Encoder

git-svn-id: https://svn.kapsi.fi/jpa/nanopb@951 e3a754e5-d11d-0410-8d38-ebb782a927b9

10 files changed, 583 insertions, 104 deletions

diff --git a/generator/nanopb_generator.py b/generator/nanopb_generator.py
index b182d01f..6aff3155 100644
--- a/generator/nanopb_generator.py
+++ b/generator/nanopb_generator.py
@@ -121,7 +121,7 @@ class Field:
             if self.max_size is None:
                 is_callback = True
             else:
-                self.ctype = 'PB_BYTES_ARRAY(%d)' % self.max_size
+                self.ctype = self.struct_name + self.name + 't'
         elif desc.type == FieldD.TYPE_MESSAGE:
             self.ltype = 'PB_LTYPE_SUBMESSAGE'
             self.ctype = names_from_type_name(desc.type_name)
@@ -146,7 +146,18 @@ class Field:
         result += '    %s %s%s;' % (self.ctype, self.name, self.array_decl)
         return result
     
-    def default_decl(self):
+    def types(self):
+        '''Return definitions for any special types this field might need.'''
+        if self.ltype == 'PB_LTYPE_BYTES' and self.max_size is not None:
+            result = 'typedef struct {\n'
+            result += '    size_t size;\n'
+            result += '    uint8_t bytes[%d];\n' % self.max_size
+            result += '} %s;\n' % self.ctype
+        else:
+            result = None
+        return result
+    
+    def default_decl(self, declaration_only = False):
         '''Return definition for this field's default value.'''
         if self.default is None:
             return None
@@ -154,7 +165,7 @@ class Field:
         if self.ltype == 'PB_LTYPE_STRING':
             ctype = 'char'
             if self.max_size is None:
-                array_decl = '[]'
+                return None # Not implemented
             else:
                 array_decl = '[%d]' % self.max_size
             default = self.default.encode('string_escape')
@@ -165,9 +176,9 @@ class Field:
             data = ['0x%02x' % ord(c) for c in data]
             
             if self.max_size is None:
-                ctype = 'PB_BYTES_ARRAY(%d)' % len(data)
+                return None # Not implemented
             else:
-                ctype = 'PB_BYTES_ARRAY(%d)' % self.max_size
+                ctype = self.ctype
             
             default = '{%d, {%s}}' % (len(data), ','.join(data))
             array_decl = ''
@@ -175,7 +186,10 @@ class Field:
             ctype, default = self.ctype, self.default
             array_decl = ''
         
-        return 'const %s %s_default%s = %s;' % (ctype, self.struct_name + self.name, array_decl, default)
+        if declaration_only:
+            return 'extern const %s %s_default%s;' % (ctype, self.struct_name + self.name, array_decl)
+        else:
+            return 'const %s %s_default%s = %s;' % (ctype, self.struct_name + self.name, array_decl, default)
     
     def pb_field_t(self, prev_field_name):
         '''Return the pb_field_t initializer to use in the constant array.
@@ -244,20 +258,32 @@ class Message:
     
     def __str__(self):
         result = 'typedef struct {\n'
-        result += '\n'.join([str(f) for f in self.fields])
+        result += '\n'.join([str(f) for f in self.ordered_fields])
         result += '\n} %s;' % self.name
         return result
     
-    def default_decl(self):
+    def types(self):
+        result = ""
+        for field in self.fields:
+            types = field.types()
+            if types is not None:
+                result += types + '\n'
+        return result
+    
+    def default_decl(self, declaration_only = False):
         result = ""
         for field in self.fields:
-            default = field.default_decl()
+            default = field.default_decl(declaration_only)
             if default is not None:
                 result += default + '\n'
         return result
 
-    def pb_field_t(self):
-        result = 'const pb_field_t %s_fields[] = {\n' % self.name
+    def fields_declaration(self):
+        result = 'extern const pb_field_t %s_fields[%d];' % (self.name, len(self.fields))
+        return result
+
+    def fields_definition(self):
+        result = 'const pb_field_t %s_fields[%d] = {\n' % (self.name, len(self.fields))
         
         prev = None
         for field in self.ordered_fields:
@@ -282,13 +308,8 @@ def iterate_messages(desc, names = Names()):
         for x in iterate_messages(submsg, sub_names):
             yield x
 
-def process_file(fdesc):
-    '''Takes a FileDescriptorProto and generate content for its header file.
-    Generates strings, which should be concatenated and stored to file.
-    '''
-    
-    yield '/* Automatically generated nanopb header */\n'
-    yield '#include <pb.h>\n\n'
+def parse_file(fdesc):
+    '''Takes a FileDescriptorProto and returns tuple (enum, messages).'''
     
