public static IEnumerable <WarcCdxItemRaw> Read(Stream cdxStream, bool gzipped)
{
List <Action <WarcCdxItemRaw, Utf8String> > fieldSetters = null;
using (cdxStream)
{
while (true)
{
var textStream = gzipped ? new GZipStream(cdxStream, CompressionMode.Decompress, true) : cdxStream;
using (var reader = new Utf8StreamReader(textStream, true))
{
if (fieldSetters == null)
{
var fields = reader.ReadLine().Split((byte)' ');
InitializeSetters();
fieldSetters = new List <Action <WarcCdxItemRaw, Utf8String> >();
var foundCdx = false;
foreach (var label in fields)
{
if (!foundCdx)
{
if (label == CDX)
{
foundCdx = true;
}
}
else
{
allSetters.TryGetValue(label, out var setter);
fieldSetters.Add(setter);
}
}
}
Utf8String[] arr = null;
while (!reader.IsCompleted)
{
var line = reader.ReadLine();
if (line.Length == 0)
{
continue;
}
line.Split((byte)' ', StringSplitOptions.None, ref arr);
var item = new WarcCdxItemRaw();
for (int i = 0; i < fieldSetters.Count; i++)
{
var value = arr[i];
if (value.Length > 0 & !(value.Length == 1 && value[0] == (byte)'-'))
{
fieldSetters[i]?.Invoke(item, value);
}
}
yield return(item);
//action(item);
}
}
if (textStream == cdxStream)
{
break;
}
var remaining = GetRemainingUnusedBytes((GZipStream)textStream);
textStream.Dispose();
if (remaining == 0)
{
break;
}
cdxStream.Position -= remaining;
}
}
}
public static void GenerateCdx(string cdx, IEnumerable <string> warcs)
{
var scratchpad = new Scratchpad();
var buf = new byte[16 * 1024];
using (var output = File.Open(cdx + ".tmp", FileMode.Create, FileAccess.Write, FileShare.Delete | FileShare.Read))
{
using (var writer = new Utf8StreamWriter(output))
{
writer.WriteClrStringLine(WarcColumns);
foreach (var warc in warcs)
{
Console.WriteLine(Path.GetFileName(warc));
using (var warcStream = File.Open(warc, FileMode.Open, FileAccess.Read, FileShare.Delete | FileShare.Read))
{
try
{
var warcname = new Utf8String(Path.GetFileName(warc));
while (warcStream.Position != warcStream.Length)
{
var startPosition = warcStream.Position;
long warcContentLength = -1;
long payloadLength = -1;
int responseCode = -1;
var contentType = Utf8String.Empty;
var date = scratchpad.Use(14);
date[0] = 0;
Utf8String url = Utf8String.Empty;
Utf8String shamanUrl = Utf8String.Empty;
DateTime? lastModified = null;
bool isresponse = false;
using (var gz = new GZipStream(warcStream, CompressionMode.Decompress, true))
{
using (var reader = new Utf8StreamReader(gz, true))
{
while (true)
{
if (reader.IsCompleted)
{
throw new EndOfStreamException();
}
var line = reader.ReadLine();
if (line.Length == 0)
{
break;
}
if (line.Equals(Warc_Response))
{
isresponse = true;
}
if (line.StartsWith(Warc_ContentLength))
{
warcContentLength = Utf8Utils.ParseInt64(WarcItem.GetHeaderValue(line));
}
else if (line.StartsWith(Warc_Date))
{
var val = WarcItem.GetHeaderValue(line).Bytes;
val.Slice(0, 4).CopyTo(date.Slice(0));
val.Slice(5, 2).CopyTo(date.Slice(4));
val.Slice(8, 2).CopyTo(date.Slice(6));
val.Slice(11, 2).CopyTo(date.Slice(8));
val.Slice(14, 2).CopyTo(date.Slice(10));
val.Slice(17, 2).CopyTo(date.Slice(12));
}
else if (line.StartsWith(Warc_URL))
{
url = scratchpad.Copy(WarcItem.GetHeaderValue(line));
}
if (line.StartsWith(Warc_Shaman_URI))
{
shamanUrl = scratchpad.Copy(WarcItem.GetHeaderValue(line));
}
}
if (warcContentLength == -1)
{
throw new InvalidOperationException();
}
if (isresponse)
{
using (var s = WarcItem.OpenHttp(reader, scratchpad, new Uri(url.ToString()), warcContentLength, out var payloadLengthFromHeader, out var redirectLocation, out responseCode, out contentType, out lastModified, null))
{
long l = 0;
while (true)
{
var m = s.Read(buf, 0, buf.Length);
if (m == 0)
{
break;
}
l += m;
}
payloadLength = l;
if (payloadLengthFromHeader != -1 && payloadLengthFromHeader != payloadLength)
{
throw new Exception("Content-Length mismatch.");
}
}
//var httpData = new LimitedStream(reader, contentLength);
var cr = reader.ReadByte();
if (cr != 13)
{
throw new InvalidDataException();
}
var lf = reader.ReadByte();
if (lf != 10)
{
throw new InvalidDataException();
}
cr = reader.ReadByte();
if (cr != 13)
{
throw new InvalidDataException();
}
lf = reader.ReadByte();
if (lf != 10)
{
throw new InvalidDataException();
}
//if (reader.ReadByte() != 13) throw new Exception();
//if (reader.ReadByte() != 10) throw new Exception();
}
else
{
