没错，问题在于这不仅仅是protobuf格式 - 它是一种混合文件格式（在此定义），其中包含了多种格式，其中包括protobuf。它还包括压缩（虽然看起来是可选的）。
我已经从规范中分离出了我能够的部分，并且我有一个使用protobuf-net处理块的C#读取器 - 它可以愉快地读取到文件末尾 - 我可以告诉你有4515个块（BlockHeader）。当它到达Blob时，我对规范如何标记OSMHeader和OSMData有点困惑 - 我乐意听取建议！我还使用ZLIB.NET来处理正在使用的zlib压缩。在没有理解这一点的情况下，我已经处理了ZLIB数据并根据所声称的大小验证了它，以检查它至少是合理的。
如果您能找出（或询问作者）他们如何区分OSMHeader和OSMData，我将很乐意再做些其他的事情。希望你不介意我就到这里了 - 但已经过去几个小时了 ;p
using System;
using System.IO;
using OpenStreetMap; // where my .proto-generated entities are living
using ProtoBuf; // protobuf-net
using zlib; // ZLIB.NET
class OpenStreetMapParser
{
static void Main()
{
using (var file = File.OpenRead("us-northeast.osm.pbf"))
{
// from http://wiki.openstreetmap.org/wiki/ProtocolBufBinary:
//A file contains a header followed by a sequence of fileblocks. The design is intended to allow future random-access to the contents of the file and skipping past not-understood or unwanted data.
//The format is a repeating sequence of:
//int4: length of the BlockHeader message in network byte order
//serialized BlockHeader message
//serialized Blob message (size is given in the header)
int length, blockCount = 0;
while (Serializer.TryReadLengthPrefix(file, PrefixStyle.Fixed32, out length))
{
// I'm just being lazy and re-using something "close enough" here
// note that v2 has a big-endian option, but Fixed32 assumes little-endian - we
// actually need the other way around (network byte order):
uint len = (uint)length;
len = ((len & 0xFF) << 24) | ((len & 0xFF00) << 8) | ((len & 0xFF0000) >> 8) | ((len & 0xFF000000) >> 24);
length = (int)len;
BlockHeader header;
// again, v2 has capped-streams built in, but I'm deliberately
// limiting myself to v1 features
using (var tmp = new LimitedStream(file, length))
{
header = Serializer.Deserialize<BlockHeader>(tmp);
}
Blob blob;
using (var tmp = new LimitedStream(file, header.datasize))
{
blob = Serializer.Deserialize<Blob>(tmp);
}
if(blob.zlib_data == null) throw new NotSupportedException("I'm only handling zlib here!");
using(var ms = new MemoryStream(blob.zlib_data))
using(var zlib = new ZLibStream(ms))
{ // at this point I'm very unclear how the OSMHeader and OSMData are packed - it isn't clear
// read this to the end, to check we can parse the zlib
int payloadLen = 0;
while (zlib.ReadByte() >= 0) payloadLen++;
if (payloadLen != blob.raw_size) throw new FormatException("Screwed that up...");
}
blockCount++;
Console.WriteLine("Read block " + blockCount.ToString());
}
Console.WriteLine("all done");
Console.ReadLine();
}
}
}
abstract class InputStream : Stream
{
protected abstract int ReadNextBlock(byte[] buffer, int offset, int count);
public sealed override int Read(byte[] buffer, int offset, int count)
{
int bytesRead, totalRead = 0;
while (count > 0 && (bytesRead = ReadNextBlock(buffer, offset, count)) > 0)
{
count -= bytesRead;
offset += bytesRead;
totalRead += bytesRead;
pos += bytesRead;
}
return totalRead;
}
long pos;
public override void Write(byte[] buffer, int offset, int count)
{
throw new NotImplementedException();
}
public override void SetLength(long value)
{
throw new NotImplementedException();
}
public override long Position
{
get
{
return pos;
}
set
{
if (pos != value) throw new NotImplementedException();
}
}
public override long Length
{
get { throw new NotImplementedException(); }
}
public override void Flush()
{
throw new NotImplementedException();
}
public override bool CanWrite
{
get { return false; }
}
public override bool CanRead
{
get { return true; }
}
public override bool CanSeek
{
get { return false; }
}
public override long Seek(long offset, SeekOrigin origin)
{
throw new NotImplementedException();
}
}
class ZLibStream : InputStream
{ // uses ZLIB.NET: http://www.componentace.com/download/download.php?editionid=25
private ZInputStream reader; // seriously, why isn't this a stream?
public ZLibStream(Stream stream)
{
reader = new ZInputStream(stream);
}
public override void Close()
{
reader.Close();
base.Close();
}
protected override int ReadNextBlock(byte[] buffer, int offset, int count)
{
// OMG! reader.Read is the base-stream, reader.read is decompressed! yeuch
return reader.read(buffer, offset, count);
}
}
// deliberately doesn't dispose the base-stream
class LimitedStream : InputStream
{
private Stream stream;
private long remaining;
public LimitedStream(Stream stream, long length)
{
if (length < 0) throw new ArgumentOutOfRangeException("length");
if (stream == null) throw new ArgumentNullException("stream");
if (!stream.CanRead) throw new ArgumentException("stream");
this.stream = stream;
this.remaining = length;
}
protected override int ReadNextBlock(byte[] buffer, int offset, int count)
{
if(count > remaining) count = (int)remaining;
int bytesRead = stream.Read(buffer, offset, count);
if (bytesRead > 0) remaining -= bytesRead;
return bytesRead;
}
}
OpenStreetMap。 - René Nyffenegger在 Mark 设定的大纲之后，我通过查看 http://git.openstreetmap.nl/index.cgi/pbf2osm.git/tree/src/main.c?h=35116112eb0066c7729a963b292faa608ddc8ad7 找到了最后一部分。
这是最终代码。
using System;
using System.Diagnostics;
using System.IO;
using crosby.binary;
using OSMPBF;
using PerlLLC.Tools;
using ProtoBuf;
using zlib;
namespace OpenStreetMapOperations
{
class OpenStreetMapParser
{
static void Main()
{
using (var file = File.OpenRead(StaticTools.AssemblyDirectory + @"\us-pacific.osm.pbf"))
{
// from http://wiki.openstreetmap.org/wiki/ProtocolBufBinary:
//A file contains a header followed by a sequence of fileblocks. The design is intended to allow future random-access to the contents of the file and skipping past not-understood or unwanted data.
