import os
import time
import sokol.audio

struct Player {
mut:
	samples  []f32
	pos      int
	finished bool
}

fn main() {
	if os.args.len < 2 {
		eprintln('Usage: play_wav file1.wav file2.wav ...')
		play_sounds([os.resource_abs_path('uhoh.wav')]) ?
		exit(1)
	}
	play_sounds(os.args[1..]) ?
}

fn play_sounds(files []string) ? {
	mut player := Player{}
	player.init()
	for f in files {
		if !os.exists(f) || os.is_dir(f) {
			eprintln('skipping "$f" (does not exist)')
			continue
		}
		fext := os.file_ext(f).to_lower()
		if fext != '.wav' {
			eprintln('skipping "$f" (not a .wav file)')
			continue
		}
		player.play_wav_file(f) ?
	}
	player.stop()
}

//
fn audio_player_callback(buffer &f32, num_frames int, num_channels int, mut p Player) {
	if p.finished {
		return
	}
	ntotal := num_channels * num_frames
	nremaining := p.samples.len - p.pos
	nsamples := if nremaining < ntotal { nremaining } else { ntotal }
	if nsamples <= 0 {
		p.finished = true
		return
	}
	unsafe { C.memcpy(buffer, &p.samples[p.pos], nsamples * int(sizeof(f32))) }
	p.pos += nsamples
}

fn (mut p Player) init() {
	audio.setup(
		num_channels: 2
		stream_userdata_cb: audio_player_callback
		user_data: p
	)
}

fn (mut p Player) stop() {
	audio.shutdown()
	p.free()
}

fn (mut p Player) play_wav_file(fpath string) ? {
	println('> play_wav_file: $fpath')
	samples := read_wav_file_samples(fpath) ?
	p.finished = true
	p.samples << samples
	p.finished = false
	for !p.finished {
		time.sleep_ms(16)
	}
	p.free()
}

fn (mut p Player) free() {
	p.finished = false
	p.samples = []f32{}
	p.pos = 0
}

// The read_wav_file_samples function below is based on the following sources:
// http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/WAVE.html
// http://www.lightlink.com/tjweber/StripWav/WAVE.html
// http://www.lightlink.com/tjweber/StripWav/Canon.html
// https://tools.ietf.org/html/draft-ema-vpim-wav-00
// NB: > The chunks MAY appear in any order except that the Format chunk
// > MUST be placed before the Sound data chunk (but not necessarily
// > contiguous to the Sound data chunk).
struct RIFFHeader {
	riff      [4]byte
	file_size u32
	form_type [4]byte
}

struct RIFFChunkHeader {
	chunk_type [4]byte
	chunk_size u32
	chunk_data voidptr
}

struct RIFFFormat {
	format_tag            u16      // PCM = 1; Values other than 1 indicate some form of compression.
	nchannels             u16      // Nc ; 1 = mono ; 2 = stereo
	sample_rate           u32      // F
	avg_bytes_per_second  u32      // F * M*Nc
	nblock_align          u16      // M*Nc
	bits_per_sample       u16      // 8 * M
	cbsize                u16      // Size of the extension: 22
	valid_bits_per_sample u16      // at most 8*M
	channel_mask          u32      // Speaker position mask
	sub_format            [16]byte // GUID
}

fn read_wav_file_samples(fpath string) ?[]f32 {
	mut res := []f32{}
	// eprintln('> read_wav_file_samples: $fpath -------------------------------------------------')
	mut bytes := os.read_bytes(fpath) ?
	mut pbytes := byteptr(bytes.data)
	mut offset := u32(0)
	rh := &RIFFHeader(pbytes)
	// eprintln('rh: $rh')
	if rh.riff != [byte(`R`), `I`, `F`, `F`]! {
		return error('WAV should start with `RIFF`')
	}
	if rh.form_type != [byte(`W`), `A`, `V`, `E`]! {
		return error('WAV should have `WAVE` form type')
	}
	if rh.file_size + 8 != bytes.len {
		return error('WAV should have valid lenght')
	}
	offset += sizeof(RIFFHeader)
	mut rf := &RIFFFormat(0)
	for {
		if offset >= bytes.len {
			break
		}
		//
		ch := &RIFFChunkHeader(unsafe { pbytes + offset })
		offset += 8 + ch.chunk_size
		// eprintln('ch: $ch')
		// eprintln('p: $pbytes | offset: $offset | bytes.len: $bytes.len')
		// ////////
		if ch.chunk_type == [byte(`L`), `I`, `S`, `T`]! {
			continue
		}
		//
		if ch.chunk_type == [byte(`i`), `d`, `3`, ` `]! {
			continue
		}
		//
		if ch.chunk_type == [byte(`f`), `m`, `t`, ` `]! {
			// eprintln('`fmt ` chunk')
			rf = unsafe { &RIFFFormat(&ch.chunk_data) }
			// eprintln('fmt riff format: $rf')
			if rf.format_tag != 1 {
				return error('only PCM encoded WAVs are supported')
			}
			if rf.nchannels < 1 || rf.nchannels > 2 {
				return error('only mono or stereo WAVs are supported')
			}
			if rf.bits_per_sample !in [u16(8), 16] {
				return error('only 8 or 16 bits per sample WAVs are supported')
			}
			continue
		}
		//
		if ch.chunk_type == [byte(`d`), `a`, `t`, `a`]! {
			if rf == 0 {
				return error('`data` chunk should be after `fmt ` chunk')
			}
			// eprintln('`fmt ` chunk: $rf\n`data` chunk: $ch')
			mut doffset := 0
			mut dp := byteptr(&ch.chunk_data)
			for doffset < ch.chunk_size {
				for c := 0; c < rf.nchannels; c++ {
					mut x := f32(0.0)
					mut step := 0
					ppos := unsafe { dp + doffset }
					if rf.bits_per_sample == 8 {
						d8 := byteptr(ppos)
						x = (f32(*d8) - 128) / 128.0
						step = 1
						doffset++
					}
					if rf.bits_per_sample == 16 {
						d16 := &i16(ppos)
						x = f32(*d16) / 32768.0
						step = 2
					}
					doffset += step
					if doffset < ch.chunk_size {
						res << x
						if rf.nchannels == 1 {
							// Duplicating single channel mono sounds,
							// produces a stereo sound, simplifying further processing:
							res << x
						}
					}
				}
			}
		}
	}
	return res
}