clang  6.0.0svn
SortJavaScriptImports.cpp
Go to the documentation of this file.
1 //===--- SortJavaScriptImports.cpp - Sort ES6 Imports -----------*- C++ -*-===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 ///
10 /// \file
11 /// \brief This file implements a sort operation for JavaScript ES6 imports.
12 ///
13 //===----------------------------------------------------------------------===//
14 
15 #include "SortJavaScriptImports.h"
16 #include "TokenAnalyzer.h"
17 #include "TokenAnnotator.h"
18 #include "clang/Basic/Diagnostic.h"
20 #include "clang/Basic/LLVM.h"
23 #include "clang/Format/Format.h"
24 #include "llvm/ADT/STLExtras.h"
25 #include "llvm/ADT/SmallVector.h"
26 #include "llvm/Support/Debug.h"
27 #include <algorithm>
28 #include <string>
29 
30 #define DEBUG_TYPE "format-formatter"
31 
32 namespace clang {
33 namespace format {
34 
35 class FormatTokenLexer;
36 
38 
39 // An imported symbol in a JavaScript ES6 import/export, possibly aliased.
41  StringRef Symbol;
42  StringRef Alias;
44 
45  bool operator==(const JsImportedSymbol &RHS) const {
46  // Ignore Range for comparison, it is only used to stitch code together,
47  // but imports at different code locations are still conceptually the same.
48  return Symbol == RHS.Symbol && Alias == RHS.Alias;
49  }
50 };
51 
52 // An ES6 module reference.
53 //
54 // ES6 implements a module system, where individual modules (~= source files)
55 // can reference other modules, either importing symbols from them, or exporting
56 // symbols from them:
57 // import {foo} from 'foo';
58 // export {foo};
59 // export {bar} from 'bar';
60 //
61 // `export`s with URLs are syntactic sugar for an import of the symbol from the
62 // URL, followed by an export of the symbol, allowing this code to treat both
63 // statements more or less identically, with the exception being that `export`s
64 // are sorted last.
65 //
66 // imports and exports support individual symbols, but also a wildcard syntax:
67 // import * as prefix from 'foo';
68 // export * from 'bar';
69 //
70 // This struct represents both exports and imports to build up the information
71 // required for sorting module references.
73  bool IsExport = false;
74  // Module references are sorted into these categories, in order.
76  SIDE_EFFECT, // "import 'something';"
77  ABSOLUTE, // from 'something'
78  RELATIVE_PARENT, // from '../*'
79  RELATIVE, // from './*'
80  };
81  ReferenceCategory Category = ReferenceCategory::SIDE_EFFECT;
82  // The URL imported, e.g. `import .. from 'url';`. Empty for `export {a, b};`.
83  StringRef URL;
84  // Prefix from "import * as prefix". Empty for symbol imports and `export *`.
85  // Implies an empty names list.
86  StringRef Prefix;
87  // Symbols from `import {SymbolA, SymbolB, ...} from ...;`.
89  // Textual position of the import/export, including preceding and trailing
90  // comments.
92 };
93 
94 bool operator<(const JsModuleReference &LHS, const JsModuleReference &RHS) {
95  if (LHS.IsExport != RHS.IsExport)
96  return LHS.IsExport < RHS.IsExport;
97  if (LHS.Category != RHS.Category)
98  return LHS.Category < RHS.Category;
99  if (LHS.Category == JsModuleReference::ReferenceCategory::SIDE_EFFECT)
100  // Side effect imports might be ordering sensitive. Consider them equal so
101  // that they maintain their relative order in the stable sort below.
102  // This retains transitivity because LHS.Category == RHS.Category here.
103  return false;
104  // Empty URLs sort *last* (for export {...};).
105  if (LHS.URL.empty() != RHS.URL.empty())
106  return LHS.URL.empty() < RHS.URL.empty();
107  if (int Res = LHS.URL.compare_lower(RHS.URL))
108  return Res < 0;
109  // '*' imports (with prefix) sort before {a, b, ...} imports.
