Implement M1 bank_model: Sample + BankIndex, hash-dedup, tiers, JSON round-trip

Pure dependency-free core with self-contained JSON writer/parser. Absolute
paths (incl. drive-relative) rejected at the add boundary; \uXXXX decoded to
UTF-8 with surrogate pairs; malformed input returns nullopt.
This commit is contained in:
2026-07-21 22:54:21 -04:00
parent 6fcb811668
commit 2862e1c865
3 changed files with 1214 additions and 16 deletions
+686 -3
View File
@@ -1,10 +1,693 @@
#include "bank_model.h"
#include <cctype>
#include <cerrno>
#include <cstdio>
#include <cstdlib>
#include <cstring>
// bank_model implementation.
//
// JSON is hand-rolled and self-contained (brief: keep the pure core
// dependency-free — no third-party JSON lib, no WDL coupling). The field set is
// a flat struct of primitives, strings, one enum, a small string array, and a
// few optionals, so a compact writer + recursive-descent parser is the simplest
// thing that works. Doubles are emitted with 17 significant digits (%.17g), the
// shortest form that round-trips every IEEE-754 double exactly, so the
// deserialize(serialize(x)) == x invariant holds bit-for-bit.
namespace reasampler {
int bankModelVersion()
{
return 1;
// ---------------------------------------------------------------------------
// equality
// ---------------------------------------------------------------------------
bool SourceRange::operator==(const SourceRange& o) const {
return startSeconds == o.startSeconds && endSeconds == o.endSeconds &&
startPpq == o.startPpq && endPpq == o.endPpq;
}
bool Provenance::operator==(const Provenance& o) const {
return parentSampleId == o.parentSampleId && fxChainSnapshot == o.fxChainSnapshot;
}
bool Levels::operator==(const Levels& o) const {
return peakDb == o.peakDb && rmsDb == o.rmsDb && lufs == o.lufs;
}
bool Sample::operator==(const Sample& o) const {
return id == o.id && displayName == o.displayName && relativePath == o.relativePath &&
sourceMode == o.sourceMode && sourceRange == o.sourceRange &&
trackGuids == o.trackGuids && wetDry == o.wetDry &&
channelCount == o.channelCount && sampleRate == o.sampleRate &&
lengthSeconds == o.lengthSeconds && lengthBeats == o.lengthBeats &&
captureTempo == o.captureTempo && key == o.key && levels == o.levels &&
clipped == o.clipped && tier == o.tier && contentHash == o.contentHash &&
provenance == o.provenance && createdTimestamp == o.createdTimestamp;
}
// ---------------------------------------------------------------------------
// path invariant
// ---------------------------------------------------------------------------
// DECISION: reject absolute paths rather than normalize them. The pure model has
// no knowledge of the project root, so it cannot correctly relativize an absolute
// path — any "normalization" would be a guess that could point at the wrong file.
// Rejecting at the boundary is honest and deterministic; the capture backend (M3)
// is responsible for handing us an already-relative path. Covers POSIX ("/x"),
// Windows drive ("C:\x", "C:/x", "C:foo" drive-relative), and UNC ("\\host\share")
// forms. Any leading <alpha>: is rejected regardless of the character that follows —
// drive-relative paths ("C:foo.wav") resolve against the drive's current directory,
// not the project root, so they violate the relative-paths-only invariant just as
// much as "C:\foo.wav" does.
