validation_assertions.hpp File Reference

Framework-agnostic correctness checks for plugin/example validation tests. More...

#include <pulp/audio/buffer.hpp>
#include <pulp/format/headless.hpp>
#include <pulp/midi/buffer.hpp>
#include <pulp/midi/ump_buffer.hpp>
#include <pulp/state/parameter.hpp>
#include <pulp/state/store.hpp>
#include <algorithm>
#include <cmath>
#include <cstdint>
#include <span>
#include <string>
#include <utility>
#include <vector>

Go to the source code of this file.

Classes
struct	pulp::format::validation::CheckResult
	Outcome of a single validation check. More...

Namespaces
namespace	pulp
namespace	pulp::format
	What the product grants a device: the quota table indexed by capability tier and live thermal state.
namespace	pulp::format::validation

Functions
CheckResult	pulp::format::validation::check_finite (std::span< const float > samples)
	True when no sample is NaN or infinite.
template<typename Sample>
CheckResult	pulp::format::validation::check_finite (const audio::BufferView< Sample > &buffer)
	True when no sample is NaN or infinite.
CheckResult	pulp::format::validation::check_any_nonzero (std::span< const float > samples, float eps=1.0e-7f)
	True when at least one sample's magnitude exceeds eps — i.e.
CheckResult	pulp::format::validation::check_silent (std::span< const float > samples, float eps=1.0e-7f)
	True when every sample's magnitude is at or below eps — the inverse of check_any_nonzero, for "bypass with zero input must be silent" style tests.
CheckResult	pulp::format::validation::check_peak_below (std::span< const float > samples, float peak_limit)
	True when every sample's magnitude is at or below peak_limit (linear, not dB).
CheckResult	pulp::format::validation::check_param_round_trip (const state::ParamRange &range, float raw, float tol=1.0e-4f)
	Verify a parameter's normalize↔denormalize map is correct for raw.
CheckResult	pulp::format::validation::check_state_round_trip (HeadlessHost &host)
	Verify plugin state actually round-trips: save a snapshot, perturb every parameter to a different value, restore the snapshot, and require the re-saved bytes to match the original.
CheckResult	pulp::format::validation::check_sysex_payload_equal (std::span< const uint8_t > expected, std::span< const uint8_t > actual)
	True when two SysEx payloads carry identical bytes.
CheckResult	pulp::format::validation::check_midi_events_equal (const midi::MidiBuffer &expected, const midi::MidiBuffer &actual)
	True when two MIDI buffers carry the same short messages (bytes + sample offset, in order), the same SysEx payloads (bytes + sample offset), and the same UMP (MIDI 2.0) sidecar packets.

Detailed Description

Framework-agnostic correctness checks for plugin/example validation tests.

These helpers cover the recurring "did the DSP/state/MIDI actually behave" questions that every example test re-implements by hand: output finiteness, silence/non-silence, peak bounds, parameter normalize↔denormalize stability, state save/load determinism, and MIDI/SysEx equality.

Each helper returns a CheckResult (a bool plus a human-readable message) so it composes with any test framework — Catch2, GoogleTest, or a bare if (!result) { ... }. The header intentionally does NOT depend on a test framework, so it ships with the SDK and is reusable by downstream example projects that consume the installed Pulp headers:

auto out = harness.process_buffer(input, 2, 512);
REQUIRE(pulp::format::validation::check_finite(out));
REQUIRE(pulp::format::validation::check_any_nonzero(out));
REQUIRE(pulp::format::validation::check_state_round_trip(harness.host()));

Screenshot content-floor checks are intentionally NOT duplicated here — use pulp::view::ScreenshotContentStats::passes_content_floor() (which already exists in core/view) so the format layer keeps no dependency on the view layer.