     enums = []
     messages = []
@@ -297,10 +318,23 @@ def process_file(fdesc):
         enums.append(Enum(Names(), enum))
     
     for names, message in iterate_messages(fdesc):
+        messages.append(Message(names, message))
         for enum in message.enum_type:
             enums.append(Enum(names, enum))
-        
-        messages.append(Message(names, message))
+    
+    return enums, messages
+
+def generate_header(headername, enums, messages):
+    '''Generate content for a header file.
+    Generates strings, which should be concatenated and stored to file.
+    '''
+    
+    yield '/* Automatically generated nanopb header */\n'
+    
+    symbol = headername.replace('.', '_').upper()
+    yield '#ifndef _PB_%s_\n' % symbol
+    yield '#define _PB_%s_\n' % symbol
+    yield '#include <pb.h>\n\n'
     
     yield '/* Enum definitions */\n'
     for enum in enums:
@@ -309,17 +343,34 @@ def process_file(fdesc):
     yield '/* Struct definitions */\n'
     messages.sort()
     for msg in messages:
+        yield msg.types()
         yield str(msg) + '\n\n'
         
     yield '/* Default values for struct fields */\n'
     for msg in messages:
-        yield msg.default_decl()
+        yield msg.default_decl(True)
     yield '\n'
     
     yield '/* Struct field encoding specification for nanopb */\n'
     for msg in messages:
-        yield msg.pb_field_t() + '\n\n'
+        yield msg.fields_declaration() + '\n'
+    
+    yield '\n#endif\n'
+
+def generate_source(headername, enums, messages):
+    '''Generate content for a source file.'''
+    
+    yield '/* Automatically generated nanopb constant definitions */\n'
+    yield '#include "%s"\n\n' % headername
+    
+    for msg in messages:
+        yield msg.default_decl(False)
     
+    yield '\n\n'
+    
+    for msg in messages:
+        yield msg.fields_definition() + '\n\n'
+
 if __name__ == '__main__':
     import sys
     import os.path
@@ -328,17 +379,26 @@ if __name__ == '__main__':
         print "Usage: " + sys.argv[0] + " file.pb"
         print "where file.pb has been compiled from .proto by:"
         print "protoc -ofile.pb file.proto"
-        print "Output fill be written to file.h"
+        print "Output fill be written to file.h and file.c"
         sys.exit(1)
     
     data = open(sys.argv[1]).read()
     fdesc = descriptor.FileDescriptorSet.FromString(data)
+    enums, messages = parse_file(fdesc.file[0])
     
-    destfile = os.path.splitext(sys.argv[1])[0] + '.h'
-    
-    print "Writing to " + destfile
+    noext = os.path.splitext(sys.argv[1])[0]
+    headername = noext + '.h'
+    sourcename = noext + '.c'
+    headerbasename = os.path.basename(headername)
     
-    destfile = open(destfile, 'w')
+    print "Writing to " + headername + " and " + sourcename
     
-    for part in process_file(fdesc.file[0]):
-        destfile.write(part)
+    header = open(headername, 'w')
+    for part in generate_header(headerbasename, enums, messages):
+        header.write(part)
+
+    source = open(sourcename, 'w')
+    for part in generate_source(headerbasename, enums, messages):
+        source.write(part)
+
+    
\ No newline at end of file
diff --git a/pb.h b/pb.h
index 27ee11fd..46d7c2ce 100644
--- a/pb.h
+++ b/pb.h
@@ -12,21 +12,13 @@
 #define pb_packed
 #endif
 
-/* Lightweight output stream. */
-typedef struct _pb_ostream_t pb_ostream_t;
-struct _pb_ostream_t
-{
-    bool (*callback)(pb_ostream_t *stream, const uint8_t *buf, size_t count);
-    void *state; /* Free field for use by callback implementation */
-    size_t bytes_written;
-};
-
-/*static inline bool pb_write(pb_ostream_t *stream, const uint8_t *buf, size_t count)
-{
-    bool status = stream->callback(stream, buf, count);
-    stream->bytes_written += count;
-    return status;
-}*/
+/* Wire types */
+typedef enum {
+    PB_WT_VARINT = 0,
+    PB_WT_64BIT  = 1,
+    PB_WT_STRING = 2,
+    PB_WT_32BIT  = 5
+} pb_wire_type_t;
 
 /* List of possible field types
  * Least-significant 4 bits tell the scalar type
@@ -83,7 +75,8 @@ typedef enum {
     
