1//===--- Rewriter.cpp - Code rewriting interface --------------------------===//
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//  This file defines the Rewriter class, which is used for code
11//  transformations.
12//
13//===----------------------------------------------------------------------===//
14
15#include "clang/Rewrite/Core/Rewriter.h"
16#include "clang/AST/Decl.h"
17#include "clang/AST/PrettyPrinter.h"
18#include "clang/AST/Stmt.h"
19#include "clang/Basic/DiagnosticIDs.h"
20#include "clang/Basic/FileManager.h"
21#include "clang/Basic/SourceManager.h"
22#include "clang/Lex/Lexer.h"
23#include "llvm/ADT/SmallString.h"
24#include "llvm/Config/llvm-config.h"
25#include "llvm/Support/FileSystem.h"
26#include "llvm/Support/raw_ostream.h"
27using namespace clang;
28
29raw_ostream &RewriteBuffer::write(raw_ostream &os) const {
30  // Walk RewriteRope chunks efficiently using MoveToNextPiece() instead of the
31  // character iterator.
32  for (RopePieceBTreeIterator I = begin(), E = end(); I != E;
33       I.MoveToNextPiece())
34    os << I.piece();
35  return os;
36}
37
38/// \brief Return true if this character is non-new-line whitespace:
39/// ' ', '\\t', '\\f', '\\v', '\\r'.
40static inline bool isWhitespace(unsigned char c) {
41  switch (c) {
42  case ' ':
43  case '\t':
44  case '\f':
45  case '\v':
46  case '\r':
47    return true;
48  default:
49    return false;
50  }
51}
52
53void RewriteBuffer::RemoveText(unsigned OrigOffset, unsigned Size,
54                               bool removeLineIfEmpty) {
55  // Nothing to remove, exit early.
56  if (Size == 0) return;
57
58  unsigned RealOffset = getMappedOffset(OrigOffset, true);
59  assert(RealOffset+Size < Buffer.size() && "Invalid location");
60
61  // Remove the dead characters.
62  Buffer.erase(RealOffset, Size);
63
64  // Add a delta so that future changes are offset correctly.
65  AddReplaceDelta(OrigOffset, -Size);
66
67  if (removeLineIfEmpty) {
68    // Find the line that the remove occurred and if it is completely empty
69    // remove the line as well.
70
71    iterator curLineStart = begin();
72    unsigned curLineStartOffs = 0;
73    iterator posI = begin();
74    for (unsigned i = 0; i != RealOffset; ++i) {
75      if (*posI == '\n') {
76        curLineStart = posI;
77        ++curLineStart;
78        curLineStartOffs = i + 1;
79      }
80      ++posI;
81    }
82
83    unsigned lineSize = 0;
84    posI = curLineStart;
85    while (posI != end() && isWhitespace(*posI)) {
86      ++posI;
87      ++lineSize;
88    }
89    if (posI != end() && *posI == '\n') {
90      Buffer.erase(curLineStartOffs, lineSize + 1/* + '\n'*/);
91      AddReplaceDelta(curLineStartOffs, -(lineSize + 1/* + '\n'*/));
92    }
93  }
94}
95
96void RewriteBuffer::InsertText(unsigned OrigOffset, StringRef Str,
97                               bool InsertAfter) {
98
99  // Nothing to insert, exit early.
100  if (Str.empty()) return;
101
102  unsigned RealOffset = getMappedOffset(OrigOffset, InsertAfter);
103  Buffer.insert(RealOffset, Str.begin(), Str.end());
104
105  // Add a delta so that future changes are offset correctly.
106  AddInsertDelta(OrigOffset, Str.size());
107}
108
109/// ReplaceText - This method replaces a range of characters in the input
110/// buffer with a new string.  This is effectively a combined "remove+insert"
111/// operation.
112void RewriteBuffer::ReplaceText(unsigned OrigOffset, unsigned OrigLength,
113                                StringRef NewStr) {
114  unsigned RealOffset = getMappedOffset(OrigOffset, true);
115  Buffer.erase(RealOffset, OrigLength);
116  Buffer.insert(RealOffset, NewStr.begin(), NewStr.end());
117  if (OrigLength != NewStr.size())
118    AddReplaceDelta(OrigOffset, NewStr.size() - OrigLength);
119}
120
121
122//===----------------------------------------------------------------------===//
123// Rewriter class
124//===----------------------------------------------------------------------===//
125
126/// getRangeSize - Return the size in bytes of the specified range if they
127/// are in the same file.  If not, this returns -1.
