/// <summary> /// Parses the header of the next section of the multipart stream and /// determines if it contains file data or parameter data. /// </summary> /// <param name="reader"> /// The StreamReader to read data from. /// </param> /// <exception cref="MultipartParseException"> /// thrown if unexpected data is hit such as end of stream. /// </exception> private void ParseSection(RebufferableBinaryReader reader) { // Our first job is to determine what type of section this is: form data or file. // This is a bit tricky because files can still be encoded with Content-Disposition: form-data // in the case of single file uploads. Multi-file uploads have Content-Disposition: file according // to the spec however in practise it seems that multiple files will be represented by // multiple Content-Disposition: form-data files. var parameters = new Dictionary <string, string>(); string line = reader.ReadLine(); while (line != string.Empty) { if (line == null) { throw new MultipartParseException("Unexpected end of stream"); } if (line == this.boundary || line == this.endBoundary) { throw new MultipartParseException("Unexpected end of section"); } // This line parses the header values into a set of key/value pairs. For example: // Content-Disposition: form-data; name="textdata" // ["content-disposition"] = "form-data" // ["name"] = "textdata" // Content-Disposition: form-data; name="file"; filename="data.txt" // ["content-disposition"] = "form-data" // ["name"] = "file" // ["filename"] = "data.txt" // Content-Type: text/plain // ["content-type"] = "text/plain" var values = SplitBySemicolonIgnoringSemicolonsInQuotes(line) .Select(x => x.Split(new[] { ':', '=' }, 2)) // Limit split to 2 splits so we don't accidently split characters in file paths. .ToDictionary( x => x[0].Trim().Replace("\"", string.Empty).ToLower(), x => x[1].Trim().Replace("\"", string.Empty)); // Here we just want to push all the values that we just retrieved into the // parameters dictionary. try { foreach (var pair in values) { parameters.Add(pair.Key, pair.Value); } } catch (ArgumentException) { throw new MultipartParseException("Duplicate field in section"); } line = reader.ReadLine(); } // Now that we've consumed all the parameters we're up to the body. We're going to do // different things depending on if we're parsing a, relatively small, form value or a // potentially large file. if (parameters.ContainsKey("filename")) { // Right now we assume that if a section contains filename then it is a file. // This assumption needs to be checked, it holds true in firefox but is untested for other // browsers. FilePart part = this.ParseFilePart(parameters, reader); this.Files.Add(part.Name, part); } else { ParameterPart part = this.ParseParameterPart(parameters, reader); this.Parameters.Add(part.Name, part); } }
/// <summary> /// Parses a section of the stream that is known to be file data. /// </summary> /// <param name="parameters"> /// The header parameters of this file, expects "name" and "filename" to be valid keys /// </param> /// <param name="reader"> /// The StreamReader to read the data from /// </param> /// <returns> /// The <see cref="FilePart"/> containing the parsed data (name, filename, stream containing file). /// </returns> private FilePart ParseFilePart(Dictionary <string, string> parameters, RebufferableBinaryReader reader) { // We want to create a stream and fill it with the data from the // file. var data = new MemoryStream(); var curBuffer = new byte[this.BinaryBufferSize]; var prevBuffer = new byte[this.BinaryBufferSize]; int curLength = 0; int prevLength = 0; prevLength = reader.Read(prevBuffer, 0, prevBuffer.Length); do { curLength = reader.Read(curBuffer, 0, curBuffer.Length); // Combine both buffers into the fullBuffer // See: http://stackoverflow.com/questions/415291/best-way-to-combine-two-or-more-byte-arrays-in-c-sharp var fullBuffer = new byte[this.BinaryBufferSize * 2]; Buffer.BlockCopy(prevBuffer, 0, fullBuffer, 0, prevLength); Buffer.BlockCopy(curBuffer, 0, fullBuffer, prevLength, curLength); // Now we want to check for a substring within the current buffer. // We need to find the closest substring greedily. That is find the // closest boundary and don't miss the end --'s if it's an end boundary. int endBoundaryPos = SubsequenceFinder.Search(fullBuffer, this.endBoundaryBinary); int endBoundaryLength = this.endBoundaryBinary.Length; int boundaryPos = SubsequenceFinder.Search(fullBuffer, this.boundaryBinary); int boundaryLength = this.boundaryBinary.Length; // We need to select the appropriate position and length // based on the smallest non-negative position. int endPos = -1; int endPosLength = 0; if (endBoundaryPos >= 0 && boundaryPos >= 0) { if (boundaryPos < endBoundaryPos) { // Select boundary endPos = boundaryPos; endPosLength = boundaryLength; } else { // Select end boundary endPos = endBoundaryPos; endPosLength = endBoundaryLength; this.readEndBoundary = true; } } else if (boundaryPos >= 0 && endBoundaryPos < 0) { // Select boundary endPos = boundaryPos; endPosLength = boundaryLength; } else if (boundaryPos < 0 && endBoundaryPos >= 0) { // Select end boundary endPos = endBoundaryPos; endPosLength = endBoundaryLength; this.readEndBoundary = true; } if (endPos != -1) { // Now we need to check if the endPos is followed by \r\n or just \n. HTTP // specifies \r\n but some clients might encode with \n. Or we might get 0 if // we are at the end of the file. int boundaryNewlineOffset = this.CalculateNewlineLength(ref fullBuffer, endPos + endPosLength); // We also need to check if the last n characters of the buffer to write // are a newline and if they are ignore them. var maxNewlineBytes = Encoding.GetMaxByteCount(2); int bufferNewlineOffset = this.FindNextNewline( ref fullBuffer, Math.Max(0, endPos - maxNewlineBytes), maxNewlineBytes); int bufferNewlineLength = this.CalculateNewlineLength(ref fullBuffer, bufferNewlineOffset); // We've found an end. We need to consume all the binary up to it // and then write the remainder back to the original stream. Then we // need to modify the original streams position to take into account // the new data. // We also want to chop off the newline that is inserted by the protocl. // We can do this by reducing endPos by the length of newline in this environment // and encoding data.Write(fullBuffer, 0, endPos - bufferNewlineLength); int writeBackOffset = endPos + endPosLength + boundaryNewlineOffset; int writeBackAmount = (prevLength + curLength) - writeBackOffset; var writeBackBuffer = new byte[writeBackAmount]; Buffer.BlockCopy(fullBuffer, writeBackOffset, writeBackBuffer, 0, writeBackAmount); reader.Buffer(writeBackBuffer); // stream.Write(fullBuffer, writeBackOffset, writeBackAmount); // stream.Position = stream.Position - writeBackAmount; // stream.Flush(); data.Position = 0; data.Flush(); break; } // No end, consume the entire previous buffer data.Write(prevBuffer, 0, prevLength); data.Flush(); // Now we want to swap the two buffers, we don't care // what happens to the data from prevBuffer so we set // curBuffer to it so it gets overwrited. byte[] tempBuffer = curBuffer; curBuffer = prevBuffer; prevBuffer = tempBuffer; // We don't need to swap the lengths because // curLength will be overwritten in the next // iteration of the loop. prevLength = curLength; }while (prevLength != 0); var contentType = parameters.ContainsKey("content-type") ? parameters["content-type"] : "text/plain"; var contentDisposition = parameters.ContainsKey("content-disposition") ? parameters["content-disposition"] : "form-data"; var part = new FilePart(parameters["name"], parameters["filename"], data.ToArray(), contentType, contentDisposition); return(part); }