     /* Works for all required/optional/repeated fields.
      * data_offset points to pb_callback_t structure.
-     * LTYPE is ignored. */
+     * LTYPE should be 0 (it is ignored, but sometimes
+     * used to speculatively index an array). */
     PB_HTYPE_CALLBACK = 0x30
 } pb_packed pb_type_t;
 
@@ -113,14 +106,13 @@ struct _pb_field_t {
 } pb_packed;
 
 /* This structure is used for 'bytes' arrays.
- * It has the number of bytes in the beginning, and after that an array. */
-#define PB_BYTES_ARRAY(buffersize) \
-struct { \
-    size_t size; \
-    uint8_t bytes[buffersize]; \
-}
-
-typedef PB_BYTES_ARRAY() pb_bytes_array_t;
+ * It has the number of bytes in the beginning, and after that an array.
+ * Note that actual structs used will have a different length of bytes array.
+ */
+typedef struct {
+    size_t size;
+    uint8_t bytes[1];
+} pb_bytes_array_t;
 
 /* This structure is used for giving the callback function.
  * It is stored in the message structure and filled in by the method that
@@ -139,11 +131,12 @@ typedef PB_BYTES_ARRAY() pb_bytes_array_t;
  * wire type. It can write multiple fields.
  */
 typedef struct _pb_istream_t pb_istream_t;
+typedef struct _pb_ostream_t pb_ostream_t;
 typedef struct _pb_callback_t pb_callback_t;
 struct _pb_callback_t {
     union {
         bool (*decode)(pb_istream_t *stream, const pb_field_t *field, void *arg);
-        bool (*encode)(pb_ostream_t *stream, const pb_field_t *field, void *arg);
+        bool (*encode)(pb_ostream_t *stream, const pb_field_t *field, const void *arg);
     } funcs;
     
     /* Free arg for use by callback */
diff --git a/pb_decode.c b/pb_decode.c
index 63008dd5..df053314 100644
--- a/pb_decode.c
+++ b/pb_decode.c
@@ -7,14 +7,19 @@
 #include "pb_decode.h"
 #include <string.h>
 
-const pb_decoder_t PB_DECODERS[PB_LTYPES_COUNT] = {
-    (pb_decoder_t)&pb_dec_varint,
-    (pb_decoder_t)&pb_dec_svarint,
-    (pb_decoder_t)&pb_dec_fixed,
+typedef bool (*pb_decoder_t)(pb_istream_t *stream, const pb_field_t *field, void *dest);
+
+/* --- Function pointers to field decoders ---
+ * Order in the array must match pb_action_t LTYPE numbering.
+ */
+static const pb_decoder_t PB_DECODERS[PB_LTYPES_COUNT] = {
+    &pb_dec_varint,
+    &pb_dec_svarint,
+    &pb_dec_fixed,
     
-    (pb_decoder_t)&pb_dec_bytes,
-    (pb_decoder_t)&pb_dec_string,
-    (pb_decoder_t)&pb_dec_submessage
+    &pb_dec_bytes,
+    &pb_dec_string,
+    &pb_dec_submessage
 };
 
 /**************
@@ -108,33 +113,27 @@ bool pb_skip_string(pb_istream_t *stream)
  * to just assume the correct type and fail safely on corrupt message.
  */
 
-enum wire_type_t {
-    WT_VARINT = 0,
-    WT_64BIT  = 1,
-    WT_STRING = 2,
-    WT_32BIT  = 5
-};
-
 static bool skip(pb_istream_t *stream, int wire_type)
 {
     switch (wire_type)
     {
-        case WT_VARINT: return pb_skip_varint(stream);
-        case WT_64BIT: return pb_read(stream, NULL, 8);
-        case WT_STRING: return pb_skip_string(stream);
-        case WT_32BIT: return pb_read(stream, NULL, 4);
+        case PB_WT_VARINT: return pb_skip_varint(stream);
+        case PB_WT_64BIT: return pb_read(stream, NULL, 8);
+        case PB_WT_STRING: return pb_skip_string(stream);
+        case PB_WT_32BIT: return pb_read(stream, NULL, 4);
         default: return false;
     }
 }
 
-/* Read a raw value to buffer, for the purpose of passing it to callback.
- * Size is maximum size on call, and actual size on return. */
-static bool read_raw_value(pb_istream_t *stream, int wire_type, uint8_t *buf, size_t *size)
+/* Read a raw value to buffer, for the purpose of passing it to callback as
+ * a substream. Size is maximum size on call, and actual size on return.
+ */
+static bool read_raw_value(pb_istream_t *stream, pb_wire_type_t wire_type, uint8_t *buf, size_t *size)
 {
     size_t max_size = *size;
     switch (wire_type)
     {
-        case WT_VARINT:
+        case PB_WT_VARINT:
             *size = 0;
             do
             {
@@ -144,11 +143,11 @@ static bool read_raw_value(pb_istream_t *stream, int wire_type, uint8_t *buf, si
             } while (*buf++ & 0x80);
             return true;
             
