Trying to figuring out how ma_encoder_init works #1046
-
|
Hi mate, thank you for developing miniaudio, and I really have quite some fun on messing around with the low level API, managed to played some audio files and generated some waveforms. Besides, since I am more familiar and feel more comfortable working on zig, I am currently contributing a zig binding for this library; however, I am currently puzzling to the function ma_encoder_init since I don't seem to find any documentation mentioning how to set up onWrite and onSeek function for a custom encoder and how it works, and seems the documentation and the example code only covers ma_encoder_init_file only. The reason I have to use this particular function instead ma_encoder_init_file is that I need to demonstrate the zig binding being functional and usable, but it is not a good idea to generate and to remove a file during the unit test, so I am thinking of building a custom encoder such that I can encode my raw sample into wav format stored in a buffer (instead of writing into an file), and verify the output. I have also looked into this discussion, but the initialization part of the encoder is missing from the example which is one of the key point to my question. I am currently having a look into the function ma_encoder_init__internal, but I am still figuring it out; thus, for the time being, I wish to know:
Please let me know if I have overlooked the documentation for this part or further details and clarification are required. Edit: I think I have figured out something for the onWrite function, but since I am working in zig, I need to find a way to explain that in C; nonetheless, I think I could answer some of the questions (I will keep on update this post when I have found a new finding): |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
|
Alright, this is the first time I have used the discussion board (where I normally use issues in other repo) and I didn't know that I need to set this board to Q&A; nonetheless, I have found my way out, and understood how things work. Before declaring the encoder, we need to have onWrite and onSeek presents so that the ma_encoder_init function can generate the file header using these two functions. To declare a onWrite function, the parameters are the following:
Thus, with knowing all these parameters, if we just want to write something into a buffer, all we need to do is to append the data into a buffer object of your choice stored in pEncoder, and update the number of pBytesWritten; however, since I am currently working in a zig binding, I don't have the c example for now, but the idea should be the same and I will convert that into c code if I have the time: fn testing_on_write(encoder: *Encoder, buffer_in: *anyopaque, bytes_to_write: usize, bytes_written: *usize) callconv(.c) Result {
const buffer_in_data: [*]u8 = @ptrCast(@alignCast(buffer_in));
const result_buffer: *TestingEncodedStorage = @ptrCast(@alignCast(encoder.getUserData()));
result_buffer.data_list.appendSlice(result_buffer.data_allocator, buffer_in_data[0..bytes_to_write]) catch return Result.out_of_memory;
bytes_written.* += bytes_to_write;
return Result.success;
}Since onSeek is not required for my case, I could just declare an empty function that fulfills the function requirement. fn testing_on_seek(encoder: *Encoder, offset: i64, origin: Vfs.SeekOrigin) callconv(.c) Result {
_ = encoder;
_ = offset;
_ = origin;
return Result.success;
}With the two functions, I can now creating the encoder with my own user data, which is a pointer to an array list: var encoder_result = TestingEncodedStorage{
.data_list = try std.ArrayList(u8).initCapacity(std.testing.allocator, 0),
.data_allocator = std.testing.allocator,
};
defer encoder_result.data_list.deinit(std.testing.allocator);
const encoder_cfg = Encoder.Config.init(.wav, .unsigned8, 1, 44100);
try expect(encoder_cfg.channels == 1);
try expect(encoder_cfg.encoding_format == .wav);
try expect(encoder_cfg.format == .unsigned8);
try expect(encoder_cfg.sample_rate == 44100);
var encoder = try Encoder.create(testing_on_write, testing_on_seek, @ptrCast(&encoder_result), encoder_cfg);
defer encoder.destroy();After all, the steps are identical to the documentation about encoder. But now I have a question: because I don't use onSeek, what is the purpose of it? |
Beta Was this translation helpful? Give feedback.
-
|
The usage of the seek callback depends on the encoding backend. Currently it's only used for WAV, and in that case, it's used during uninitialization to write out the final sample count in the header section of the WAV file. The encoding API is not as well developed as other parts of the library, and as a result does not have as detailed documentation. I need to do another pass on it. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you! This explains why the file size in the header doesn't align to the actual file size. Perhaps I could try to use the seek callback function to do the similar. |
Beta Was this translation helpful? Give feedback.
Alright, this is the first time I have used the discussion board (where I normally use issues in other repo) and I didn't know that I need to set this board to Q&A; nonetheless, I have found my way out, and understood how things work.
Before declaring the encoder, we need to have onWrite and onSeek presents so that the ma_encoder_init function can generate the file header using these two functions.
To declare a onWrite function, the parameters are the following:
ma_encoder* pEncoder: self explanatory, and if you need to return the encoded result into an arraylist or a buffer, you many put the buffer in pEncoder->userDataconst void* pBufferIn: this parameter stores the data to be written…