110  if (LHS.Prefix.empty() != RHS.Prefix.empty())
111  return LHS.Prefix.empty() < RHS.Prefix.empty();
112  if (LHS.Prefix != RHS.Prefix)
113  return LHS.Prefix > RHS.Prefix;
114  return false;
115 }
116 
117 // JavaScriptImportSorter sorts JavaScript ES6 imports and exports. It is
118 // implemented as a TokenAnalyzer because ES6 imports have substantial syntactic
119 // structure, making it messy to sort them using regular expressions.
121 public:
123  : TokenAnalyzer(Env, Style),
124  FileContents(Env.getSourceManager().getBufferData(Env.getFileID())) {}
125 
128  SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
129  FormatTokenLexer &Tokens) override {
130  tooling::Replacements Result;
131  AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
132  AnnotatedLines.end());
133 
134  const AdditionalKeywords &Keywords = Tokens.getKeywords();
136  AnnotatedLine *FirstNonImportLine;
137  std::tie(References, FirstNonImportLine) =
138  parseModuleReferences(Keywords, AnnotatedLines);
139 
140  if (References.empty())
141  return Result;
142 
144  for (unsigned i = 0, e = References.size(); i != e; ++i)
145  Indices.push_back(i);
146  std::stable_sort(Indices.begin(), Indices.end(),
147  [&](unsigned LHSI, unsigned RHSI) {
148  return References[LHSI] < References[RHSI];
149  });
150  bool ReferencesInOrder = std::is_sorted(Indices.begin(), Indices.end());
151 
152  std::string ReferencesText;
153  bool SymbolsInOrder = true;
154  for (unsigned i = 0, e = Indices.size(); i != e; ++i) {
155  JsModuleReference Reference = References[Indices[i]];
156  if (appendReference(ReferencesText, Reference))
157  SymbolsInOrder = false;
158  if (i + 1 < e) {
159  // Insert breaks between imports and exports.
160  ReferencesText += "\n";
161  // Separate imports groups with two line breaks, but keep all exports
162  // in a single group.
163  if (!Reference.IsExport &&
164  (Reference.IsExport != References[Indices[i + 1]].IsExport ||
165  Reference.Category != References[Indices[i + 1]].Category))
166  ReferencesText += "\n";
167  }
168  }
169 
170  if (ReferencesInOrder && SymbolsInOrder)
171  return Result;
172 
173  SourceRange InsertionPoint = References[0].Range;
174  InsertionPoint.setEnd(References[References.size() - 1].Range.getEnd());
175 
176  // The loop above might collapse previously existing line breaks between
177  // import blocks, and thus shrink the file. SortIncludes must not shrink
178  // overall source length as there is currently no re-calculation of ranges
179  // after applying source sorting.
180  // This loop just backfills trailing spaces after the imports, which are
181  // harmless and will be stripped by the subsequent formatting pass.
182  // FIXME: A better long term fix is to re-calculate Ranges after sorting.
183  unsigned PreviousSize = getSourceText(InsertionPoint).size();
184  while (ReferencesText.size() < PreviousSize) {
185  ReferencesText += " ";
186  }
187 
188  // Separate references from the main code body of the file.
189  if (FirstNonImportLine && FirstNonImportLine->First->NewlinesBefore < 2)
190  ReferencesText += "\n";
191 
192  DEBUG(llvm::dbgs() << "Replacing imports:\n"
193  << getSourceText(InsertionPoint) << "\nwith:\n"
194  << ReferencesText << "\n");
195  auto Err = Result.add(tooling::Replacement(
196  Env.getSourceManager(), CharSourceRange::getCharRange(InsertionPoint),
197  ReferencesText));
198  // FIXME: better error handling. For now, just print error message and skip
199  // the replacement for the release version.
200  if (Err) {
201  llvm::errs() << llvm::toString(std::move(Err)) << "\n";
202  assert(false);
203  }
204 
205  return Result;
206  }
207 
208 private:
209  FormatToken *Current;
210  FormatToken *LineEnd;
211 
212  FormatToken invalidToken;
213 
214  StringRef FileContents;
215 
216  void skipComments() { Current = skipComments(Current); }
217 
218  FormatToken *skipComments(FormatToken *Tok) {
219  while (Tok && Tok->is(tok::comment))
220  Tok = Tok->Next;
221  return Tok;
222  }
223 
224  void nextToken() {
225  Current = Current->Next;
226  skipComments();
227  if (!Current || Current == LineEnd->Next) {
228  // Set the current token to an invalid token, so that further parsing on
229  // this line fails.