static bool isAbsolutePath(const std::string& p) {
if (p.empty()) return false;
if (p[0] == '/' || p[0] == '\\') return true; // POSIX root or UNC
if (p.size() >= 2 && std::isalpha(static_cast<unsigned char>(p[0])) && p[1] == ':')
return true; // Windows drive (C:\, C:/, C:foo, C:)
return false;
}
// ---------------------------------------------------------------------------
// BankIndex
// ---------------------------------------------------------------------------
AddResult BankIndex::add(const Sample& sample) {
if (sample.id.empty()) return AddResult::RejectedEmptyId;
if (isAbsolutePath(sample.relativePath)) return AddResult::RejectedAbsolutePath;
if (findByHash(sample.contentHash) != nullptr)
return AddResult::Collapsed;
samples_.push_back(sample);
return AddResult::Added;
}
bool BankIndex::remove(const std::string& id) {
for (auto it = samples_.begin(); it != samples_.end(); ++it) {
if (it->id == id) {
samples_.erase(it);
return true;
}
}
return false;
}
const Sample* BankIndex::query(const std::string& id) const {
for (const auto& s : samples_)
if (s.id == id) return &s;
return nullptr;
}
const Sample* BankIndex::findByHash(const std::string& contentHash) const {
if (contentHash.empty()) return nullptr; // empty hashes never dedup
for (const auto& s : samples_)
if (s.contentHash == contentHash) return &s;
return nullptr;
}
bool BankIndex::moveTier(const std::string& id, Tier tier) {
for (auto& s : samples_) {
if (s.id == id) {
s.tier = tier;
return true;
}
}
return false;
}
std::vector<Sample> BankIndex::byTier(Tier tier) const {
std::vector<Sample> out;
for (const auto& s : samples_)
if (s.tier == tier) out.push_back(s);
return out;
}
// ---------------------------------------------------------------------------
// JSON writer
// ---------------------------------------------------------------------------
namespace {
void writeEscaped(std::string& out, const std::string& s) {
out += '"';
for (char c : s) {
switch (c) {
case '"': out += "\\\""; break;
case '\\': out += "\\\\"; break;
case '\b': out += "\\b"; break;
case '\f': out += "\\f"; break;
case '\n': out += "\\n"; break;
case '\r': out += "\\r"; break;
case '\t': out += "\\t"; break;
default:
if (static_cast<unsigned char>(c) < 0x20) {
char buf[8];
std::snprintf(buf, sizeof(buf), "\\u%04x", static_cast<unsigned char>(c));
out += buf;
} else {
out += c;
}
}
}
out += '"';
}
std::string numToStr(double v) {
char buf[32];
std::snprintf(buf, sizeof(buf), "%.17g", v);
return buf;
}
std::string numToStr(std::int64_t v) {
char buf[32];
std::snprintf(buf, sizeof(buf), "%lld", static_cast<long long>(v));
return buf;
}
std::string numToStr(int v) { return numToStr(static_cast<std::int64_t>(v)); }
class ObjWriter {
public:
explicit ObjWriter(std::string& out) : out_(out) { out_ += '{'; }
~ObjWriter() { out_ += '}'; }
void keyRaw(const char* key, const std::string& rawValue) {
sep();
writeEscaped(out_, key);
out_ += ':';
out_ += rawValue;
}
void keyStr(const char* key, const std::string& value) {
sep();
writeEscaped(out_, key);
out_ += ':';
writeEscaped(out_, value);
}
// Begin a nested value; caller writes the value immediately after.
void keyBegin(const char* key) {
sep();
writeEscaped(out_, key);
out_ += ':';
}
private:
void sep() {
if (first_) first_ = false; else out_ += ',';
}
std::string& out_;
bool first_ = true;
};
void writeStringArray(std::string& out, const std::vector<std::string>& v) {
out += '[';
for (std::size_t i = 0; i < v.size(); ++i) {
if (i) out += ',';
writeEscaped(out, v[i]);
}
out += ']';
}
void writeSample(std::string& out, const Sample& s) {
ObjWriter w(out);
w.keyStr("id", s.id);
w.keyStr("displayName", s.displayName);
w.keyStr("relativePath", s.relativePath);
w.keyRaw("sourceMode", numToStr(static_cast<int>(s.sourceMode)));
w.keyBegin("sourceRange");
{
ObjWriter r(out);
r.keyRaw("startSeconds", numToStr(s.sourceRange.startSeconds));
r.keyRaw("endSeconds", numToStr(s.sourceRange.endSeconds));
r.keyRaw("startPpq", numToStr(s.sourceRange.startPpq));
r.keyRaw("endPpq", numToStr(s.sourceRange.endPpq));
}
w.keyBegin("trackGuids");
writeStringArray(out, s.trackGuids);
w.keyRaw("wetDry", numToStr(s.wetDry));
w.keyRaw("channelCount", numToStr(s.channelCount));
w.keyRaw("sampleRate", numToStr(s.sampleRate));
w.keyRaw("lengthSeconds", numToStr(s.lengthSeconds));
w.keyRaw("lengthBeats", numToStr(s.lengthBeats));
w.keyRaw("captureTempo", numToStr(s.captureTempo));
// Optionals are emitted as null when absent so present/absent round-trips.
