Audio examples #644

rothmichaels · 2024-11-19T17:20:10Z

No description provided.

example/lfo.cpp

mpusz · 2024-11-19T20:01:52Z

example/lfo.cpp

+
+  musical_context m_context;
+  quantity<si::hertz, float> m_frequency;
+  quantity_point<angular::radian, float> m_phase{0.f};


SI also has radian but does not have revolutions. I am not sure which one is the better use case considering that you use SI already.

I'm not sure I have a strong opinion. I went with this both because of angular::revolutions and having a base dimension angle seems more natural than angle being m / m.

If you think it would be better for the example to use si::radian and 2π * si::radian instead of angle::revolutions. Any strong opinions?

I can also poll my colleagues (who are not on holiday) tomorrow about this to get their thoughts.

This was just an FYI so you can use the best tool for the job. You know better what is needed than me 😉

Split out audio units and musical context functions into individual headers to be used in future examples in addition to the LFO example.

Add an example of wrapping an unsafe OS or host application API with a type-safe quanitites API. The `audio::get_musical_context` API used in this example will also be used in other audio examples.

example/audio/audio.h

example/audio/third_party_audio_api.h

example/audio/wrapped_third_party_audio_api.h

mpusz · 2024-12-20T08:43:07Z

example/audio/1-oscillator.cpp

+
+  void set_period(QuantityOf<isq::time> auto period)
+  {
+    m_frequency = 1.f / value_cast<float>(period);


You can also use inverse() from math.h that provides additional safety (https://aurora-opensource.github.io/au/main/reference/math/#inverse_as-inverse_in). However, it is not that mandatory as you are using floating-point representations.

Oh nice, I like this.

mpusz · 2024-12-20T08:44:55Z

example/audio/1-oscillator.cpp

+  void set_period(QuantityOf<audio::beat_count> auto period)
+  {
+    std::cout << MP_UNITS_STD_FMT::format("Setting period to {} -- ", period);
+    set_period(value_cast<float>(period) / m_context.current_tempo);


This value_cast might not be needed here as current_tempo is already a floating point.

mpusz · 2024-12-20T08:48:54Z

example/audio/1-oscillator.cpp

+  void reset() { m_phase = phase_t{0.f * angular::radian}; }
+
+private:
+  using phase_t = quantity_point<angular::radian, mp_units::default_point_origin(angular::radian), float>;


Yes, it is unfortunate that a point origin must be provided when we want to specify a specific representation type. I have no ideas on how to improve here. I believe that for points, most people will use double, and the main interest will be to work against different origins. This is why I provided arguments in such order. But maybe I am wrong, or is there another way to improve here?

@JohelEGP and @chiphogg, do you have any ideas here?

If the origin is an expression, {}.

I am not sure what you mean by that?

I meant something like this, but it turns out that it doesn't work with auto NTTPs, whereas it does with function parameters (https://cpp1.godbolt.org/z/hPY9KeYfa):

#include <initializer_list> #include <type_traits> template<int Origin = 1, class Rep = double> struct point; static_assert(std::is_same_v<point<{}, int>, point<0, int>>);

mpusz · 2024-12-20T09:00:40Z

example/audio/1-oscillator.cpp

+  // duration equal to 2 measures of 4/4 music (i.e. 2 whole notes at
+  // the current tempo):
+  const auto beats = 2 * audio::whole_note;
+  const auto buffer_duration = value_cast<float>(beats) / context.current_tempo;


value_cast is probably not needed here?

mpusz · 2024-12-20T09:04:41Z

example/audio/1-oscillator.cpp

+  // of audio samples.  In this example we will create a buffer with
+  // duration equal to 2 measures of 4/4 music (i.e. 2 whole notes at
+  // the current tempo):
+  const auto beats = 2 * audio::whole_note;


It is up to you, and this is purely a matter of personal preference. 😉 AAA (Almost Always Auto) is a nice style, and I like it in many cases (especially not to repeat myself).

However, with time, I have learned to appreciate CTAD as it makes the code easier to reason about. Using quantity and quantity_point CTAD placeholders instead of auto probably makes the code easier for a new user to understand or maintain.

const Quantity auto beats is also an option, but it is more to type and does not have any benefits here over just const quantity beats.

mpusz · 2024-12-20T09:11:00Z

example/audio/1-oscillator.cpp

+  std::cout << MP_UNITS_STD_FMT::format("Filling buffer with values from LFO @ {}", sin_gen.get_frequency());
+  std::generate(begin(buffer_1), end(buffer_1), sin_gen);
+
+  assert(buffer_1.size() > 0u);


I am not sure what this assert does as you set the vector's size explicitly in line 132. Did you mean reserve() and then using back_inserter in the algorithm?

mpusz · 2024-12-20T09:11:44Z

example/audio/1-oscillator.cpp

+  for (std::size_t i = 1u; i < buffer_1.size(); ++i) {
+    std::cout << MP_UNITS_STD_FMT::format(", {}", buffer_1[i]);
+  }


Again, it is up to you but I would use a range-for here 😉

mpusz · 2024-12-20T09:30:30Z

example/audio/1-oscillator.cpp

+  auto buffer_2 = buffer_t(buffer_1.size());
+  std::generate(begin(buffer_2), end(buffer_2), sin_gen);
+
+  return buffer_1 == buffer_2;


The error in the CI is strange, but #include <vector> is missing, so it might be the reason why the comparison operator is not found on one of the configurations.

Co-authored-by: Mateusz Pusz <[email protected]>

- Try including <vector> to fix CI error - Use `mp_units::inverse` - Use range-based for-loop - Use CTAG and `quantity` instead of `auto`

…mples

I was suprised by the need to use `.in(one)` here as I would have assumed the result was already in unit `one`.