230  invalidToken.Tok.setKind(tok::unknown);
231  Current = &invalidToken;
232  }
233  }
234 
235  StringRef getSourceText(SourceRange Range) {
236  return getSourceText(Range.getBegin(), Range.getEnd());
237  }
238 
239  StringRef getSourceText(SourceLocation Begin, SourceLocation End) {
240  const SourceManager &SM = Env.getSourceManager();
241  return FileContents.substr(SM.getFileOffset(Begin),
242  SM.getFileOffset(End) - SM.getFileOffset(Begin));
243  }
244 
245  // Appends ``Reference`` to ``Buffer``, returning true if text within the
246  // ``Reference`` changed (e.g. symbol order).
247  bool appendReference(std::string &Buffer, JsModuleReference &Reference) {
248  // Sort the individual symbols within the import.
249  // E.g. `import {b, a} from 'x';` -> `import {a, b} from 'x';`
250  SmallVector<JsImportedSymbol, 1> Symbols = Reference.Symbols;
251  std::stable_sort(
252  Symbols.begin(), Symbols.end(),
253  [&](const JsImportedSymbol &LHS, const JsImportedSymbol &RHS) {
254  return LHS.Symbol.compare_lower(RHS.Symbol) < 0;
255  });
256  if (Symbols == Reference.Symbols) {
257  // No change in symbol order.
258  StringRef ReferenceStmt = getSourceText(Reference.Range);
259  Buffer += ReferenceStmt;
260  return false;
261  }
262  // Stitch together the module reference start...
263  SourceLocation SymbolsStart = Reference.Symbols.front().Range.getBegin();
264  SourceLocation SymbolsEnd = Reference.Symbols.back().Range.getEnd();
265  Buffer += getSourceText(Reference.Range.getBegin(), SymbolsStart);
266  // ... then the references in order ...
267  for (auto I = Symbols.begin(), E = Symbols.end(); I != E; ++I) {
268  if (I != Symbols.begin())
269  Buffer += ",";
270  Buffer += getSourceText(I->Range);
271  }
272  // ... followed by the module reference end.
273  Buffer += getSourceText(SymbolsEnd, Reference.Range.getEnd());
274  return true;
275  }
276 
277  // Parses module references in the given lines. Returns the module references,
278  // and a pointer to the first "main code" line if that is adjacent to the
279  // affected lines of module references, nullptr otherwise.
280  std::pair<SmallVector<JsModuleReference, 16>, AnnotatedLine *>
281  parseModuleReferences(const AdditionalKeywords &Keywords,
282  SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
284  SourceLocation Start;
285  AnnotatedLine *FirstNonImportLine = nullptr;
286  bool AnyImportAffected = false;
287  for (auto Line : AnnotatedLines) {
288  Current = Line->First;
289  LineEnd = Line->Last;
290  skipComments();
291  if (Start.isInvalid() || References.empty())
292  // After the first file level comment, consider line comments to be part
293  // of the import that immediately follows them by using the previously
294  // set Start.
295  Start = Line->First->Tok.getLocation();
296  if (!Current) {
297  // Only comments on this line. Could be the first non-import line.
298  FirstNonImportLine = Line;
299  continue;
300  }
301  JsModuleReference Reference;
302  Reference.Range.setBegin(Start);
303  if (!parseModuleReference(Keywords, Reference)) {
304  if (!FirstNonImportLine)
305  FirstNonImportLine = Line; // if no comment before.
306  break;
307  }
308  FirstNonImportLine = nullptr;
309  AnyImportAffected = AnyImportAffected || Line->Affected;
310  Reference.Range.setEnd(LineEnd->Tok.getEndLoc());
311  DEBUG({
312  llvm::dbgs() << "JsModuleReference: {"
313  << "is_export: " << Reference.IsExport
314  << ", cat: " << Reference.Category
315  << ", url: " << Reference.URL
316  << ", prefix: " << Reference.Prefix;
317  for (size_t i = 0; i < Reference.Symbols.size(); ++i)
318  llvm::dbgs() << ", " << Reference.Symbols[i].Symbol << " as "
319  << Reference.Symbols[i].Alias;
320  llvm::dbgs() << ", text: " << getSourceText(Reference.Range);
321  llvm::dbgs() << "}\n";
322  });
323  References.push_back(Reference);
324  Start = SourceLocation();
325  }
326  // Sort imports if any import line was affected.
