| 1 |
//===-- HTMLForest.cpp - browser-based parse forest explorer |
| 2 |
//---------------===// |
| 3 |
// |
| 4 |
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 5 |
// See https://llvm.org/LICENSE.txt for license information. |
| 6 |
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 7 |
// |
| 8 |
//===----------------------------------------------------------------------===// |
| 9 |
// |
| 10 |
// The plain text forest node dump (clang-pseudo -print-forest) is useful but |
| 11 |
// hard to reconcile with the code being examined, especially when it is large. |
| 12 |
// |
| 13 |
// HTMLForest produces a self-contained HTML file containing both the code and |
| 14 |
// the forest representation, linking them interactively with javascript. |
| 15 |
// At any given time, a single parse tree is shown (ambiguities resolved). |
| 16 |
// The user can switch between ambiguous alternatives. |
| 17 |
// |
| 18 |
// +-------+---------------+ |
| 19 |
// | | +-----+| |
| 20 |
// | #tree | #code |#info|| |
| 21 |
// | | +-----+| |
| 22 |
// | | | |
| 23 |
// +-------+---------------+ |
| 24 |
// |
| 25 |
// #tree is a hierarchical view of the nodes (nested <ul>s), like -print-forest. |
| 26 |
// (It is a simple tree, not a DAG, because ambiguities have been resolved). |
| 27 |
// Like -print-forest, trivial sequences are collapsed (expression~IDENTIFIER). |
| 28 |
// |
| 29 |
// #code is the source code, annotated with <span>s marking the node ranges. |
| 30 |
// These spans are usually invisible (exception: ambiguities are marked), but |
| 31 |
// they are used to show and change the selection. |
| 32 |
// |
| 33 |
// #info is a floating box that shows details of the currently selected node: |
| 34 |
// - rule (for sequence nodes). Abbreviated rules are also shown. |
| 35 |
// - alternatives (for ambiguous nodes). The user can choose an alternative. |
| 36 |
// - ancestors. The parent nodes show how this node fits in translation-unit. |
| 37 |
// |
| 38 |
// There are two types of 'active' node: |
| 39 |
// - *highlight* is what the cursor is over, and is colored blue. |
| 40 |
// Near ancestors are shaded faintly (onion-skin) to show local structure. |
| 41 |
// - *selection* is set by clicking. |
| 42 |
// The #info box shows the selection, and selected nodes have a dashed ring. |
| 43 |
// |
| 44 |
//===----------------------------------------------------------------------===// |
| 45 |
|
| 46 |
#include "clang-pseudo/Disambiguate.h" |
| 47 |
#include "clang-pseudo/Forest.h" |
| 48 |
#include "clang-pseudo/grammar/Grammar.h" |
| 49 |
#include "llvm/ADT/StringExtras.h" |
| 50 |
#include "llvm/Support/JSON.h" |
| 51 |
#include "llvm/Support/raw_ostream.h" |
| 52 |
namespace clang { |
| 53 |
namespace pseudo { |
| 54 |
namespace { |
| 55 |
|
| 56 |
// Defines const char HTMLForest_css[] = "...contents of HTMLForest.css..."; etc |
| 57 |
#include "HTMLForestResources.inc" |
| 58 |
|
| 59 |
struct Writer { |
| 60 |
llvm::raw_ostream &Out; |
| 61 |
const Grammar &G; |
| 62 |
const ForestNode &Root; |
| 63 |
const TokenStream &Stream; |
| 64 |
const Disambiguation &Disambig; |
| 65 |
|
| 66 |
void write() { |
| 67 |
Out << "<!doctype html>\n"; |
| 68 |
tag("html", [&] { |
| 69 |
tag("head", [&] { |
| 70 |
tag("title", [&] { Out << "HTMLForest"; }); |
| 71 |
tag("script", [&] { Out << HTMLForest_js; }); |
| 72 |
tag("style", [&] { Out << HTMLForest_css; }); |
| 73 |
tag("script", [&] { |
| 74 |
Out << "var forest="; |
| 75 |
writeForestJSON(); |
| 76 |
Out << ";"; |
| 77 |
}); |
| 78 |
tag("pre id='hidden-code' hidden", [&] { writeCode(); }); |
| 79 |
}); |
| 80 |
tag("body", [&] { Out << HTMLForest_html; }); |
| 81 |
}); |
| 82 |
} |
| 83 |
|
| 84 |
void writeCode(); |
| 85 |
void writeForestJSON(); |
| 86 |
void tag(llvm::StringRef Opener, llvm::function_ref<void()> Body) { |
| 87 |
Out << "<" << Opener << ">"; |
| 88 |
Body(); |
| 89 |
Out << "</" << Opener.split(' ').first << ">\n"; |
| 90 |
} |
| 91 |
}; |
| 92 |
|
| 93 |
void Writer::writeCode() { |