-        case WT_64BIT:
+        case PB_WT_64BIT:
             *size = 8;
             return pb_read(stream, buf, 8);
         
-        case WT_32BIT:
+        case PB_WT_32BIT:
             *size = 4;
             return pb_read(stream, buf, 4);
         
@@ -239,7 +238,7 @@ bool decode_field(pb_istream_t *stream, int wire_type, pb_field_iterator_t *iter
             return func(stream, iter->current, iter->pData);
     
         case PB_HTYPE_ARRAY:
-            if (wire_type == WT_STRING
+            if (wire_type == PB_WT_STRING
                 && PB_LTYPE(iter->current->type) <= PB_LTYPE_LAST_PACKABLE)
             {
                 /* Packed array */
@@ -270,7 +269,7 @@ bool decode_field(pb_istream_t *stream, int wire_type, pb_field_iterator_t *iter
             }
         
         case PB_HTYPE_CALLBACK:
-            if (wire_type == WT_STRING)
+            if (wire_type == PB_WT_STRING)
             {
                 pb_callback_t *pCallback = (pb_callback_t*)iter->pData;
                 pb_istream_t substream;
@@ -419,7 +418,7 @@ bool pb_dec_fixed(pb_istream_t *stream, const pb_field_t *field, void *dest)
 #ifdef __BIG_ENDIAN__
     uint8_t bytes[8] = {0};
     bool status = pb_read(stream, bytes, field->data_size);
-    uint8_t lebytes[8] = {bytes[7], bytes[6], bytes[5], bytes[4], 
+    uint8_t bebytes[8] = {bytes[7], bytes[6], bytes[5], bytes[4], 
                           bytes[3], bytes[2], bytes[1], bytes[0]};
     endian_copy(dest, lebytes, field->data_size, 8);
     return status;
@@ -428,7 +427,7 @@ bool pb_dec_fixed(pb_istream_t *stream, const pb_field_t *field, void *dest)
 #endif
 }
 
-bool pb_dec_bytes(pb_istream_t *stream, const pb_field_t *field, uint8_t *dest)
+bool pb_dec_bytes(pb_istream_t *stream, const pb_field_t *field, void *dest)
 {
     pb_bytes_array_t *x = (pb_bytes_array_t*)dest;
     
@@ -445,7 +444,7 @@ bool pb_dec_bytes(pb_istream_t *stream, const pb_field_t *field, uint8_t *dest)
     return pb_read(stream, x->bytes, x->size);
 }
 
-bool pb_dec_string(pb_istream_t *stream, const pb_field_t *field, uint8_t *dest)
+bool pb_dec_string(pb_istream_t *stream, const pb_field_t *field, void *dest)
 {
     uint32_t size;
     bool status;
diff --git a/pb_decode.h b/pb_decode.h
index 448b55db..011efdd9 100644
--- a/pb_decode.h
+++ b/pb_decode.h
@@ -1,18 +1,32 @@
 #ifndef _PB_DECODE_H_
 #define _PB_DECODE_H_
 
+/* pb_decode.h: Functions to decode protocol buffers. Depends on pb_decode.c.
+ * The main function is pb_decode. You will also need to create an input
+ * stream, which is easiest to do with pb_istream_t.
+ * 
+ * You also need structures and their corresponding pb_field_t descriptions.
+ * These are usually generated from .proto-files with a script.
+ */
+
 #include <stdbool.h>
 #include "pb.h"
 
 /* Lightweight input stream.
- * If buf is NULL, read but don't store bytes.
- * You can to provide a callback function for reading or use
+ * You can provide a callback function for reading or use
  * pb_istream_from_buffer.
  * 
- * You can use state to store your own data (e.g. buffer pointer),
+ * Rules for callback:
+ * 1) Return false on IO errors. This will cause decoding to abort.
+ * 
+ * 2) If buf is NULL, read but don't store bytes ("skip input").
+ * 
+ * 3) You can use state to store your own data (e.g. buffer pointer),
  * and rely on pb_read to verify that no-body reads past bytes_left.
- * However, substreams may change bytes_left so don't use that to
- * compute any pointers.
+ * 
+ * 4) Your callback may be used with substreams, in which case bytes_left
+ * is different than from the main stream. Don't use bytes_left to compute
+ * any pointers.
  */
 struct _pb_istream_t
 {
@@ -25,6 +39,7 @@ pb_istream_t pb_istream_from_buffer(uint8_t *buf, size_t bufsize);
 bool pb_read(pb_istream_t *stream, uint8_t *buf, size_t count);
 
