Added initial ability to pause and resume voice recording.

Telegram/CMakeLists.txt

@@ -1019,6 +1019,7 @@ PRIVATE
     media/audio/media_audio.h
     media/audio/media_audio_capture.cpp
     media/audio/media_audio_capture.h
+    media/audio/media_audio_capture_common.h
     media/audio/media_audio_ffmpeg_loader.cpp
     media/audio/media_audio_ffmpeg_loader.h
     media/audio/media_audio_loader.cpp

Telegram/SourceFiles/history/view/controls/history_view_voice_record_bar.cpp

@@ -84,16 +84,14 @@ enum class FilterType {
 	const int duration = kPrecision
 		* (float64(samples) / ::Media::Player::kDefaultFrequency);
 	const auto durationString = Ui::FormatDurationText(duration / kPrecision);
-	const auto decimalPart = duration % kPrecision;
+	const auto decimalPart = QString::number(duration % kPrecision);
-	return QString("%1%2%3")
+	return durationString + QLocale().decimalPoint() + decimalPart;
-		.arg(durationString, QLocale().decimalPoint())
-		.arg(decimalPart);
 }
 
 [[nodiscard]] std::unique_ptr<VoiceData> ProcessCaptureResult(
-		const ::Media::Capture::Result &data) {
+		const VoiceWaveform &waveform) {
 	auto voiceData = std::make_unique<VoiceData>();
-	voiceData->waveform = data.waveform;
+	voiceData->waveform = waveform;
 	voiceData->wavemax = voiceData->waveform.empty()
 		? uchar(0)
 		: *ranges::max_element(voiceData->waveform);
@@ -427,12 +425,11 @@ public:
 		not_null<Ui::RpWidget*> parent,
 		const style::RecordBar &st,
 		not_null<Main::Session*> session,
-		::Media::Capture::Result &&data,
+		::Media::Capture::Result *data,
 		const style::font &font);
 
 	void requestPaintProgress(float64 progress);
 	rpl::producer<> stopRequests() const;
-	::Media::Capture::Result *data() const;
 
 	void playPause();
 
@@ -456,7 +453,7 @@ private:
 	const not_null<DocumentData*> _document;
 	const std::unique_ptr<VoiceData> _voiceData;
 	const std::shared_ptr<Data::DocumentMedia> _mediaView;
-	const std::unique_ptr<::Media::Capture::Result> _data;
+	const not_null<::Media::Capture::Result*> _data;
 	const base::unique_qptr<Ui::IconButton> _delete;
 	const style::font &_durationFont;
 	const QString _duration;
@@ -486,15 +483,15 @@ ListenWrap::ListenWrap(
 	not_null<Ui::RpWidget*> parent,
 	const style::RecordBar &st,
 	not_null<Main::Session*> session,
-	::Media::Capture::Result &&data,
+	::Media::Capture::Result *data,
 	const style::font &font)
 : _parent(parent)
 , _st(st)
 , _session(session)
 , _document(DummyDocument(&session->data()))
-, _voiceData(ProcessCaptureResult(data))
+, _voiceData(ProcessCaptureResult(data->waveform))
 , _mediaView(_document->createMediaView())
-, _data(std::make_unique<::Media::Capture::Result>(std::move(data)))
+, _data(data)
 , _delete(base::make_unique_q<Ui::IconButton>(parent, _st.remove))
 , _durationFont(font)
 , _duration(Ui::FormatDurationText(
@@ -817,10 +814,6 @@ rpl::producer<> ListenWrap::stopRequests() const {
 	return _delete->clicks() | rpl::to_empty;
 }
 
-::Media::Capture::Result *ListenWrap::data() const {
-	return _data.get();
-}
-
 rpl::lifetime &ListenWrap::lifetime() {
 	return _lifetime;
 }
@@ -1293,12 +1286,14 @@ void VoiceRecordBar::updateTTLGeometry(
 		const auto from = -_ttlButton->width();
 		const auto right = anim::interpolate(from, finalRight, progress);
 		_ttlButton->moveToRight(right, _ttlButton->y());
+#if 0
 	} else if (type == TTLAnimationType::TopBottom) {
 		const auto ttlFrom = anyTop - _ttlButton->height() * 2;
 		const auto ttlTo = anyTop - _lock->height();
 		_ttlButton->moveToLeft(
 			_ttlButton->x(),
 			anim::interpolate(ttlFrom, ttlTo, 1. - progress));
+#endif
 	} else if (type == TTLAnimationType::RightTopStatic) {
 		_ttlButton->moveToRight(
 			-_ttlButton->width(),
@@ -1408,48 +1403,7 @@ void VoiceRecordBar::init() {
 		_showLockAnimation.start(std::move(callback), from, to, duration);
 	}, lifetime());
 