128int Rewriter::getRangeSize(const CharSourceRange &Range,
129                           RewriteOptions opts) const {
130  if (!isRewritable(Range.getBegin()) ||
131      !isRewritable(Range.getEnd())) return -1;
132
133  FileID StartFileID, EndFileID;
134  unsigned StartOff, EndOff;
135
136  StartOff = getLocationOffsetAndFileID(Range.getBegin(), StartFileID);
137  EndOff   = getLocationOffsetAndFileID(Range.getEnd(), EndFileID);
138
139  if (StartFileID != EndFileID)
140    return -1;
141
142  // If edits have been made to this buffer, the delta between the range may
143  // have changed.
144  std::map<FileID, RewriteBuffer>::const_iterator I =
145    RewriteBuffers.find(StartFileID);
146  if (I != RewriteBuffers.end()) {
147    const RewriteBuffer &RB = I->second;
148    EndOff = RB.getMappedOffset(EndOff, opts.IncludeInsertsAtEndOfRange);
149    StartOff = RB.getMappedOffset(StartOff, !opts.IncludeInsertsAtBeginOfRange);
150  }
151
152
153  // Adjust the end offset to the end of the last token, instead of being the
154  // start of the last token if this is a token range.
155  if (Range.isTokenRange())
156    EndOff += Lexer::MeasureTokenLength(Range.getEnd(), *SourceMgr, *LangOpts);
157
158  return EndOff-StartOff;
159}
160
161int Rewriter::getRangeSize(SourceRange Range, RewriteOptions opts) const {
162  return getRangeSize(CharSourceRange::getTokenRange(Range), opts);
163}
164
165
166/// getRewrittenText - Return the rewritten form of the text in the specified
167/// range.  If the start or end of the range was unrewritable or if they are
168/// in different buffers, this returns an empty string.
169///
170/// Note that this method is not particularly efficient.
171///
172std::string Rewriter::getRewrittenText(SourceRange Range) const {
173  if (!isRewritable(Range.getBegin()) ||
174      !isRewritable(Range.getEnd()))
175    return "";
176
177  FileID StartFileID, EndFileID;
178  unsigned StartOff, EndOff;
179  StartOff = getLocationOffsetAndFileID(Range.getBegin(), StartFileID);
180  EndOff   = getLocationOffsetAndFileID(Range.getEnd(), EndFileID);
181
182  if (StartFileID != EndFileID)
183    return ""; // Start and end in different buffers.
184
185  // If edits have been made to this buffer, the delta between the range may
186  // have changed.
187  std::map<FileID, RewriteBuffer>::const_iterator I =
188    RewriteBuffers.find(StartFileID);
189  if (I == RewriteBuffers.end()) {
190    // If the buffer hasn't been rewritten, just return the text from the input.
191    const char *Ptr = SourceMgr->getCharacterData(Range.getBegin());
192
193    // Adjust the end offset to the end of the last token, instead of being the
194    // start of the last token.
195    EndOff += Lexer::MeasureTokenLength(Range.getEnd(), *SourceMgr, *LangOpts);
196    return std::string(Ptr, Ptr+EndOff-StartOff);
197  }
198
199  const RewriteBuffer &RB = I->second;
200  EndOff = RB.getMappedOffset(EndOff, true);
201  StartOff = RB.getMappedOffset(StartOff);
202
203  // Adjust the end offset to the end of the last token, instead of being the
204  // start of the last token.
205  EndOff += Lexer::MeasureTokenLength(Range.getEnd(), *SourceMgr, *LangOpts);
206
207  // Advance the iterators to the right spot, yay for linear time algorithms.
208  RewriteBuffer::iterator Start = RB.begin();
209  std::advance(Start, StartOff);
210  RewriteBuffer::iterator End = Start;
211  std::advance(End, EndOff-StartOff);
212
213  return std::string(Start, End);
214}
215
216unsigned Rewriter::getLocationOffsetAndFileID(SourceLocation Loc,
217                                              FileID &FID) const {
218  assert(Loc.isValid() && "Invalid location");
219  std::pair<FileID,unsigned> V = SourceMgr->getDecomposedLoc(Loc);
220  FID = V.first;
221  return V.second;
222}
223
224
225/// getEditBuffer - Get or create a RewriteBuffer for the specified FileID.
226///
227RewriteBuffer &Rewriter::getEditBuffer(FileID FID) {
228  std::map<FileID, RewriteBuffer>::iterator I =
229    RewriteBuffers.lower_bound(FID);
230  if (I != RewriteBuffers.end() && I->first == FID)
231    return I->second;
232  I = RewriteBuffers.insert(I, std::make_pair(FID, RewriteBuffer()));
233
234  StringRef MB = SourceMgr->getBufferData(FID);
235  I->second.Initialize(MB.begin(), MB.end());
236
237  return I->second;
238}
239
240/// InsertText - Insert the specified string at the specified location in the
241/// original buffer.