 /* Decode from stream to destination struct.
+ * Returns true on success, false on any failure.
  * The actual struct pointed to by dest must match the description in fields.
  */
 bool pb_decode(pb_istream_t *stream, const pb_field_t fields[], void *dest_struct);
@@ -49,15 +64,8 @@ bool pb_dec_varint(pb_istream_t *stream, const pb_field_t *field, void *dest);
 bool pb_dec_svarint(pb_istream_t *stream, const pb_field_t *field, void *dest);
 bool pb_dec_fixed(pb_istream_t *stream, const pb_field_t *field, void *dest);
 
-bool pb_dec_bytes(pb_istream_t *stream, const pb_field_t *field, uint8_t *dest);
-bool pb_dec_string(pb_istream_t *stream, const pb_field_t *field, uint8_t *dest);
+bool pb_dec_bytes(pb_istream_t *stream, const pb_field_t *field, void *dest);
+bool pb_dec_string(pb_istream_t *stream, const pb_field_t *field, void *dest);
 bool pb_dec_submessage(pb_istream_t *stream, const pb_field_t *field, void *dest);
 
-typedef bool (*pb_decoder_t)(pb_istream_t *stream, const pb_field_t *field, void *dest);
-
-/* --- Function pointers to field decoders ---
- * Order in the array must match pb_action_t LTYPE numbering.
- */
-const pb_decoder_t PB_DECODERS[PB_LTYPES_COUNT];
-
 #endif
diff --git a/pb_encode.c b/pb_encode.c
new file mode 100644
index 00000000..ae185c1a
--- /dev/null
+++ b/pb_encode.c
@@ -0,0 +1,326 @@
+/* pb_encode.c -- encode a protobuf using minimal resources
+ *
+ * 2011 Petteri Aimonen <jpa@kapsi.fi>
+ */
+
+#include "pb.h"
+#include "pb_encode.h"
+#include <string.h>
+
+typedef bool (*pb_encoder_t)(pb_ostream_t *stream, const pb_field_t *field, const void *src);
+
+/* --- Function pointers to field encoders ---
+ * Order in the array must match pb_action_t LTYPE numbering.
+ */
+static const pb_encoder_t PB_ENCODERS[PB_LTYPES_COUNT] = {
+    &pb_enc_varint,
+    &pb_enc_svarint,
+    &pb_enc_fixed,
+    
+    &pb_enc_bytes,
+    &pb_enc_string,
+    &pb_enc_submessage
+};
+
+/* pb_ostream_t implementation */
+
+static bool buf_write(pb_ostream_t *stream, const uint8_t *buf, size_t count)
+{
+    uint8_t *dest = (uint8_t*)stream->state;
+    memcpy(dest, buf, count);
+    stream->state = dest + count;
+    return true;
+}
+
+pb_ostream_t pb_ostream_from_buffer(uint8_t *buf, size_t bufsize)
+{
+    pb_ostream_t stream;
+    stream.callback = &buf_write;
+    stream.state = buf;
+    stream.max_size = bufsize;
+    stream.bytes_written = 0;
+    return stream;
+}
+
+bool pb_write(pb_ostream_t *stream, const uint8_t *buf, size_t count)
+{
+    stream->bytes_written += count;
+    
+    if (stream->callback == NULL)
+        return true;
+    
+    if (stream->bytes_written > stream->max_size)
+        return false;
+    
+    return stream->callback(stream, buf, count);
+}
+
+/* Main encoding stuff */
+
+static bool encode_array(pb_ostream_t *stream, const pb_field_t *field,
+                         const void *pData, size_t count, pb_encoder_t func)
+{
+    int i;
+    const void *p;
+    size_t size;
+    
+    if (PB_LTYPE(field->type) < PB_LTYPE_LAST_PACKABLE)
+    {
+        if (!pb_encode_tag(stream, PB_WT_STRING, field->tag))
+            return false;
+        
+        /* Determine the total size of packed array. */
+        if (PB_LTYPE(field->type) == PB_LTYPE_FIXED)
+        {
+            size = field->data_size * count;
+        }
+        else
+        {
+            pb_ostream_t sizestream = {0};
+            p = pData;
+            for (i = 0; i < count; i++)
+            {
+                if (!func(&sizestream, field, p))
+                    return false;
+                p = (const char*)p + field->data_size;
+            }
+            size = sizestream.bytes_written;
+        }
+        
+        pb_encode_varint(stream, size);
+        
+        if (stream->callback == NULL)
+            return pb_write(stream, NULL, size); /* Just sizing.. */
+        
+        /* Write the data */
+        p = pData;
+        for (i = 0; i < count; i++)
+        {
+            if (!func(stream, field, p))
+                return false;
+            p = (const char*)p + field->data_size;
+        }
+    }
+    else
+    {
+        p = pData;
+        for (i = 0; i < count; i++)
+        {
+            if (!pb_encode_tag_for_field(stream, field))
+                return false;
+            if (!func(stream, field, p))
+                return false;
+            p = (const char*)p + field->data_size;
+        }
+    }
+    
+    return true;
+}
+
+bool pb_encode(pb_ostream_t *stream, const pb_field_t fields[], const void *src_struct)
+{
+    const pb_field_t *field = fields;
+    const void *pData = src_struct;
