| 1 | //===----------------------------------------------------------------------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "llvm/CAS/NamedValuesSchema.h" |
| 10 | #include "llvm/Support/Endian.h" |
| 11 | #include "llvm/Support/EndianStream.h" |
| 12 | #include "llvm/Support/StringSaver.h" |
| 13 | |
| 14 | using namespace llvm; |
| 15 | using namespace llvm::cas; |
| 16 | |
| 17 | char NamedValuesSchema::ID = 0; |
| 18 | constexpr StringLiteral NamedValuesSchema::SchemaName; |
| 19 | |
| 20 | void NamedValuesSchema::anchor() {} |
| 21 | |
| 22 | bool NamedValuesSchema::isNode(const ObjectProxy &Node) const { |
| 23 | // Load the first ref to check its content. |
| 24 | if (Node.getNumReferences() < 1) |
| 25 | return false; |
| 26 | |
| 27 | auto FirstRef = Node.getReference(I: 0); |
| 28 | return FirstRef == *NamedValuesKindRef; |
| 29 | } |
| 30 | |
| 31 | NamedValuesSchema::NamedValuesSchema(cas::ObjectStore &CAS, Error &E) |
| 32 | : NamedValuesSchema::RTTIExtends(CAS) { |
| 33 | ErrorAsOutParameter EAOP(E); |
| 34 | auto Kind = CAS.storeFromString(Refs: {}, String: SchemaName); |
| 35 | if (!Kind) { |
| 36 | E = Kind.takeError(); |
| 37 | return; |
| 38 | } |
| 39 | NamedValuesKindRef = *Kind; |
| 40 | } |
| 41 | |
| 42 | Expected<NamedValuesSchema> NamedValuesSchema::create(ObjectStore &CAS) { |
| 43 | Error E = Error::success(); |
| 44 | NamedValuesSchema S(CAS, E); |
| 45 | if (E) |
| 46 | return std::move(E); |
| 47 | return S; |
| 48 | } |
| 49 | |
| 50 | size_t NamedValuesSchema::getNumEntries(NamedValuesProxy Values) const { |
| 51 | return Values.getNumReferences() - 1; |
| 52 | } |
| 53 | |
| 54 | Error NamedValuesSchema::forEachEntry( |
| 55 | NamedValuesProxy Values, |
| 56 | function_ref<Error(const NamedValuesEntry &)> Callback) const { |
| 57 | for (size_t I = 0, IE = getNumEntries(Values); I != IE; ++I) |
| 58 | if (Error E = Callback(loadEntry(Values, I))) |
| 59 | return E; |
| 60 | |
| 61 | return Error::success(); |
| 62 | } |
| 63 | |
| 64 | NamedValuesEntry NamedValuesSchema::loadEntry(NamedValuesProxy Values, |
| 65 | size_t I) const { |
| 66 | StringRef Name = Values.getName(I); |
| 67 | auto ObjectRef = Values.getReference(I: I + 1); |
| 68 | |
| 69 | return {Name, ObjectRef}; |
| 70 | } |
| 71 | |
| 72 | std::optional<size_t> NamedValuesSchema::lookupEntry(NamedValuesProxy Values, |
| 73 | StringRef Name) const { |
| 74 | size_t NumNames = getNumEntries(Values); |
| 75 | if (!NumNames) |
| 76 | return std::nullopt; |
| 77 | |
| 78 | // Start with a binary search, if there are enough entries. |
| 79 | // FIXME: MaxLinearSearchSize is a heuristic and not optimized. |
| 80 | const size_t MaxLinearSearchSize = 4; |
| 81 | size_t Last = NumNames; |
| 82 | size_t First = 0; |
| 83 | while (Last - First > MaxLinearSearchSize) { |
| 84 | auto I = First + (Last - First) / 2; |
| 85 | StringRef NameI = Values.getName(I); |
| 86 | switch (Name.compare(RHS: NameI)) { |
| 87 | case 0: |
| 88 | return I; |
| 89 | case -1: |
| 90 | Last = I; |
| 91 | break; |
| 92 | case 1: |
| 93 | First = I + 1; |
| 94 | break; |
| 95 | } |
| 96 | } |
| 97 | |
| 98 | // Use a linear search for small list. |
| 99 | for (; First != Last; ++First) |
| 100 | if (Name == Values.getName(I: First)) |
| 101 | return First; |
| 102 | |
| 103 | return std::nullopt; |
| 104 | } |
| 105 | |
| 106 | Expected<NamedValuesProxy> NamedValuesSchema::load(ObjectRef Object) const { |