-	_lock->setClickedCallback([=] {
+	const auto setLevelAsSend = [=] {
-		if (!_lock->isStopState()) {
-			return;
-		}
-
-		::Media::Capture::instance()->startedChanges(
-		) | rpl::filter([=](bool capturing) {
-			return !capturing && _listen;
-		}) | rpl::take(1) | rpl::start_with_next([=] {
-			_lockShowing = false;
-
-			const auto to = 1.;
-			const auto &duration = st::historyRecordVoiceShowDuration;
-			auto callback = [=](float64 value) {
-				_listen->requestPaintProgress(value);
-				const auto reverseValue = to - value;
-				_level->requestPaintProgress(reverseValue);
-				update();
-				if (to == value) {
-					_recordingLifetime.destroy();
-				}
-				updateTTLGeometry(TTLAnimationType::TopBottom, 1. - value);
-			};
-			_showListenAnimation.stop();
-			_showListenAnimation.start(std::move(callback), 0., to, duration);
-		}, lifetime());
-
-		stopRecording(StopType::Listen);
-	});
-
-	_lock->locks(
-	) | rpl::start_with_next([=] {
-		if (_hasTTLFilter && _hasTTLFilter()) {
-			if (!_ttlButton) {
-				_ttlButton = std::make_unique<TTLButton>(
-					_outerContainer,
-					_st);
-			}
-			_ttlButton->show();
-		}
-		updateTTLGeometry(TTLAnimationType::RightTopStatic, 0);
-
 		_level->setType(VoiceRecordButton::Type::Send);
 
 		_level->clicks(
@@ -1464,6 +1418,58 @@ void VoiceRecordBar::init() {
 		) | rpl::start_with_next([=](bool enter) {
 			_inField = enter;
 		}, _recordingLifetime);
+	};
+
+	_lock->setClickedCallback([=] {
+		if (isListenState()) {
+			startRecording();
+			_listen = nullptr;
+			setLevelAsSend();
+
+			return;
+		}
+		if (!_lock->isStopState()) {
+			return;
+		}
+
+		stopRecording(StopType::Listen);
+	});
+
+	_paused.value() | rpl::distinct_until_changed(
+	) | rpl::start_with_next([=](bool paused) {
+		if (!paused) {
+			return;
+		}
+		// _lockShowing = false;
+
+		const auto to = 1.;
+		const auto &duration = st::historyRecordVoiceShowDuration;
+		auto callback = [=](float64 value) {
+			_listen->requestPaintProgress(value);
+			const auto reverseValue = to - value;
+			_level->requestPaintProgress(reverseValue);
+			update();
+			if (to == value) {
+				_recordingLifetime.destroy();
+			}
+		};
+		_showListenAnimation.stop();
+		_showListenAnimation.start(std::move(callback), 0., to, duration);
+	}, lifetime());
+
+	_lock->locks(
+	) | rpl::start_with_next([=] {
+		if (_hasTTLFilter && _hasTTLFilter()) {
+			if (!_ttlButton) {
+				_ttlButton = std::make_unique<TTLButton>(
+					_outerContainer,
+					_st);
+			}
+			_ttlButton->show();
+		}
+		updateTTLGeometry(TTLAnimationType::RightTopStatic, 0);
+
+		setLevelAsSend();
 
 		const auto &duration = st::historyRecordVoiceShowDuration;
 		const auto from = 0.;
@@ -1616,7 +1622,12 @@ void VoiceRecordBar::startRecording() {
 		startRedCircleAnimation();
 