327  if (!AnyImportAffected)
328  References.clear();
329  return std::make_pair(References, FirstNonImportLine);
330  }
331 
332  // Parses a JavaScript/ECMAScript 6 module reference.
333  // See http://www.ecma-international.org/ecma-262/6.0/#sec-scripts-and-modules
334  // for grammar EBNF (production ModuleItem).
335  bool parseModuleReference(const AdditionalKeywords &Keywords,
336  JsModuleReference &Reference) {
337  if (!Current || !Current->isOneOf(Keywords.kw_import, tok::kw_export))
338  return false;
339  Reference.IsExport = Current->is(tok::kw_export);
340 
341  nextToken();
342  if (Current->isStringLiteral() && !Reference.IsExport) {
343  // "import 'side-effect';"
344  Reference.Category = JsModuleReference::ReferenceCategory::SIDE_EFFECT;
345  Reference.URL =
346  Current->TokenText.substr(1, Current->TokenText.size() - 2);
347  return true;
348  }
349 
350  if (!parseModuleBindings(Keywords, Reference))
351  return false;
352 
353  if (Current->is(Keywords.kw_from)) {
354  // imports have a 'from' clause, exports might not.
355  nextToken();
356  if (!Current->isStringLiteral())
357  return false;
358  // URL = TokenText without the quotes.
359  Reference.URL =
360  Current->TokenText.substr(1, Current->TokenText.size() - 2);
361  if (Reference.URL.startswith(".."))
362  Reference.Category =
363  JsModuleReference::ReferenceCategory::RELATIVE_PARENT;
364  else if (Reference.URL.startswith("."))
365  Reference.Category = JsModuleReference::ReferenceCategory::RELATIVE;
366  else
367  Reference.Category = JsModuleReference::ReferenceCategory::ABSOLUTE;
368  } else {
369  // w/o URL groups with "empty".
370  Reference.Category = JsModuleReference::ReferenceCategory::RELATIVE;
371  }
372  return true;
373  }
374 
375  bool parseModuleBindings(const AdditionalKeywords &Keywords,
376  JsModuleReference &Reference) {
377  if (parseStarBinding(Keywords, Reference))
378  return true;
379  return parseNamedBindings(Keywords, Reference);
380  }
381 
382  bool parseStarBinding(const AdditionalKeywords &Keywords,
383  JsModuleReference &Reference) {
384  // * as prefix from '...';
385  if (Current->isNot(tok::star))
386  return false;
387  nextToken();
388  if (Current->isNot(Keywords.kw_as))
389  return false;
390  nextToken();
391  if (Current->isNot(tok::identifier))
392  return false;
393  Reference.Prefix = Current->TokenText;
394  nextToken();
395  return true;
396  }
397 
398  bool parseNamedBindings(const AdditionalKeywords &Keywords,
399  JsModuleReference &Reference) {
400  if (Current->is(tok::identifier)) {
401  nextToken();
402  if (Current->is(Keywords.kw_from))
403  return true;
404  if (Current->isNot(tok::comma))
405  return false;
406  nextToken(); // eat comma.
407  }
408  if (Current->isNot(tok::l_brace))
409  return false;
410 
411  // {sym as alias, sym2 as ...} from '...';
412  while (Current->isNot(tok::r_brace)) {
413  nextToken();
414  if (Current->is(tok::r_brace))
415  break;
416  if (!Current->isOneOf(tok::identifier, tok::kw_default))
417  return false;
418 
420  Symbol.Symbol = Current->TokenText;
421  // Make sure to include any preceding comments.
422  Symbol.Range.setBegin(
424  nextToken();
425 
426  if (Current->is(Keywords.kw_as)) {
427  nextToken();
428  if (!Current->isOneOf(tok::identifier, tok::kw_default))
429  return false;
430  Symbol.Alias = Current->TokenText;
431  nextToken();
432  }
433  Symbol.Range.setEnd(Current->Tok.getLocation());
434  Reference.Symbols.push_back(Symbol);
435 
436  if (!Current->isOneOf(tok::r_brace, tok::comma))
437  return false;
438  }
439  nextToken(); // consume r_brace
440  return true;
441  }
442 };
443 
445  StringRef Code,
447  StringRef FileName) {
448  // FIXME: Cursor support.