| 107 | auto Node = CAS.getProxy(Ref: Object); |
| 108 | if (!Node) |
| 109 | return Node.takeError(); |
| 110 | |
| 111 | return load(Object: *Node); |
| 112 | } |
| 113 | |
| 114 | Expected<NamedValuesProxy> NamedValuesSchema::load(ObjectProxy Object) const { |
| 115 | if (!isNode(Node: Object)) |
| 116 | return createStringError(EC: inconvertibleErrorCode(), |
| 117 | S: "object does not conform to NamedValuesSchema" ); |
| 118 | |
| 119 | return NamedValuesProxy(*this, Object); |
| 120 | } |
| 121 | |
| 122 | Expected<NamedValuesProxy> |
| 123 | NamedValuesSchema::construct(ArrayRef<NamedValuesEntry> Entries) { |
| 124 | // ScratchPad for output. |
| 125 | SmallString<256> Data; |
| 126 | SmallVector<ObjectRef, 16> Refs; |
| 127 | Refs.push_back(Elt: *NamedValuesKindRef); |
| 128 | |
| 129 | // Ensure a stable order for entries and ignore name collisions. |
| 130 | SmallVector<NamedValuesEntry> Sorted(Entries); |
| 131 | llvm::stable_sort(Range&: Sorted); |
| 132 | |
| 133 | if (llvm::unique(R&: Sorted) != Sorted.end()) |
| 134 | return createStringError(Fmt: "entry names are not unique" ); |
| 135 | |
| 136 | raw_svector_ostream OS(Data); |
| 137 | support::endian::Writer Writer(OS, endianness::little); |
| 138 | // Encode the entries in the Data. The layout of the named values schema |
| 139 | // object is: |
| 140 | // * Name offset table: The offset of in the data blob for where to find the |
| 141 | // string. It has N + 1 entries and you can find the name of n-th entry at |
| 142 | // offset[n] -> offset[n+1]. Each offset is encoded as little-endian |
| 143 | // uint32_t. |
| 144 | // * Object: ObjectRef for each entry is at n + 1 refs for the object (with |
| 145 | // the first one being the named value kind ID). |
| 146 | |
| 147 | // Write Name. |
| 148 | // The start of the string table index. |
| 149 | uint32_t StrIdx = sizeof(uint32_t) * (Sorted.size() + 1); |
| 150 | for (auto &Entry : Sorted) { |
| 151 | Writer.write(Val: StrIdx); |
| 152 | StrIdx += Entry.Name.size(); |
| 153 | |
| 154 | // Append refs. |
| 155 | Refs.push_back(Elt: Entry.Ref); |
| 156 | } |
| 157 | // Write the end index for the last string. |
| 158 | Writer.write(Val: StrIdx); |
| 159 | |
| 160 | // Write names in the end of the block. |
| 161 | for (auto &Entry : Sorted) |
| 162 | OS << Entry.Name; |
| 163 | |
| 164 | auto Proxy = CAS.createProxy(Refs, Data); |
| 165 | if (!Proxy) |
| 166 | return Proxy.takeError(); |
| 167 | |
| 168 | return NamedValuesProxy(*this, *Proxy); |
| 169 | } |
| 170 | |
| 171 | void NamedValuesSchema::Builder::add(StringRef Name, ObjectRef Ref) { |
| 172 | StringSaver Saver(Alloc); |
| 173 | Nodes.emplace_back(Args: Saver.save(S: Name), Args&: Ref); |
| 174 | } |
| 175 | |
| 176 | Expected<NamedValuesProxy> NamedValuesSchema::Builder::build() { |
| 177 | auto Schema = NamedValuesSchema::create(CAS); |
| 178 | if (!Schema) |
| 179 | return Schema.takeError(); |
| 180 | return Schema->construct(Entries: Nodes); |
| 181 | } |
| 182 | |
| 183 | StringRef NamedValuesProxy::getName(size_t I) const { |
| 184 | uint32_t StartIdx = |
| 185 | support::endian::read32le(P: getData().data() + sizeof(uint32_t) * I); |
| 186 | uint32_t EndIdx = |
| 187 | support::endian::read32le(P: getData().data() + sizeof(uint32_t) * (I + 1)); |
| 188 | |
| 189 | return StringRef(getData().data() + StartIdx, EndIdx - StartIdx); |
| 190 | } |
| 191 | |