242bool Rewriter::InsertText(SourceLocation Loc, StringRef Str,
243                          bool InsertAfter, bool indentNewLines) {
244  if (!isRewritable(Loc)) return true;
245  FileID FID;
246  unsigned StartOffs = getLocationOffsetAndFileID(Loc, FID);
247
248  SmallString<128> indentedStr;
249  if (indentNewLines && Str.find('\n') != StringRef::npos) {
250    StringRef MB = SourceMgr->getBufferData(FID);
251
252    unsigned lineNo = SourceMgr->getLineNumber(FID, StartOffs) - 1;
253    const SrcMgr::ContentCache *
254        Content = SourceMgr->getSLocEntry(FID).getFile().getContentCache();
255    unsigned lineOffs = Content->SourceLineCache[lineNo];
256
257    // Find the whitespace at the start of the line.
258    StringRef indentSpace;
259    {
260      unsigned i = lineOffs;
261      while (isWhitespace(MB[i]))
262        ++i;
263      indentSpace = MB.substr(lineOffs, i-lineOffs);
264    }
265
266    SmallVector<StringRef, 4> lines;
267    Str.split(lines, "\n");
268
269    for (unsigned i = 0, e = lines.size(); i != e; ++i) {
270      indentedStr += lines[i];
271      if (i < e-1) {
272        indentedStr += '\n';
273        indentedStr += indentSpace;
274      }
275    }
276    Str = indentedStr.str();
277  }
278
279  getEditBuffer(FID).InsertText(StartOffs, Str, InsertAfter);
280  return false;
281}
282
283bool Rewriter::InsertTextAfterToken(SourceLocation Loc, StringRef Str) {
284  if (!isRewritable(Loc)) return true;
285  FileID FID;
286  unsigned StartOffs = getLocationOffsetAndFileID(Loc, FID);
287  RewriteOptions rangeOpts;
288  rangeOpts.IncludeInsertsAtBeginOfRange = false;
289  StartOffs += getRangeSize(SourceRange(Loc, Loc), rangeOpts);
290  getEditBuffer(FID).InsertText(StartOffs, Str, /*InsertAfter*/true);
291  return false;
292}
293
294/// RemoveText - Remove the specified text region.
295bool Rewriter::RemoveText(SourceLocation Start, unsigned Length,
296                          RewriteOptions opts) {
297  if (!isRewritable(Start)) return true;
298  FileID FID;
299  unsigned StartOffs = getLocationOffsetAndFileID(Start, FID);
300  getEditBuffer(FID).RemoveText(StartOffs, Length, opts.RemoveLineIfEmpty);
301  return false;
302}
303
304/// ReplaceText - This method replaces a range of characters in the input
305/// buffer with a new string.  This is effectively a combined "remove/insert"
306/// operation.
307bool Rewriter::ReplaceText(SourceLocation Start, unsigned OrigLength,
308                           StringRef NewStr) {
309  if (!isRewritable(Start)) return true;
310  FileID StartFileID;
311  unsigned StartOffs = getLocationOffsetAndFileID(Start, StartFileID);
312
313  getEditBuffer(StartFileID).ReplaceText(StartOffs, OrigLength, NewStr);
314  return false;
315}
316
317bool Rewriter::ReplaceText(SourceRange range, SourceRange replacementRange) {
318  if (!isRewritable(range.getBegin())) return true;
319  if (!isRewritable(range.getEnd())) return true;
320  if (replacementRange.isInvalid()) return true;
321  SourceLocation start = range.getBegin();
322  unsigned origLength = getRangeSize(range);
323  unsigned newLength = getRangeSize(replacementRange);
324  FileID FID;
325  unsigned newOffs = getLocationOffsetAndFileID(replacementRange.getBegin(),
326                                                FID);
327  StringRef MB = SourceMgr->getBufferData(FID);
328  return ReplaceText(start, origLength, MB.substr(newOffs, newLength));
329}
330
331/// ReplaceStmt - This replaces a Stmt/Expr with another, using the pretty
332/// printer to generate the replacement code.  This returns true if the input
333/// could not be rewritten, or false if successful.
334bool Rewriter::ReplaceStmt(Stmt *From, Stmt *To) {
335  assert(From != nullptr && To != nullptr && "Expected non-null Stmt's");
336
337  // Measaure the old text.