var remaining = warcContentLength;
while (remaining != 0)
{
var m = reader.Read((int)Math.Min(remaining, int.MaxValue));
if (m.Count == 0)
{
throw new Exception();
}
remaining -= m.Count;
}
var e = reader.ReadLine();
if (e.Length != 0)
{
throw new InvalidDataException();
}
e = reader.ReadLine();
if (e.Length != 0)
{
throw new InvalidDataException();
}
e = reader.ReadLine();
if (!reader.IsCompleted)
{
throw new InvalidDataException();
}
}
//var r = reader.RemainingBufferedData;
var end = reader.ReadByte();
if (end != -1)
{
throw new InvalidDataException();
}
//Console.WriteLine($"Remaining: {r.Length}");
}
warcStream.Position -= GetRemainingUnusedBytes(gz);
}
if (isresponse)
{
if (shamanUrl.Length > 0)
{
writer.Write(shamanUrl);
}
else
{
writer.Write(url);
}
writer.Write((byte)' ');
writer.Write(startPosition);
writer.Write((byte)' ');
writer.Write(warcStream.Position - startPosition);
writer.Write((byte)' ');
if (date[0] != 0)
{
writer.Write(date);
}
else
{
writer.Write((byte)'-');
}
writer.Write((byte)' ');
writer.Write(warcname);
writer.Write((byte)' ');
if (responseCode != -1)
{
writer.Write(responseCode);
}
else
{
writer.Write((byte)'-');
}
writer.Write((byte)' ');
writer.Write(contentType);
writer.Write((byte)' ');
writer.Write(payloadLength);
writer.Write((byte)' ');
if (lastModified != null)
{
WriteDate(writer, lastModified.Value);
}
else
{
writer.Write((byte)'-');
}
writer.Write((byte)' ');
writer.WriteLine();
}
scratchpad.Reset();
}
}
catch
{
if (warcStream.Position == warcStream.Length)
{
Console.WriteLine("WARNING: truncated WARC.");;
}
else
{
throw;
}
}
}
}
}
}
File.Delete(cdx);
File.Move(cdx + ".tmp", cdx);
}
public static Stream OpenHttp(Utf8StreamReader httpReader, Scratchpad scratchpad, Uri requestedUrl, long responseLength, out long payloadLength, out Uri location, out int responseCode, out Utf8String contentType, out DateTime?lastModified, Action <Utf8String, Utf8String> onHttpHeader)
{
var startPosition = httpReader.Position;
payloadLength = -1;
location = null;
lastModified = null;
bool chunked = false;
bool gzipped = false;
bool brotli = false;
var responseLine = httpReader.ReadLine();
responseCode = (int)Utf8Utils.ParseInt64(responseLine.TryCaptureBetween((byte)' ', (byte)' ') ?? responseLine.CaptureAfter((byte)' '));
while (true)
{
var line = httpReader.ReadLine();
if (httpReader.IsCompleted)
{
throw new InvalidDataException();
}
if (line.Length == 0)
{
break;
}
if (onHttpHeader != null)
{
var d = line.IndexOf((byte)':');
onHttpHeader(line.Substring(0, d).Trim(), line.Substring(d + 1).Trim());
}
if (line.StartsWith(Http_TransferEncoding))
{
var value = GetHeaderValue(line);
if (value.Equals("chunked"))
{
chunked = true;
}
}
else if (line.StartsWith(Http_ContentLength))
{
payloadLength = Utf8Utils.ParseInt64(GetHeaderValue(line));
}
else if (line.StartsWith(Http_ContentEncoding))
{
var value = GetHeaderValue(line);
if (value == Http_Gzip)
{
gzipped = true;
}
else if (value == Http_Brotli)
{
brotli = true;
}
}
else if (line.StartsWith(Http_Location))
{
var val = GetHeaderValue(line).ToString();
try
{
if (val.StartsWith("//"))
{
location = new Uri(requestedUrl.Scheme + ":" + val);
}
else
{
location = new Uri(requestedUrl, val);
}
}
catch (Exception ex)
{
}
}
else if (line.StartsWith(Http_ContentType) && scratchpad != null)
{
var value = GetHeaderValue(line);
value = value.TryCaptureBefore((byte)' ') ?? value;
value = value.TryCaptureBefore((byte)';') ?? value;
contentType = scratchpad.Copy(value);
}
else if (line.StartsWith(Http_LastModified))
{
try
{
lastModified = WarcCdxItemRaw.ParseHttpDate(GetHeaderValue(line));
}
catch { }
}
}
var compressed = gzipped || brotli;
if (compressed || chunked)
{
payloadLength = -1;
}
Stream s;
if (responseLength != -1)
{
var currentPos = httpReader.Position - startPosition;
var httpBodyLength = responseLength - currentPos;
if (!compressed && !chunked && payloadLength != -1 && httpBodyLength != payloadLength)
{
throw new Exception("Unexpected Content-Length.");
}
s = new LimitedStream(httpReader, httpBodyLength);
}
else
{
s = httpReader;
}
if (chunked)
{
s = new ChunkedStream(s);
}
if (compressed && chunked)
{
s = new OnDisposeConsumeStream(s);
}
if (gzipped)
{
s = new GZipStream(s, CompressionMode.Decompress);
}
else if (brotli)
{
s = new BrotliStream(s, CompressionMode.Decompress);
}
return(s);
}