449  std::unique_ptr<Environment> Env =
450  Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
451  JavaScriptImportSorter Sorter(*Env, Style);
452  return Sorter.process();
453 }
454 
455 } // end namespace format
456 } // end namespace clang
if(T->getSizeExpr()) TRY_TO(TraverseStmt(T -> getSizeExpr()))
Token Tok
The Token.
Definition: FormatToken.h:124
Defines the SourceManager interface.
const AdditionalKeywords & getKeywords()
Maintains a set of replacements that are conflict-free.
Definition: Replacement.h:205
void setBegin(SourceLocation b)
unsigned NewlinesBefore
The number of newlines immediately before the Token.
Definition: FormatToken.h:130
FormatToken * Next
The next token in the unwrapped line.
Definition: FormatToken.h:283
JavaScriptImportSorter(const Environment &Env, const FormatStyle &Style)
bool operator==(const JsImportedSymbol &RHS) const
This file implements a token annotator, i.e.
int Category
Definition: Format.cpp:1319
void setKind(tok::TokenKind K)
Definition: Token.h:91
bool isNot(T Kind) const
Definition: FormatToken.h:313
const FormatToken & Tok
Forward-declares and imports various common LLVM datatypes that clang wants to use unqualified...
Defines the Diagnostic-related interfaces.
FormatToken * getPreviousNonComment() const
Returns the previous token ignoring comments.
Definition: FormatToken.h:448
tooling::Replacements analyze(TokenAnnotator &Annotator, SmallVectorImpl< AnnotatedLine *> &AnnotatedLines, FormatTokenLexer &Tokens) override
bool isOneOf(A K1, B K2) const
Definition: FormatToken.h:306
A text replacement.
Definition: Replacement.h:81
Determines extra information about the tokens comprising an UnwrappedLine.
SourceLocation End
const AnnotatedLine * Line
SourceLocation getLocation() const
Return a source location identifier for the specified offset in the current file. ...
Definition: Token.h:124
SourceLocation Begin
A wrapper around a Token storing information about the whitespace characters preceding it...
Definition: FormatToken.h:120
SourceLocation getEnd() const
const SourceManager & SM
Definition: Format.cpp:1308
tooling::Replacements process()
static CharSourceRange getCharRange(SourceRange R)
unsigned getFileOffset(SourceLocation SpellingLoc) const
Returns the offset from the start of the file that the specified SourceLocation represents.
Encodes a location in the source.
bool is(tok::TokenKind Kind) const
Definition: FormatToken.h:297
Various functions to configurably format source code.
Encapsulates keywords that are context sensitive or for languages not properly supported by Clang&#39;s l...
Definition: FormatToken.h:634
SourceRange WhitespaceRange
The range of the whitespace immediately preceding the Token.
Definition: FormatToken.h:137
StringRef TokenText
The raw text of the token.
Definition: FormatToken.h:168
The FormatStyle is used to configure the formatting to follow specific guidelines.
Definition: Format.h:46
static std::unique_ptr< Environment > CreateVirtualEnvironment(StringRef Code, StringRef FileName, ArrayRef< tooling::Range > Ranges)
Dataflow Directional Tag Classes.
This file implements a sorter for JavaScript ES6 imports.
bool operator<(const JsModuleReference &LHS, const JsModuleReference &RHS)
std::string toString(const til::SExpr *E)
SmallVector< JsImportedSymbol, 1 > Symbols
This file declares an abstract TokenAnalyzer, and associated helper classes.
Defines the clang::SourceLocation class and associated facilities.
llvm::Error add(const Replacement &R)
Adds a new replacement R to the current set of replacements.
void setEnd(SourceLocation e)
A trivial tuple used to represent a source range.
tooling::Replacements sortJavaScriptImports(const FormatStyle &Style, StringRef Code, ArrayRef< tooling::Range > Ranges, StringRef FileName)
bool isStringLiteral() const
Definition: FormatToken.h:329
SourceLocation getBegin() const
This class handles loading and caching of source files into memory.
const FormatStyle & Style
SourceLocation getEndLoc() const
Definition: Token.h:151