+    const void *pSize;
+    
+    while (field->tag != 0)
+    {
+        pData = (const char*)pData + field->data_offset;
+        pSize = (const char*)pData + field->size_offset;
+        
+        pb_encoder_t func = PB_ENCODERS[PB_LTYPE(field->type)];
+        
+        switch (PB_HTYPE(field->type))
+        {
+            case PB_HTYPE_REQUIRED:
+                if (!pb_encode_tag_for_field(stream, field))
+                    return false;
+                if (!func(stream, field, pData))
+                    return false;
+                break;
+            
+            case PB_HTYPE_OPTIONAL:
+                if (*(bool*)pSize)
+                {
+                    if (!pb_encode_tag_for_field(stream, field))
+                        return false;
+                
+                    if (!func(stream, field, pData))
+                        return false;
+                }
+                break;
+            
+            case PB_HTYPE_ARRAY:
+                if (!encode_array(stream, field, pData, *(size_t*)pSize, func))
+                    return false;
+                break;
+            
+            case PB_HTYPE_CALLBACK:
+            {
+                pb_callback_t *callback = (pb_callback_t*)pData;
+                if (callback->funcs.encode != NULL)
+                {
+                    if (!callback->funcs.encode(stream, field, callback->arg))
+                        return false;
+                }
+                break;
+            }
+        }
+    
+        field++;
+    }
+    
+    return true;
+}
+
+/* Helper functions */
+bool pb_encode_varint(pb_ostream_t *stream, uint64_t value)
+{
+    uint8_t buffer[10];
+    int i = 0;
+    
+    if (value == 0)
+        return pb_write(stream, (uint8_t*)&value, 1);
+    
+    while (value)
+    {
+        buffer[i] = (value & 0x7F) | 0x80;
+        value >>= 7;
+        i++;
+    }
+    buffer[i-1] &= 0x7F; /* Unset top bit on last byte */
+    
+    return pb_write(stream, buffer, i);
+}
+
+bool pb_encode_tag(pb_ostream_t *stream, pb_wire_type_t wiretype, int field_number)
+{
+    int tag = wiretype | (field_number << 3);
+    return pb_encode_varint(stream, tag);
+}
+
+bool pb_encode_tag_for_field(pb_ostream_t *stream, const pb_field_t *field)
+{
+    pb_wire_type_t wiretype;
+    switch (PB_LTYPE(field->type))
+    {
+        case PB_LTYPE_VARINT:
+        case PB_LTYPE_SVARINT:
+            wiretype = PB_WT_VARINT;
+            break;
+        
+        case PB_LTYPE_FIXED:
+            if (field->data_size == 4)
+                wiretype = PB_WT_32BIT;
+            else if (field->data_size == 8)
+                wiretype = PB_WT_64BIT;
+            else
+                return false;
+            break;
+        
+        case PB_LTYPE_BYTES:
+        case PB_LTYPE_STRING:
+        case PB_LTYPE_SUBMESSAGE:
+            wiretype = PB_WT_STRING;
+            break;
+        
+        default:
+            return false;
+    }
+    
+    return pb_encode_tag(stream, wiretype, field->tag);
+}
+
+bool pb_encode_string(pb_ostream_t *stream, uint8_t *buffer, size_t size)
+{
+    if (!pb_encode_varint(stream, size))
+        return false;
+    
+    return pb_write(stream, buffer, size);
+}
+
+/* Field encoders */
+
+/* Copy srcsize bytes from src so that values are casted properly.
+ * On little endian machine, copy to start of dest
+ * On big endian machine, copy to end of dest
+ * destsize must always be larger than srcsize
+ * 
+ * Note: This is the reverse of the endian_copy in pb_decode.c.
+ */
+static void endian_copy(void *dest, const void *src, size_t destsize, size_t srcsize)
+{
+#ifdef __BIG_ENDIAN__
+    memcpy((char*)dest + (destsize - srcsize), src, srcsize);
+#else
+    memcpy(dest, src, srcsize);
+#endif
+}
+
+bool pb_enc_varint(pb_ostream_t *stream, const pb_field_t *field, const void *src)
+{
+    uint64_t value = 0;
+    endian_copy(&value, src, sizeof(value), field->data_size);
+    return pb_encode_varint(stream, value);
+}
+
+bool pb_enc_svarint(pb_ostream_t *stream, const pb_field_t *field, const void *src)
+{
+    uint64_t value = 0;
+    uint64_t zigzagged;
+    uint64_t mask;
+    endian_copy(&value, src, sizeof(value), field->data_size);
+    
+    mask = 0x80 << (field->data_size * 8);
+    zigzagged = (value & ~mask) << 1;
+    if (value & mask) zigzagged |= 1;
+    
+    return pb_encode_varint(stream, value);
+}
+
+bool pb_enc_fixed(pb_ostream_t *stream, const pb_field_t *field, const void *src)
+{
+    #ifdef __BIG_ENDIAN__
+    uint8_t bytes[8] = {0};
+    endian_copy(bytes, src, sizeof(bytes), field->data_size);