 		_recording = true;
+		if (_paused.current()) {
+			_paused = false;
+			instance()->pause(false, nullptr);
+		} else {
 			instance()->start();
+		}
 		instance()->updated(
 		) | rpl::start_with_next_error([=](const Update &update) {
 			_recordingTipRequired = (update.samples < kMinSamples);
@@ -1685,7 +1696,7 @@ void VoiceRecordBar::stop(bool send) {
 		const auto type = send ? StopType::Send : StopType::Cancel;
 		stopRecording(type, ttlBeforeHide);
 	};
-	_lockShowing = false;
+	// _lockShowing = false;
 	visibilityAnimate(false, std::move(disappearanceCallback));
 }
 
@@ -1695,6 +1706,7 @@ void VoiceRecordBar::finish() {
 	_inField = false;
 	_redCircleProgress = 0.;
 	_recordingSamples = 0;
+	_paused = false;
 
 	_showAnimation.stop();
 	_lockToStopAnimation.stop();
@@ -1704,6 +1716,8 @@ void VoiceRecordBar::finish() {
 	[[maybe_unused]] const auto s = takeTTLState();
 
 	_sendActionUpdates.fire({ Api::SendProgressType::RecordVoice, -1 });
+
+	_data = {};
 }
 
 void VoiceRecordBar::hideFast() {
@@ -1719,43 +1733,53 @@ void VoiceRecordBar::stopRecording(StopType type, bool ttlBeforeHide) {
 		instance()->stop(crl::guard(this, [=](Result &&data) {
 			_cancelRequests.fire({});
 		}));
+	} else if (type == StopType::Listen) {
+		instance()->pause(true, crl::guard(this, [=](Result &&data) {
+			if (data.bytes.isEmpty()) {
+				// Close everything.
+				stop(false);
 				return;
 			}
+			_paused = true;
+			_data = std::move(data);
+
+			window()->raise();
+			window()->activateWindow();
+			_listen = std::make_unique<ListenWrap>(
+				this,
+				_st,
+				&_show->session(),
+				&_data,
+				_cancelFont);
+			_listenChanges.fire({});
+
+			// _lockShowing = false;
+		}));
+	} else if (type == StopType::Send) {
 		instance()->stop(crl::guard(this, [=](Result &&data) {
 			if (data.bytes.isEmpty()) {
 				// Close everything.
 				stop(false);
 				return;
 			}
+			_data = std::move(data);
 
 			window()->raise();
 			window()->activateWindow();
-		const auto duration = Duration(data.samples);
-		if (type == StopType::Send) {
 			const auto options = Api::SendOptions{
 				.ttlSeconds = (ttlBeforeHide
 					? std::numeric_limits<int>::max()
 					: 0),
 			};
 			_sendVoiceRequests.fire({
-				data.bytes,
+				_data.bytes,
-				data.waveform,
+				_data.waveform,
-				duration,
+				Duration(_data.samples),
 				options,
 			});
-		} else if (type == StopType::Listen) {
-			_listen = std::make_unique<ListenWrap>(
-				this,
-				_st,
-				&_show->session(),
-				std::move(data),
-				_cancelFont);
-			_listenChanges.fire({});
-
-			_lockShowing = false;
-		}
 		}));
 	}
+}
 
 void VoiceRecordBar::drawDuration(QPainter &p) {
 	const auto duration = FormatVoiceDuration(_recordingSamples);
@@ -1811,14 +1835,13 @@ void VoiceRecordBar::drawMessage(QPainter &p, float64 recordActive) {
 
 void VoiceRecordBar::requestToSendWithOptions(Api::SendOptions options) {
 	if (isListenState()) {
-		const auto data = _listen->data();
 		if (takeTTLState()) {
 			options.ttlSeconds = std::numeric_limits<int>::max();
 		}
 		_sendVoiceRequests.fire({
-			data->bytes,
+			_data.bytes,
-			data->waveform,
+			_data.waveform,
-			Duration(data->samples),
+			Duration(_data.samples),
 			options,
 		});
 	}
@@ -1837,7 +1860,7 @@ rpl::producer<> VoiceRecordBar::cancelRequests() const {
 }
 
 bool VoiceRecordBar::isRecording() const {
-	return _recording.current();
+	return _recording.current() && !_paused.current();
 }
 