| 94 |
// This loop (whitespace logic) is cribbed from TokenStream::Print. |
| 95 |
bool FirstToken = true; |
| 96 |
unsigned LastLine = -1; |
| 97 |
StringRef LastText; |
| 98 |
for (const auto &T : Stream.tokens()) { |
| 99 |
StringRef Text = T.text(); |
| 100 |
if (FirstToken) { |
| 101 |
FirstToken = false; |
| 102 |
} else if (T.Line == LastLine) { |
| 103 |
if (LastText.data() + LastText.size() != Text.data()) |
| 104 |
Out << ' '; |
| 105 |
} else { |
| 106 |
Out << " \n"; // Extra space aids selection. |
| 107 |
Out.indent(T.Indent); |
| 108 |
} |
| 109 |
Out << "<span class='token' id='t" << Stream.index(T) << "'>"; |
| 110 |
llvm::printHTMLEscaped(Text, Out); |
| 111 |
Out << "</span>"; |
| 112 |
LastLine = T.Line; |
| 113 |
LastText = Text; |
| 114 |
} |
| 115 |
if (!FirstToken) |
| 116 |
Out << '\n'; |
| 117 |
} |
| 118 |
|
| 119 |
// Writes a JSON array of forest nodes. Items are e.g.: |
| 120 |
// {kind:'sequence', symbol:'compound-stmt', children:[5,8,33], |
| 121 |
// rule:'compound-stmt := ...'} {kind:'terminal', symbol:'VOID', token:'t52'} |
| 122 |
// {kind:'ambiguous', symbol:'type-specifier', children:[3,100] selected:3} |
| 123 |
// {kind:'opaque', symbol:'statement-seq', firstToken:'t5', lastToken:'t6'} |
| 124 |
void Writer::writeForestJSON() { |
| 125 |
// This is the flat array of nodes: the index into this array is the node ID. |
| 126 |
std::vector<std::pair<const ForestNode *, /*End*/ Token::Index>> Sequence; |
| 127 |
llvm::DenseMap<const ForestNode *, unsigned> Index; |
| 128 |
auto AssignID = [&](const ForestNode *N, Token::Index End) -> unsigned { |
| 129 |
auto R = Index.try_emplace(N, Sequence.size()); |
| 130 |
if (R.second) |
| 131 |
Sequence.push_back({N, End}); |
| 132 |
return R.first->second; |
| 133 |
}; |
| 134 |
AssignID(&Root, Stream.tokens().size()); |
| 135 |
auto TokenID = [](Token::Index I) { return ("t" + llvm::Twine(I)).str(); }; |
| 136 |
|
| 137 |
llvm::json::OStream Out(this->Out, 2); |
| 138 |
Out.array([&] { |
| 139 |
for (unsigned I = 0; I < Sequence.size(); ++I) { |
| 140 |
const ForestNode *N = Sequence[I].first; |
| 141 |
Token::Index End = Sequence[I].second; |
| 142 |
Out.object([&] { |
| 143 |
Out.attribute("symbol", G.symbolName(N->symbol())); |
| 144 |
switch (N->kind()) { |
| 145 |
case ForestNode::Terminal: |
| 146 |
Out.attribute("kind", "terminal"); |
| 147 |
Out.attribute("token", TokenID(N->startTokenIndex())); |
| 148 |
break; |
| 149 |
case ForestNode::Sequence: |
| 150 |
Out.attribute("kind", "sequence"); |
| 151 |
Out.attribute("rule", G.dumpRule(N->rule())); |
| 152 |
break; |
| 153 |
case ForestNode::Ambiguous: |
| 154 |
Out.attribute("kind", "ambiguous"); |
| 155 |
Out.attribute("selected", |
| 156 |
AssignID(N->children()[Disambig.lookup(N)], End)); |
| 157 |
break; |
| 158 |
case ForestNode::Opaque: |
| 159 |
Out.attribute("kind", "opaque"); |
| 160 |
Out.attribute("firstToken", TokenID(N->startTokenIndex())); |
| 161 |
// [firstToken, lastToken] is a closed range. |
| 162 |
// If empty, lastToken is omitted. |
| 163 |
if (N->startTokenIndex() != End) |
| 164 |
Out.attribute("lastToken", TokenID(End - 1)); |
| 165 |
break; |
| 166 |
} |
| 167 |
auto Children = N->children(); |
| 168 |
if (!Children.empty()) |
| 169 |
Out.attributeArray("children", [&] { |
| 170 |
for (unsigned I = 0; I < Children.size(); ++I) |
| 171 |
Out.value(AssignID(Children[I], |
| 172 |
I + 1 == Children.size() |
| 173 |
? End |
| 174 |
: Children[I + 1]->startTokenIndex())); |
| 175 |
}); |
| 176 |
}); |
| 177 |
} |
| 178 |
}); |
| 179 |
} |
| 180 |
|
| 181 |
} // namespace |
| 182 |
|
| 183 |
// We only accept the derived stream here. |
| 184 |
// FIXME: allow the original stream instead? |
| 185 |
void writeHTMLForest(llvm::raw_ostream &OS, const Grammar &G, |
| 186 |
const ForestNode &Root, const Disambiguation &Disambig, |
| 187 |
const TokenStream &Stream) { |
| 188 |
Writer{OS, G, Root, Stream, Disambig}.write(); |
| 189 |
} |
| 190 |
|
| 191 |
} // namespace pseudo |
| 192 |
} // namespace clang |