parse_tree.cpp 8.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272
  1. // Part of the Carbon Language project, under the Apache License v2.0 with LLVM
  2. // Exceptions. See /LICENSE for license information.
  3. // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
  4. #include "toolchain/parser/parse_tree.h"
  5. #include <cstdlib>
  6. #include "common/check.h"
  7. #include "llvm/ADT/ArrayRef.h"
  8. #include "llvm/ADT/Optional.h"
  9. #include "llvm/ADT/Sequence.h"
  10. #include "llvm/ADT/SmallSet.h"
  11. #include "llvm/ADT/SmallVector.h"
  12. #include "llvm/ADT/iterator.h"
  13. #include "llvm/Support/raw_ostream.h"
  14. #include "toolchain/lexer/token_kind.h"
  15. #include "toolchain/parser/parse_node_kind.h"
  16. #include "toolchain/parser/parser2.h"
  17. #include "toolchain/parser/parser_impl.h"
  18. namespace Carbon {
  19. auto ParseTree::Parse(TokenizedBuffer& tokens, DiagnosticConsumer& consumer)
  20. -> ParseTree {
  21. TokenizedBuffer::TokenLocationTranslator translator(
  22. tokens, /*last_line_lexed_to_column=*/nullptr);
  23. TokenDiagnosticEmitter emitter(translator, consumer);
  24. // Delegate to the parser.
  25. // TODO: Edit this to swap between Parser and Parser2. This is manual in order
  26. // to avoid test duplication.
  27. return Parser::Parse(tokens, emitter);
  28. }
  29. auto ParseTree::postorder() const -> llvm::iterator_range<PostorderIterator> {
  30. return {PostorderIterator(Node(0)),
  31. PostorderIterator(Node(node_impls_.size()))};
  32. }
  33. auto ParseTree::postorder(Node n) const
  34. -> llvm::iterator_range<PostorderIterator> {
  35. CARBON_CHECK(n.is_valid());
  36. // The postorder ends after this node, the root, and begins at the start of
  37. // its subtree.
  38. int end_index = n.index_ + 1;
  39. int start_index = end_index - node_impls_[n.index_].subtree_size;
  40. return {PostorderIterator(Node(start_index)),
  41. PostorderIterator(Node(end_index))};
  42. }
  43. auto ParseTree::children(Node n) const
  44. -> llvm::iterator_range<SiblingIterator> {
  45. CARBON_CHECK(n.is_valid());
  46. int end_index = n.index_ - node_impls_[n.index_].subtree_size;
  47. return {SiblingIterator(*this, Node(n.index_ - 1)),
  48. SiblingIterator(*this, Node(end_index))};
  49. }
  50. auto ParseTree::roots() const -> llvm::iterator_range<SiblingIterator> {
  51. return {
  52. SiblingIterator(*this, Node(static_cast<int>(node_impls_.size()) - 1)),
  53. SiblingIterator(*this, Node(-1))};
  54. }
  55. auto ParseTree::node_has_error(Node n) const -> bool {
  56. CARBON_CHECK(n.is_valid());
  57. return node_impls_[n.index_].has_error;
  58. }
  59. auto ParseTree::node_kind(Node n) const -> ParseNodeKind {
  60. CARBON_CHECK(n.is_valid());
  61. return node_impls_[n.index_].kind;
  62. }
  63. auto ParseTree::node_token(Node n) const -> TokenizedBuffer::Token {
  64. CARBON_CHECK(n.is_valid());
  65. return node_impls_[n.index_].token;
  66. }
  67. auto ParseTree::node_subtree_size(Node n) const -> int32_t {
  68. CARBON_CHECK(n.is_valid());
  69. return node_impls_[n.index_].subtree_size;
  70. }
  71. auto ParseTree::GetNodeText(Node n) const -> llvm::StringRef {
  72. CARBON_CHECK(n.is_valid());
  73. return tokens_->GetTokenText(node_impls_[n.index_].token);
  74. }
  75. auto ParseTree::PrintNode(llvm::raw_ostream& output, Node n, int depth,
  76. bool preorder) const -> bool {
  77. const auto& n_impl = node_impls_[n.index()];
  78. output.indent(2 * depth);
  79. output << "{";
  80. // If children are being added, include node_index in order to disambiguate
  81. // nodes.
  82. if (preorder) {
  83. output << "node_index: " << n.index_ << ", ";
  84. }
  85. output << "kind: '" << n_impl.kind.name() << "', text: '"
  86. << tokens_->GetTokenText(n_impl.token) << "'";
  87. if (n_impl.has_error) {
  88. output << ", has_error: yes";
  89. }
  90. if (n_impl.subtree_size > 1) {
  91. output << ", subtree_size: " << n_impl.subtree_size;
  92. if (preorder) {
  93. output << ", children: [\n";
  94. return true;
  95. }
  96. }
  97. output << "}";
  98. return false;
  99. }
  100. auto ParseTree::Print(llvm::raw_ostream& output) const -> void {
  101. // Walk the tree just to calculate depths for each node.