 bool VoiceRecordBar::isRecordingLocked() const {

Telegram/SourceFiles/history/view/controls/history_view_voice_record_bar.h

@@ -10,6 +10,7 @@ https://github.com/telegramdesktop/tdesktop/blob/master/LEGAL
 #include "api/api_common.h"
 #include "base/timer.h"
 #include "history/view/controls/compose_controls_common.h"
+#include "media/audio/media_audio_capture_common.h"
 #include "ui/effects/animations.h"
 #include "ui/round_rect.h"
 #include "ui/rp_widget.h"
@@ -162,6 +163,9 @@ private:
 	std::unique_ptr<Ui::AbstractButton> _ttlButton;
 	std::unique_ptr<ListenWrap> _listen;
 
+	::Media::Capture::Result _data;
+	rpl::variable<bool> _paused;
+
 	base::Timer _startTimer;
 
 	rpl::event_stream<SendActionUpdate> _sendActionUpdates;

Telegram/SourceFiles/media/audio/media_audio_capture.cpp

@@ -7,6 +7,7 @@ https://github.com/telegramdesktop/tdesktop/blob/master/LEGAL
 */
 #include "media/audio/media_audio_capture.h"
 
+#include "media/audio/media_audio_capture_common.h"
 #include "media/audio/media_audio_ffmpeg_loader.h"
 #include "ffmpeg/ffmpeg_utility.h"
 #include "base/timer.h"
@@ -37,6 +38,45 @@ bool ErrorHappened(ALCdevice *device) {
 	return false;
 }
 
+[[nodiscard]] VoiceWaveform CollectWaveform(
+		const QVector<uchar> &waveformVector) {
+	if (waveformVector.isEmpty()) {
+		return {};
+	}
+	auto waveform = VoiceWaveform();
+	auto count = int64(waveformVector.size());
+	auto sum = int64(0);
+	if (count >= Player::kWaveformSamplesCount) {
+		auto peaks = QVector<uint16>();
+		peaks.reserve(Player::kWaveformSamplesCount);
+
+		auto peak = uint16(0);
+		for (auto i = int32(0); i < count; ++i) {
+			auto sample = uint16(waveformVector.at(i)) * 256;
+			if (peak < sample) {
+				peak = sample;
+			}
+			sum += Player::kWaveformSamplesCount;
+			if (sum >= count) {
+				sum -= count;
+				peaks.push_back(peak);
+				peak = 0;
+			}
+		}
+
+		auto sum = std::accumulate(peaks.cbegin(), peaks.cend(), 0LL);
+		peak = qMax(int32(sum * 1.8 / peaks.size()), 2500);
+
+		waveform.resize(peaks.size());
+		for (int32 i = 0, l = peaks.size(); i != l; ++i) {
+			waveform[i] = char(qMin(
+				31U,
+				uint32(qMin(peaks.at(i), peak)) * 31 / peak));
+		}
+	}
+	return waveform;
+}
+
 } // namespace
 
 class Instance::Inner final : public QObject {
@@ -46,6 +86,7 @@ public:
 
 	void start(Fn<void(Update)> updated, Fn<void()> error);
 	void stop(Fn<void(Result&&)> callback = nullptr);
+	void pause(bool value, Fn<void(Result&&)> callback);
 
 private:
 	void process();
@@ -67,6 +108,8 @@ private:
 	base::Timer _timer;
 	QByteArray _captured;
 
+	bool _paused = false;
+
 };
 
 void Start() {
@@ -118,6 +161,17 @@ void Instance::stop(Fn<void(Result&&)> callback) {
 	});
 }
 
+void Instance::pause(bool value, Fn<void(Result&&)> callback) {
+	Expects(callback != nullptr || !value);
+	InvokeQueued(_inner.get(), [=] {
+		_inner->pause(value, [=](Result &&result) {
+			crl::on_main([=, result = std::move(result)]() mutable {
+				callback(std::move(result));
+			});
+		});
+	});
+}
+
 void Instance::check() {
 	_available = false;
 	if (auto device = alcGetString(0, ALC_CAPTURE_DEFAULT_DEVICE_SPECIFIER)) {
@@ -241,6 +295,9 @@ void Instance::Inner::fail() {
 void Instance::Inner::start(Fn<void(Update)> updated, Fn<void()> error) {
 	_updated = std::move(updated);
 	_error = std::move(error);
+	if (_paused) {
+		_paused = false;
+	}
 