w.keyBegin("key");
if (s.key) writeEscaped(out, *s.key); else out += "null";
w.keyBegin("levels");
{
ObjWriter l(out);
l.keyRaw("peakDb", numToStr(s.levels.peakDb));
l.keyRaw("rmsDb", numToStr(s.levels.rmsDb));
l.keyRaw("lufs", numToStr(s.levels.lufs));
}
w.keyRaw("clipped", s.clipped ? "true" : "false");
w.keyRaw("tier", numToStr(static_cast<int>(s.tier)));
w.keyStr("contentHash", s.contentHash);
w.keyBegin("provenance");
if (s.provenance) {
ObjWriter p(out);
p.keyStr("parentSampleId", s.provenance->parentSampleId);
p.keyStr("fxChainSnapshot", s.provenance->fxChainSnapshot);
} else {
out += "null";
}
w.keyRaw("createdTimestamp", numToStr(s.createdTimestamp));
}
} // namespace
std::string BankIndex::serialize() const {
std::string out;
{
ObjWriter root(out);
root.keyRaw("version", numToStr(1));
root.keyBegin("samples");
out += '[';
for (std::size_t i = 0; i < samples_.size(); ++i) {
if (i) out += ',';
writeSample(out, samples_[i]);
}
out += ']';
} // root closes the object here — not deferred to function return (NRVO would
// otherwise let the caller observe `out` before the closing brace is appended)
return out;
}
// ---------------------------------------------------------------------------
// JSON parser (recursive descent). Returns false on any malformed input; never
// reads out of bounds. Only supports the subset our writer emits.
// ---------------------------------------------------------------------------
namespace {
class Parser {
public:
explicit Parser(const std::string& s) : s_(s) {}
bool parseIndex(BankIndex& out);
private:
const std::string& s_;
std::size_t pos_ = 0;
bool eof() const { return pos_ >= s_.size(); }
char peek() const { return s_[pos_]; }
void skipWs() {
while (!eof()) {
char c = s_[pos_];
if (c == ' ' || c == '\t' || c == '\n' || c == '\r') ++pos_;
else break;
}
}
bool consume(char c) {
skipWs();
if (eof() || s_[pos_] != c) return false;
++pos_;
return true;
}
bool parseString(std::string& out);
bool parseRawScalar(std::string& out); // number / true / false / null token
bool parseDouble(double& out);
bool parseInt64(std::int64_t& out);
bool parseInt(int& out);
bool parseBool(bool& out);
bool expectNullOr(bool& wasNull); // peeks for `null`; consumes if present
bool parseSample(Sample& out);
bool parseKey(std::string& key); // an object member key + ':'
bool skipValue(); // for forward-compat unknown keys
};
// Parses a JSON string literal (with the escapes our writer emits, plus \uXXXX
// for control chars). Positioned at the opening quote after whitespace.
bool Parser::parseString(std::string& out) {
skipWs();
if (eof() || s_[pos_] != '"') return false;
++pos_;
out.clear();
while (!eof()) {
char c = s_[pos_++];
if (c == '"') return true;
if (c == '\\') {
if (eof()) return false;
char e = s_[pos_++];
switch (e) {
case '"': out += '"'; break;
case '\\': out += '\\'; break;
case '/': out += '/'; break;
case 'b': out += '\b'; break;
case 'f': out += '\f'; break;
case 'n': out += '\n'; break;
case 'r': out += '\r'; break;
case 't': out += '\t'; break;
case 'u': {
// Decode a \uXXXX escape to its code point.
auto readHex4 = [&](unsigned int& cp) -> bool {
if (pos_ + 4 > s_.size()) return false;
cp = 0;
for (int i = 0; i < 4; ++i) {
char h = s_[pos_++];
cp <<= 4;
if (h >= '0' && h <= '9') cp |= static_cast<unsigned>(h - '0');
else if (h >= 'a' && h <= 'f') cp |= static_cast<unsigned>(h - 'a' + 10);
else if (h >= 'A' && h <= 'F') cp |= static_cast<unsigned>(h - 'A' + 10);
else return false;
}
return true;
};
unsigned int hi = 0;
if (!readHex4(hi)) return false;
unsigned int codePoint = hi;
if (hi >= 0xD800 && hi <= 0xDBFF) {
// High surrogate — must be followed by \uDC00\uDFFF.
if (pos_ + 6 > s_.size()) return false;
if (s_[pos_] != '\\' || s_[pos_ + 1] != 'u') return false;
pos_ += 2;
unsigned int lo = 0;
if (!readHex4(lo)) return false;
if (lo < 0xDC00 || lo > 0xDFFF) return false; // unpaired high surrogate
codePoint = 0x10000 + ((hi - 0xD800) << 10) + (lo - 0xDC00);
} else if (hi >= 0xDC00 && hi <= 0xDFFF) {
return false; // unpaired low surrogate — malformed
}
// Encode codePoint as UTF-8.