…ture/audio-examples

…erload set

mpusz · 2025-01-09T08:28:46Z

example/audio/1-oscillator.cpp

+  const quantity buffer_size =
+    quantity_cast<audio::sample_count>((buffer_duration * context.current_sample_rate).in(one));


Is a quantity_cast required here? I assume that the implicit conversion did not compile? Did you try an explicit conversion?

Suggested change

const quantity buffer_size =

quantity_cast<audio::sample_count>((buffer_duration * context.current_sample_rate).in(one));

const quantity buffer_size =

audio::sample_count((buffer_duration * context.current_sample_rate).in(one));

Isn't audio::sample a good unit here?

Yes, audio::sample is a good unit here. This is what I had before the changes to dimensionless quantities:

const quantity buffer_size = (buffer_duration * context.current_sample_rate).in(audio::sample);

Maybe I didn't know the best way to do this after the dimensionless changes.

I've fixed the previous audio example in the paper (https://godbolt.org/z/z7sPqe66e) and improved inverse(). Maybe this will give you some ideas on how to refactor this.

…mples

The audio examples were including the deprecated mp-units/format.h header which generates deprecation warnings that are treated as errors with -Werror. Removed the deprecated includes and ensured proper headers are included. 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <[email protected]>

Changes MP_UNITS_CONSTEVAL to constexpr for unit multiplication, division, and `inverse` operations in unit.h. This was discovered with feedback on PR mpusz#644 to use `inverse`. Specifically fixes compilation error with clang (and likely other compilers?) where consteval functions cannot be called in non-constant contexts, such as: ```c++ m_frequency = inverse<si::hertz>(period); ``` Added runtime tests for `inverse()` function covering time-to-frequency conversion, runtime parameters, and unit conversion.

…mples

mpusz · 2025-12-21T21:30:02Z

@rothmichaels Do you plan to work on those a bit more? They need at least pre-commit run for styling and also there are a lot of unresolved threads here.

…mples

mpusz reviewed Nov 19, 2024

View reviewed changes

rothmichaels added 4 commits December 19, 2024 16:11

WIP LFO implementation

2a9265d

Refactor LFO example and get to compile

39acaff

Split out audio units and musical context functions into individual headers to be used in future examples in addition to the LFO example.

Add 0-wrapping_unsafe_apis audio example

f4fe2bd

Add an example of wrapping an unsafe OS or host application API with a type-safe quanitites API. The `audio::get_musical_context` API used in this example will also be used in other audio examples.

Finish LFO example

aec61bc

rothmichaels force-pushed the feature/audio-examples branch from b0fbe94 to aec61bc Compare December 20, 2024 00:28

Uncomment unit symbols

fbfecdb

rothmichaels commented Dec 20, 2024

View reviewed changes

example/audio/audio.h Show resolved Hide resolved

rothmichaels added 5 commits December 19, 2024 19:48

Include <algorithm> in 1-lfo example

8176bdc

Try value_cast<float> to fix int->float conversion warning

50a323b

Fix reuse of sample_rate name for analysis

b0a6ac9

Try adding more value_cast to fix analysis errors

9629004

Rename 1-lfo -> 1-oscillator

8be556f

mpusz requested changes Dec 20, 2024

View reviewed changes

rothmichaels and others added 9 commits December 20, 2024 10:30

Update example/audio/third_party_audio_api.h

2e0a824

Co-authored-by: Mateusz Pusz <[email protected]>

Update example/audio/wrapped_third_party_audio_api.h

9fbe556

Co-authored-by: Mateusz Pusz <[email protected]>

Audio examples 1-oscillator PR feedback

03775fd

- Try including <vector> to fix CI error - Use `mp_units::inverse` - Use range-based for-loop - Use CTAG and `quantity` instead of `auto`

Merge remote-tracking branch 'upstream/master' into feature/audio-exa…

b790489

…mples

Update audio example for dimensionless changes

e5a5948

I was suprised by the need to use `.in(one)` here as I would have assumed the result was already in unit `one`.

Remove unused header

7091c1e

Merge remote-tracking branch 'origin/feature/audio-examples' into fea…

8223e15

…ture/audio-examples

Add missing call to update_step in 1-oscillator example

607d1ea

1-oscillator example combine set_frequency and set_period into one ov…

acf4d8d

…erload set

mpusz reviewed Jan 9, 2025

View reviewed changes

rothmichaels and others added 3 commits May 6, 2025 03:52

Merge remote-tracking branch 'upstream/master' into feature/audio-exa…

52f7b88

…mples

Merge remote-tracking branch 'upstream/master' into feature/audio-exa…

afacfb2

…mples

This was referenced Sep 26, 2025

Replace MP_UNITS_CONSTEVAL with constexpr in unit operations #725

Closed

fix: enable runtime usage of inverse function by using compile-time unit extraction #726

Merged

Merge remote-tracking branch 'upstream/master' into feature/audio-exa…

f58981d

…mples

rothmichaels force-pushed the feature/audio-examples branch from ca795f2 to f58981d Compare November 11, 2025 09:24

Merge branch 'master' into feature/audio-examples

936a414

Merge remote-tracking branch 'upstream/master' into feature/audio-exa…

947fbee

…mples

		const quantity buffer_size =
		quantity_cast<audio::sample_count>((buffer_duration * context.current_sample_rate).in(one));

Uh oh!

Audio examples #644

Are you sure you want to change the base?

Audio examples #644

Uh oh!

Conversation

rothmichaels commented Nov 19, 2024

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rothmichaels Dec 20, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mpusz commented Dec 21, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

rothmichaels Dec 20, 2024 •

edited

Loading