  102. llvm::SmallVector<int> indents;
  103. indents.append(size(), 0);
  104. llvm::SmallVector<std::pair<Node, int>, 16> node_stack;
  105. for (Node n : roots()) {
  106. node_stack.push_back({n, 0});
  107. }
  108. while (!node_stack.empty()) {
  109. Node n;
  110. int depth;
  111. std::tie(n, depth) = node_stack.pop_back_val();
  112. for (Node sibling_n : children(n)) {
  113. indents[sibling_n.index()] = depth + 1;
  114. node_stack.push_back({sibling_n, depth + 1});
  115. }
  116. }
  117. output << "[\n";
  118. for (Node n : postorder()) {
  119. PrintNode(output, n, indents[n.index()], /*adding_children=*/false);
  120. output << ",\n";
  121. }
  122. output << "]\n";
  123. }
  124. auto ParseTree::Print(llvm::raw_ostream& output, bool preorder) const -> void {
  125. if (!preorder) {
  126. Print(output);
  127. return;
  128. }
  129. output << "[\n";
  130. // The parse tree is stored in postorder. The preorder can be constructed
  131. // by reversing the order of each level of siblings within an RPO. The
  132. // sibling iterators are directly built around RPO and so can be used with a
  133. // stack to produce preorder.
  134. // The roots, like siblings, are in RPO (so reversed), but we add them in
  135. // order here because we'll pop off the stack effectively reversing then.
  136. llvm::SmallVector<std::pair<Node, int>, 16> node_stack;
  137. for (Node n : roots()) {
  138. node_stack.push_back({n, 0});
  139. }
  140. while (!node_stack.empty()) {
  141. Node n;
  142. int depth;
  143. std::tie(n, depth) = node_stack.pop_back_val();
  144. if (PrintNode(output, n, depth, /*adding_children=*/true)) {
  145. // Has children, so we descend. We append the children in order here as
  146. // well because they will get reversed when popped off the stack.
  147. for (Node sibling_n : children(n)) {
  148. node_stack.push_back({sibling_n, depth + 1});
  149. }
  150. continue;
  151. }
  152. int next_depth = node_stack.empty() ? 0 : node_stack.back().second;
  153. CARBON_CHECK(next_depth <= depth) << "Cannot have the next depth increase!";
  154. for (int close_children_count : llvm::seq(0, depth - next_depth)) {
  155. (void)close_children_count;
  156. output << "]}";
  157. }
  158. // We always end with a comma and a new line as we'll move to the next
  159. // node at whatever the current level ends up being.
  160. output << ",\n";
  161. }
  162. output << "]\n";
  163. }
  164. auto ParseTree::Verify() const -> bool {
  165. // Verify basic tree structure invariants.
  166. llvm::SmallVector<ParseTree::Node, 16> ancestors;
  167. for (Node n : llvm::reverse(postorder())) {
  168. const auto& n_impl = node_impls_[n.index()];
  169. if (n_impl.has_error && !has_errors_) {
  170. llvm::errs()
  171. << "Node #" << n.index()
  172. << " has errors, but the tree is not marked as having any.\n";
  173. return false;
  174. }
  175. if (n_impl.subtree_size > 1) {
  176. if (!ancestors.empty()) {
  177. auto parent_n = ancestors.back();
  178. const auto& parent_n_impl = node_impls_[parent_n.index()];
  179. int end_index = n.index() - n_impl.subtree_size;
  180. int parent_end_index = parent_n.index() - parent_n_impl.subtree_size;
  181. if (parent_end_index > end_index) {
  182. llvm::errs() << "Node #" << n.index() << " has a subtree size of "
  183. << n_impl.subtree_size
  184. << " which extends beyond its parent's (node #"
  185. << parent_n.index() << ") subtree (size "
  186. << parent_n_impl.subtree_size << ")\n";
  187. return false;
  188. }
  189. }
  190. // Has children, so we descend.
  191. ancestors.push_back(n);
  192. continue;
  193. }
  194. if (n_impl.subtree_size < 1) {
  195. llvm::errs() << "Node #" << n.index()
  196. << " has an invalid subtree size of " << n_impl.subtree_size
  197. << "!\n";
  198. return false;
  199. }
  200. // We're going to pop off some levels of the tree. Check each ancestor to
  201. // make sure the offsets are correct.
  202. int next_index = n.index() - 1;
  203. while (!ancestors.empty()) {
  204. ParseTree::Node parent_n = ancestors.back();
  205. if ((parent_n.index() - node_impls_[parent_n.index()].subtree_size) !=
  206. next_index) {
  207. break;
  208. }
  209. ancestors.pop_back();
  210. }
  211. }
  212. if (!ancestors.empty()) {
  213. llvm::errs()
  214. << "Finished walking the parse tree and there are still ancestors:\n";
  215. for (Node ancestor_n : ancestors) {
  216. llvm::errs() << " Node #" << ancestor_n.index() << "\n";
  217. }
  218. return false;
  219. }
  220. return true;
  221. }
  222. auto ParseTree::Node::Print(llvm::raw_ostream& output) const -> void {
  223. output << index();
  224. }
  225. auto ParseTree::PostorderIterator::Print(llvm::raw_ostream& output) const
  226. -> void {
  227. output << node_.index();
  228. }
  229. auto ParseTree::SiblingIterator::Print(llvm::raw_ostream& output) const
  230. -> void {
  231. output << node_.index();
  232. }
  233. } // namespace Carbon