if (codePoint <= 0x7F) {
out += static_cast<char>(codePoint);
} else if (codePoint <= 0x7FF) {
out += static_cast<char>(0xC0 | (codePoint >> 6));
out += static_cast<char>(0x80 | (codePoint & 0x3F));
} else if (codePoint <= 0xFFFF) {
out += static_cast<char>(0xE0 | (codePoint >> 12));
out += static_cast<char>(0x80 | ((codePoint >> 6) & 0x3F));
out += static_cast<char>(0x80 | (codePoint & 0x3F));
} else {
out += static_cast<char>(0xF0 | (codePoint >> 18));
out += static_cast<char>(0x80 | ((codePoint >> 12) & 0x3F));
out += static_cast<char>(0x80 | ((codePoint >> 6) & 0x3F));
out += static_cast<char>(0x80 | (codePoint & 0x3F));
}
break;
}
default: return false;
}
} else {
out += c;
}
}
return false; // unterminated string
}
// Reads a bare token (number, true, false, null) up to the next structural char.
bool Parser::parseRawScalar(std::string& out) {
skipWs();
std::size_t start = pos_;
while (!eof()) {
char c = s_[pos_];
if (c == ',' || c == '}' || c == ']' || c == ' ' || c == '\t' ||
c == '\n' || c == '\r')
break;
++pos_;
}
if (pos_ == start) return false;
out.assign(s_, start, pos_ - start);
return true;
}
bool Parser::parseDouble(double& out) {
std::string tok;
if (!parseRawScalar(tok)) return false;
const char* b = tok.c_str();
char* end = nullptr;
errno = 0;
double v = std::strtod(b, &end);
if (end != b + tok.size()) return false;
if (errno == ERANGE) return false; // overflow / underflow → malformed
out = v;
return true;
}
bool Parser::parseInt64(std::int64_t& out) {
std::string tok;
if (!parseRawScalar(tok)) return false;
const char* b = tok.c_str();
char* end = nullptr;
errno = 0;
long long v = std::strtoll(b, &end, 10);
if (end != b + tok.size()) return false;
if (errno == ERANGE) return false; // overflow → malformed
out = static_cast<std::int64_t>(v);
return true;
}
bool Parser::parseInt(int& out) {
std::int64_t v = 0;
if (!parseInt64(v)) return false;
out = static_cast<int>(v);
return true;
}
bool Parser::parseBool(bool& out) {
std::string tok;
if (!parseRawScalar(tok)) return false;
if (tok == "true") { out = true; return true; }
if (tok == "false") { out = false; return true; }
return false;
}
// If the next value is the `null` token, consumes it and sets wasNull=true.
// Otherwise leaves the position untouched and sets wasNull=false. Returns false
// only on eof.
bool Parser::expectNullOr(bool& wasNull) {
skipWs();
if (eof()) return false;
if (s_.compare(pos_, 4, "null") == 0) {
pos_ += 4;
wasNull = true;
} else {
wasNull = false;
}
return true;
}
bool Parser::parseKey(std::string& key) {
if (!parseString(key)) return false;
if (!consume(':')) return false;
return true;
}
// Skips one JSON value (object / array / string / scalar) for forward-compat
// with keys we don't recognize. Assumes position is at the start of the value.
bool Parser::skipValue() {
skipWs();
if (eof()) return false;
char c = s_[pos_];
if (c == '"') {
std::string tmp;
return parseString(tmp);
}
if (c == '{' || c == '[') {
char open = c, close = (c == '{') ? '}' : ']';
++pos_;
int depth = 1;
while (!eof() && depth > 0) {
char d = s_[pos_];
if (d == '"') {
std::string tmp;
if (!parseString(tmp)) return false;
continue;
}
if (d == open) ++depth;
else if (d == close) --depth;
++pos_;
}
return depth == 0;
}
std::string tmp;
return parseRawScalar(tmp);
}
bool Parser::parseSample(Sample& s) {
if (!consume('{')) return false;
skipWs();
if (consume('}')) return true; // empty object (shouldn't happen, but valid)
do {
std::string key;
if (!parseKey(key)) return false;
if (key == "id") {
if (!parseString(s.id)) return false;
} else if (key == "displayName") {
if (!parseString(s.displayName)) return false;
} else if (key == "relativePath") {
if (!parseString(s.relativePath)) return false;
} else if (key == "sourceMode") {
int v = 0;
if (!parseInt(v)) return false;
// Valid range: MasterMix(0) .. Realtime(5).
