MIGRATION

0.5.0 => 1.0.0

  Messages

    Raw pointer access has been removed, because unaligned access of floats
    and doubles may cause undefined behavior on some architectures. Now, the
    values should be loaded and stored with the get and put accessors.

    This makes it necessary to regenerate the bindings and change all accessors
    that write values through raw pointers.

    < 1.0.0:

      float *value = person_raw_weight(&person);

    = 1.0.0:

      float value;
      person_get_weight(&person, &value);
      person_put_weight(&person, &value);

0.4.0 => 0.5.0

  The bindings have to be regenerated.

0.3.x => 0.4.0

  Encoder

    In order to handle packed fields, the encoder does now take a fourth
    argument, namely the amount of values to be encoded. This is a prerequisite
    for the support of packed fields. Now, multiple values can be encoded all
    at once, regardless of whether a field is packed or not. This also applies
    for messages, which have already been implemented as nested encoders.

    This makes it necessary to regenerate the bindings and change all accessors
    that write values to repeated fields.

    < 0.4.0:

      pb_encoder_t person = person_encoder_create(...);
      person_encode_phone(&person, &phonenumbers[0]);
      person_encode_phone(&person, &phonenumbers[1]);

    = 0.4.0:

      pb_encoder_t person = person_encoder_create(...);
      person_encode_phone(&person, &phonenumbers, 2);

0.2.x => 0.3.0

  Binaries

    The "binary" which holds the underlying serialized data in wire format was
    renamed to be a "buffer". Additionally, the journaling functionality is now
    part of a "journal" which itself contains a buffer that is tracked. The
    former is necessary for encoding and decoding messages, the later for
    performing random access on encoded messages.

    < 0.3.0:

      pb_binary_t binary = pb_binary_create(...);
      pb_binary_destroy(&binary);

    = 0.3.0:

      pb_buffer_t buffer = pb_buffer_create(...);
      pb_buffer_destroy(&buffer);

      pb_journal_t journal = pb_journal_create(...);
      pb_journal_destroy(&journal);

  Validation

    Message validation was refactored into a separate "validator" structure,
    so that validation can directly take place on a buffer. This step was
    necessary to make validation available to the lite runtime.

    < 0.3.0:

      pb_message_t message = pb_message_create(...);
      pb_message_check(&message);

    = 0.3.0:

      pb_validator_t validator = pb_validator_create(...);
      pb_validator_check(&validator, &buffer);

0.1.x => 0.2.0

  Validation

    The function for message validation was renamed into pb_message_check() in
    order to remove any ambiguities with pb_message_valid() which checks if the
    message is still valid in terms of alignment.

    < 0.2.0

      pb_message_t message = pb_message_create(...);
      pb_message_validate(&message);

    = 0.2.0

      pb_message_t message = pb_message_create(...);
      pb_message_check(&message);