 	// Start OpenAL Capture
 	d->device = alcCaptureOpenDevice(nullptr, kCaptureFrequency, AL_FORMAT_MONO16, kCaptureFrequency / 5);
@@ -404,10 +461,23 @@ void Instance::Inner::start(Fn<void(Update)> updated, Fn<void()> error) {
 	DEBUG_LOG(("Audio Capture: started!"));
 }
 
+void Instance::Inner::pause(bool value, Fn<void(Result&&)> callback) {
+	_paused = value;
+	if (!_paused) {
+		return;
+	}
+	callback({
+		d->fullSamples ? d->data : QByteArray(),
+		d->fullSamples ? CollectWaveform(d->waveform) : VoiceWaveform(),
+		qint32(d->fullSamples),
+	});
+}
+
 void Instance::Inner::stop(Fn<void(Result&&)> callback) {
 	if (!_timer.isActive()) {
 		return; // in stop() already
 	}
+	_paused = false;
 	_timer.cancel();
 
 	const auto needResult = (callback != nullptr);
@@ -480,33 +550,7 @@ void Instance::Inner::stop(Fn<void(Result&&)> callback) {
 	VoiceWaveform waveform;
 	qint32 samples = d->fullSamples;
 	if (needResult && samples && !d->waveform.isEmpty()) {
-		int64 count = d->waveform.size(), sum = 0;
+		waveform = CollectWaveform(d->waveform);
-		if (count >= Player::kWaveformSamplesCount) {
-			QVector<uint16> peaks;
-			peaks.reserve(Player::kWaveformSamplesCount);
-
-			uint16 peak = 0;
-			for (int32 i = 0; i < count; ++i) {
-				uint16 sample = uint16(d->waveform.at(i)) * 256;
-				if (peak < sample) {
-					peak = sample;
-				}
-				sum += Player::kWaveformSamplesCount;
-				if (sum >= count) {
-					sum -= count;
-					peaks.push_back(peak);
-					peak = 0;
-				}
-			}
-
-			auto sum = std::accumulate(peaks.cbegin(), peaks.cend(), 0LL);
-			peak = qMax(int32(sum * 1.8 / peaks.size()), 2500);
-
-			waveform.resize(peaks.size());
-			for (int32 i = 0, l = peaks.size(); i != l; ++i) {
-				waveform[i] = char(qMin(31U, uint32(qMin(peaks.at(i), peak)) * 31 / peak));
-			}
-		}
 	}
 	if (hadDevice) {
 		if (d->codecContext) {
@@ -568,6 +612,10 @@ void Instance::Inner::stop(Fn<void(Result&&)> callback) {
 void Instance::Inner::process() {
 	Expects(!d->processing);
 
+	if (_paused) {
+		return;
+	}
+
 	d->processing = true;
 	const auto guard = gsl::finally([&] { d->processing = false; });
 

Telegram/SourceFiles/media/audio/media_audio_capture.h

@@ -19,11 +19,7 @@ struct Update {
 	ushort level = 0;
 };
 
-struct Result {
+struct Result;
-	QByteArray bytes;
-	VoiceWaveform waveform;
-	int samples = 0;
-};
 
 void Start();
 void Finish();
@@ -51,6 +47,7 @@ public:
 
 	void start();
 	void stop(Fn<void(Result&&)> callback = nullptr);
+	void pause(bool value, Fn<void(Result&&)> callback);
 
 private:
 	class Inner;

Telegram/SourceFiles/media/audio/media_audio_capture_common.h

@@ -0,0 +1,18 @@
+/*
+This file is part of Telegram Desktop,
+the official desktop application for the Telegram messaging service.
+
+For license and copyright information please follow this link:
+https://github.com/telegramdesktop/tdesktop/blob/master/LEGAL
+*/
+#pragma once
+
+namespace Media::Capture {
+
+struct Result {
+	QByteArray bytes;
+	VoiceWaveform waveform;
+	int samples = 0;
+};
+
+} // namespace Media::Capture