if (v < static_cast<int>(SourceMode::MasterMix) ||
v > static_cast<int>(SourceMode::Realtime))
return false;
s.sourceMode = static_cast<SourceMode>(v);
} else if (key == "sourceRange") {
if (!consume('{')) return false;
do {
std::string rk;
if (!parseKey(rk)) return false;
double dv = 0.0;
if (!parseDouble(dv)) return false;
if (rk == "startSeconds") s.sourceRange.startSeconds = dv;
else if (rk == "endSeconds") s.sourceRange.endSeconds = dv;
else if (rk == "startPpq") s.sourceRange.startPpq = dv;
else if (rk == "endPpq") s.sourceRange.endPpq = dv;
} while (consume(','));
if (!consume('}')) return false;
} else if (key == "trackGuids") {
if (!consume('[')) return false;
skipWs();
if (!consume(']')) {
do {
std::string g;
if (!parseString(g)) return false;
s.trackGuids.push_back(g);
} while (consume(','));
if (!consume(']')) return false;
}
} else if (key == "wetDry") {
if (!parseDouble(s.wetDry)) return false;
} else if (key == "channelCount") {
if (!parseInt(s.channelCount)) return false;
} else if (key == "sampleRate") {
if (!parseInt(s.sampleRate)) return false;
} else if (key == "lengthSeconds") {
if (!parseDouble(s.lengthSeconds)) return false;
} else if (key == "lengthBeats") {
if (!parseDouble(s.lengthBeats)) return false;
} else if (key == "captureTempo") {
if (!parseDouble(s.captureTempo)) return false;
} else if (key == "key") {
bool wasNull = false;
if (!expectNullOr(wasNull)) return false;
if (wasNull) {
s.key.reset();
} else {
std::string k;
if (!parseString(k)) return false;
s.key = k;
}
} else if (key == "levels") {
if (!consume('{')) return false;
do {
std::string lk;
if (!parseKey(lk)) return false;
double dv = 0.0;
if (!parseDouble(dv)) return false;
if (lk == "peakDb") s.levels.peakDb = dv;
else if (lk == "rmsDb") s.levels.rmsDb = dv;
else if (lk == "lufs") s.levels.lufs = dv;
} while (consume(','));
if (!consume('}')) return false;
} else if (key == "clipped") {
if (!parseBool(s.clipped)) return false;
} else if (key == "tier") {
int v = 0;
if (!parseInt(v)) return false;
// Valid range: Scratch(0) .. Archive(1).
if (v < static_cast<int>(Tier::Scratch) ||
v > static_cast<int>(Tier::Archive))
return false;
s.tier = static_cast<Tier>(v);
} else if (key == "contentHash") {
if (!parseString(s.contentHash)) return false;
} else if (key == "provenance") {
bool wasNull = false;
if (!expectNullOr(wasNull)) return false;
if (wasNull) {
s.provenance.reset();
} else {
if (!consume('{')) return false;
Provenance p;
do {
std::string pk;
if (!parseKey(pk)) return false;
std::string pv;
if (!parseString(pv)) return false;
if (pk == "parentSampleId") p.parentSampleId = pv;
else if (pk == "fxChainSnapshot") p.fxChainSnapshot = pv;
} while (consume(','));
if (!consume('}')) return false;
s.provenance = p;
}
} else if (key == "createdTimestamp") {
if (!parseInt64(s.createdTimestamp)) return false;
} else {
if (!skipValue()) return false; // forward-compat: ignore unknown
}
} while (consume(','));
return consume('}');
}
bool Parser::parseIndex(BankIndex& out) {
if (!consume('{')) return false;
skipWs();
if (consume('}')) return true; // empty object — vacuously an empty index
std::vector<Sample> parsed;
do {
std::string key;
if (!parseKey(key)) return false;
if (key == "samples") {
if (!consume('[')) return false;
skipWs();
if (!consume(']')) {
do {
Sample s;
if (!parseSample(s)) return false;
parsed.push_back(std::move(s));
} while (consume(','));
if (!consume(']')) return false;
}
} else {
if (!skipValue()) return false; // version, or unknown keys
}
} while (consume(','));
if (!consume('}')) return false;
// Trailing garbage after the root object is malformed.
skipWs();
if (!eof()) return false;
// Rebuild via add() so the same invariants (relative-path, dedup) that guard
// live inserts also guard deserialized data. Rejected/collapsed entries are
// dropped silently — a well-formed serialized index never triggers them.
for (auto& s : parsed) out.add(s);
return true;
}
} // namespace
std::optional<BankIndex> BankIndex::deserialize(const std::string& json) {
BankIndex idx;
Parser p(json);
if (!p.parseIndex(idx)) return std::nullopt;
return idx;
}
} // namespace reasampler