//The format is a repeating sequence of:
//int4: length of the BlockHeader message in network byte order
//serialized BlockHeader message
//serialized Blob message (size is given in the header)
int length, blockCount = 0;
while (Serializer.TryReadLengthPrefix(file, PrefixStyle.Fixed32, out length))
{
// I'm just being lazy and re-using something "close enough" here
// note that v2 has a big-endian option, but Fixed32 assumes little-endian - we
// actually need the other way around (network byte order):
length = IntLittleEndianToBigEndian((uint)length);
BlockHeader header;
// again, v2 has capped-streams built in, but I'm deliberately
// limiting myself to v1 features
using (var tmp = new LimitedStream(file, length))
{
header = Serializer.Deserialize<BlockHeader>(tmp);
}
Blob blob;
using (var tmp = new LimitedStream(file, header.datasize))
{
blob = Serializer.Deserialize<Blob>(tmp);
}
if (blob.zlib_data == null) throw new NotSupportedException("I'm only handling zlib here!");
HeaderBlock headerBlock;
PrimitiveBlock primitiveBlock;
using (var ms = new MemoryStream(blob.zlib_data))
using (var zlib = new ZLibStream(ms))
{
if (header.type == "OSMHeader")
headerBlock = Serializer.Deserialize<HeaderBlock>(zlib);
if (header.type == "OSMData")
primitiveBlock = Serializer.Deserialize<PrimitiveBlock>(zlib);
}
blockCount++;
Trace.WriteLine("Read block " + blockCount.ToString());
}
Trace.WriteLine("all done");
}
}
// 4-byte number
static int IntLittleEndianToBigEndian(uint i)
{
return (int)(((i & 0xff) << 24) + ((i & 0xff00) << 8) + ((i & 0xff0000) >> 8) + ((i >> 24) & 0xff));
}
}
abstract class InputStream : Stream
{
protected abstract int ReadNextBlock(byte[] buffer, int offset, int count);
public sealed override int Read(byte[] buffer, int offset, int count)
{
int bytesRead, totalRead = 0;
while (count > 0 && (bytesRead = ReadNextBlock(buffer, offset, count)) > 0)
{
count -= bytesRead;
offset += bytesRead;
totalRead += bytesRead;
pos += bytesRead;
}
return totalRead;
}
long pos;
public override void Write(byte[] buffer, int offset, int count)
{
throw new NotImplementedException();
}
public override void SetLength(long value)
{
throw new NotImplementedException();
}
public override long Position
{
get
{
return pos;
}
set
{
if (pos != value) throw new NotImplementedException();
}
}
public override long Length
{
get { throw new NotImplementedException(); }
}
public override void Flush()
{
throw new NotImplementedException();
}
public override bool CanWrite
{
get { return false; }
}
public override bool CanRead
{
get { return true; }
}
public override bool CanSeek
{
get { return false; }
}
public override long Seek(long offset, SeekOrigin origin)
{
throw new NotImplementedException();
}
}
class ZLibStream : InputStream
{ // uses ZLIB.NET: http://www.componentace.com/download/download.php?editionid=25
private ZInputStream reader; // seriously, why isn't this a stream?
public ZLibStream(Stream stream)
{
reader = new ZInputStream(stream);
}
public override void Close()
{
reader.Close();
base.Close();
}
protected override int ReadNextBlock(byte[] buffer, int offset, int count)
{
// OMG! reader.Read is the base-stream, reader.read is decompressed! yeuch
return reader.read(buffer, offset, count);
}
}
// deliberately doesn't dispose the base-stream
class LimitedStream : InputStream
{
private Stream stream;
private long remaining;
public LimitedStream(Stream stream, long length)
{
if (length < 0) throw new ArgumentOutOfRangeException("length");
if (stream == null) throw new ArgumentNullException("stream");
if (!stream.CanRead) throw new ArgumentException("stream");
this.stream = stream;
this.remaining = length;
}
protected override int ReadNextBlock(byte[] buffer, int offset, int count)
{
if (count > remaining) count = (int)remaining;
int bytesRead = stream.Read(buffer, offset, count);
if (bytesRead > 0) remaining -= bytesRead;
return bytesRead;
}
}
}
package OSM_PROTO;
message Blob {
optional bytes raw = 1;
optional int32 raw_size = 2;
optional bytes zlib_data = 3;
optional bytes lzma_data = 4;
optional bytes bzip2_data = 5;
}
message BlockHeader {
required string type = 1;
optional bytes indexdata = 2;
required int32 datasize = 3;
}
这是在生成的文件中BlockHeader的声明：
public sealed partial class BlockHeader : pb::GeneratedMessage<BlockHeader, BlockHeader.Builder> {...}
-> 使用 pb = global::Google.ProtocolBuffers;
(ProtocolBuffers.dll) 随此软件包一同提供：
你尝试过获取一些较小的区域吗？例如us-pacific.osm.pbf
最终发布错误消息会很有用。