+    uint8_t lebytes[8] = {bytes[7], bytes[6], bytes[5], bytes[4], 
+                          bytes[3], bytes[2], bytes[1], bytes[0]};
+    return pb_write(stream, lebytes, field->data_size);
+    #else
+    return pb_write(stream, (uint8_t*)src, field->data_size);
+    #endif
+}
+
+bool pb_enc_bytes(pb_ostream_t *stream, const pb_field_t *field, const void *src)
+{
+    pb_bytes_array_t *bytes = (pb_bytes_array_t*)src;
+    return pb_encode_string(stream, bytes->bytes, bytes->size);
+}
+
+bool pb_enc_string(pb_ostream_t *stream, const pb_field_t *field, const void *src)
+{
+    return pb_encode_string(stream, (uint8_t*)src, strlen((char*)src));
+}
+
+bool pb_enc_submessage(pb_ostream_t *stream, const pb_field_t *field, const void *src)
+{
+    pb_ostream_t sizestream = {0};
+    size_t size;
+    
+    if (field->ptr == NULL)
+        return false;
+    
+    if (!pb_encode(&sizestream, (pb_field_t*)field->ptr, src))
+        return false;
+    
+    size = sizestream.bytes_written;
+    
+    if (!pb_encode_varint(stream, size))
+        return false;
+    
+    if (stream->callback == NULL)
+        return pb_write(stream, NULL, size); /* Just sizing */
+    
+    return pb_encode(stream, (pb_field_t*)field->ptr, src);
+}
+
diff --git a/pb_encode.h b/pb_encode.h
new file mode 100644
index 00000000..95f3483a
--- /dev/null
+++ b/pb_encode.h
@@ -0,0 +1,69 @@
+#ifndef _PB_ENCODE_H_
+#define _PB_ENCODE_H_
+
+/* pb_encode.h: Functions to encode protocol buffers. Depends on pb_encode.c.
+ * The main function is pb_encode. You also need an output stream, structures
+ * and their field descriptions (just like with pb_decode).
+ */
+
+#include <stdbool.h>
+#include "pb.h"
+
+/* Lightweight output stream.
+ * You can provide callback for writing or use pb_ostream_from_buffer.
+ * 
+ * Alternatively, callback can be NULL in which case the stream will just
+ * count the number of bytes that would have been written. In this case
+ * max_size is not checked.
+ *
+ * Rules for callback:
+ * 1) Return false on IO errors. This will cause decoding to abort.
+ * 
+ * 2) You can use state to store your own data (e.g. buffer pointer).
+ * 
+ * 3) pb_write will update bytes_written before your callback runs.
+ * 
+ * 4) Your callback will be always used with the same pb_ostream_t.
+ * There are no substreams when encoding.
+ */
+struct _pb_ostream_t
+{
+    bool (*callback)(pb_ostream_t *stream, const uint8_t *buf, size_t count);
+    void *state; /* Free field for use by callback implementation */
+    size_t max_size; /* Limit number of output bytes written (or use SIZE_MAX). */
+    size_t bytes_written;
+};
+
+pb_ostream_t pb_ostream_from_buffer(uint8_t *buf, size_t bufsize);
+bool pb_write(pb_ostream_t *stream, const uint8_t *buf, size_t count);
+
+/* Encode struct to given output stream.
+ * Returns true on success, false on any failure.
+ * The actual struct pointed to by src_struct must match the description in fields.
+ * All required fields in the struct are assumed to have been filled in.
+ */
+bool pb_encode(pb_ostream_t *stream, const pb_field_t fields[], const void *src_struct);
+
+/* --- Helper functions ---
+ * You may want to use these from your caller or callbacks.
+ */
+
+bool pb_encode_varint(pb_ostream_t *stream, uint64_t value);
+bool pb_encode_tag(pb_ostream_t *stream, pb_wire_type_t wiretype, int field_number);
+bool pb_encode_tag_for_field(pb_ostream_t *stream, const pb_field_t *field);
+bool pb_encode_string(pb_ostream_t *stream, uint8_t *buffer, size_t size);
+
+/* --- Field encoders ---
+ * Each encoder writes the content for the field.
+ * The tag/wire type has been written already.
+ */
+
+bool pb_enc_varint(pb_ostream_t *stream, const pb_field_t *field, const void *src);
+bool pb_enc_svarint(pb_ostream_t *stream, const pb_field_t *field, const void *src);
+bool pb_enc_fixed(pb_ostream_t *stream, const pb_field_t *field, const void *src);
+
+bool pb_enc_bytes(pb_ostream_t *stream, const pb_field_t *field, const void *src);
+bool pb_enc_string(pb_ostream_t *stream, const pb_field_t *field, const void *src);
+bool pb_enc_submessage(pb_ostream_t *stream, const pb_field_t *field, const void *src);
+
+#endif
\ No newline at end of file
diff --git a/tests/Makefile b/tests/Makefile
index 35c6f977..7450e1ed 100644
--- a/tests/Makefile
+++ b/tests/Makefile
@@ -1,14 +1,14 @@
 CFLAGS=-ansi -Wall -Werror -I .. -g -O0
-DEPS=../pb_decode.c ../pb_decode.h ../pb.h person.h
-TESTS=test_decode1 decode_unittests
+DEPS=../pb_decode.c ../pb_decode.h ../pb_encode.c ../pb_encode.h ../pb.h person.h
+TESTS=test_decode1 test_encode1 decode_unittests
 
