontName.Length > 0)
{
sbFontName.Append(";");
}
sbFontName.Append(fontNameAux);
}
fontName = sbFontName.ToString();
}
lines.Add(string.Format("Text({0}, {1})({2}, {3})[{4}]: \"{5}\"",
Math.Round(textElement.Matrix.Matrix[0, 2], 2),
Math.Round(textElement.Matrix.Matrix[1, 2], 2),
Math.Round(textElement.VisibleWidth, 2),
Math.Round(textElement.VisibleHeight, 2),
fontName,
textElement.VisibleText));
}
}
txtOutput.Lines = lines.ToArray();
}