 all: $(TESTS)
 
 clean:
-	rm -f test_decode1
+	rm -f $(TESTS)
 
 %: %.c $(DEPS)
-	$(CC) $(CFLAGS) -o $@ $< ../pb_decode.c
+	$(CC) $(CFLAGS) -o $@ $< ../pb_decode.c ../pb_encode.c
 
 run_unittests: decode_unittests
 	./decode_unittests
diff --git a/tests/person.proto b/tests/person.proto
index 01b2d4e1..5befb076 100644
--- a/tests/person.proto
+++ b/tests/person.proto
@@ -4,7 +4,7 @@ message Person {
   required string name = 1 [(nanopb).max_size = 40];
   required int32 id = 2;
   optional string email = 3 [(nanopb).max_size = 40];
-  optional bytes test = 5 [default = "abc\x00\x01\x02"];
+  optional bytes test = 5 [default="\x00\x01\x02", (nanopb).max_size = 20];
 
   enum PhoneType {
     MOBILE = 0;
diff --git a/tests/test_decode1.c b/tests/test_decode1.c
index 362c404a..cc4688dc 100644
--- a/tests/test_decode1.c
+++ b/tests/test_decode1.c
@@ -1,9 +1,10 @@
 #include <stdio.h>
-#include <string.h>
-#include <stddef.h>
 #include <pb_decode.h>
 #include "person.h"
 
+/* This test has only one source file anyway.. */
+#include "person.c"
+
 bool print_person(pb_istream_t *stream)
 {
     int i;
diff --git a/tests/test_encode1.c b/tests/test_encode1.c
new file mode 100644
index 00000000..b4998f47
--- /dev/null
+++ b/tests/test_encode1.c
@@ -0,0 +1,23 @@
+#include <stdio.h>
+#include <pb_encode.h>
+#include "person.h"
+
+/* This test has only one source file anyway.. */
+#include "person.c"
+
+bool callback(pb_ostream_t *stream, const uint8_t *buf, size_t count)
+{
+    return fwrite(buf, 1, count, stdout) == count;
+}
+
+int main()
+{
+    Person person = {"Test Person 99", 99, true, "test@person.com",
+        1, {{"555-12345678", true, Person_PhoneType_MOBILE}}};
+    
+    pb_ostream_t stream = {&callback, 0, SIZE_MAX, 0};
+    
+    pb_encode(&stream, Person_fields, &person);
+    
+    return 0;
+}