338  int Size = getRangeSize(From->getSourceRange());
339  if (Size == -1)
340    return true;
341
342  // Get the new text.
343  std::string SStr;
344  llvm::raw_string_ostream S(SStr);
345  To->printPretty(S, nullptr, PrintingPolicy(*LangOpts));
346  const std::string &Str = S.str();
347
348  ReplaceText(From->getLocStart(), Size, Str);
349  return false;
350}
351
352std::string Rewriter::ConvertToString(Stmt *From) {
353  assert(From != nullptr && "Expected non-null Stmt");
354  std::string SStr;
355  llvm::raw_string_ostream S(SStr);
356  From->printPretty(S, nullptr, PrintingPolicy(*LangOpts));
357  return S.str();
358}
359
360bool Rewriter::IncreaseIndentation(CharSourceRange range,
361                                   SourceLocation parentIndent) {
362  if (range.isInvalid()) return true;
363  if (!isRewritable(range.getBegin())) return true;
364  if (!isRewritable(range.getEnd())) return true;
365  if (!isRewritable(parentIndent)) return true;
366
367  FileID StartFileID, EndFileID, parentFileID;
368  unsigned StartOff, EndOff, parentOff;
369
370  StartOff = getLocationOffsetAndFileID(range.getBegin(), StartFileID);
371  EndOff   = getLocationOffsetAndFileID(range.getEnd(), EndFileID);
372  parentOff = getLocationOffsetAndFileID(parentIndent, parentFileID);
373
374  if (StartFileID != EndFileID || StartFileID != parentFileID)
375    return true;
376  if (StartOff > EndOff)
377    return true;
378
379  FileID FID = StartFileID;
380  StringRef MB = SourceMgr->getBufferData(FID);
381
382  unsigned parentLineNo = SourceMgr->getLineNumber(FID, parentOff) - 1;
383  unsigned startLineNo = SourceMgr->getLineNumber(FID, StartOff) - 1;
384  unsigned endLineNo = SourceMgr->getLineNumber(FID, EndOff) - 1;
385
386  const SrcMgr::ContentCache *
387      Content = SourceMgr->getSLocEntry(FID).getFile().getContentCache();
388
389  // Find where the lines start.
390  unsigned parentLineOffs = Content->SourceLineCache[parentLineNo];
391  unsigned startLineOffs = Content->SourceLineCache[startLineNo];
392
393  // Find the whitespace at the start of each line.
394  StringRef parentSpace, startSpace;
395  {
396    unsigned i = parentLineOffs;
397    while (isWhitespace(MB[i]))
398      ++i;
399    parentSpace = MB.substr(parentLineOffs, i-parentLineOffs);
400
401    i = startLineOffs;
402    while (isWhitespace(MB[i]))
403      ++i;
404    startSpace = MB.substr(startLineOffs, i-startLineOffs);
405  }
406  if (parentSpace.size() >= startSpace.size())
407    return true;
408  if (!startSpace.startswith(parentSpace))
409    return true;
410
411  StringRef indent = startSpace.substr(parentSpace.size());
412
413  // Indent the lines between start/end offsets.
414  RewriteBuffer &RB = getEditBuffer(FID);
415  for (unsigned lineNo = startLineNo; lineNo <= endLineNo; ++lineNo) {
416    unsigned offs = Content->SourceLineCache[lineNo];
417    unsigned i = offs;
418    while (isWhitespace(MB[i]))
419      ++i;
420    StringRef origIndent = MB.substr(offs, i-offs);
421    if (origIndent.startswith(startSpace))
422      RB.InsertText(offs, indent, /*InsertAfter=*/false);
423  }
424
425  return false;
426}
427
428namespace {
429// A wrapper for a file stream that atomically overwrites the target.
430//
431// Creates a file output stream for a temporary file in the constructor,
432// which is later accessible via getStream() if ok() return true.
433// Flushes the stream and moves the temporary file to the target location
434// in the destructor.
435class AtomicallyMovedFile {
436public:
437  AtomicallyMovedFile(DiagnosticsEngine &Diagnostics, StringRef Filename,
438                      bool &AllWritten)
439    : Diagnostics(Diagnostics), Filename(Filename), AllWritten(AllWritten) {
440    TempFilename = Filename;
441    TempFilename += "-%%%%%%%%";
442    int FD;
443    if (llvm::sys::fs::createUniqueFile(TempFilename.str(), FD, TempFilename)) {
444      AllWritten = false;
445      Diagnostics.Report(clang::diag::err_unable_to_make_temp)
446        << TempFilename;
447    } else {
448      FileStream.reset(new llvm::raw_fd_ostream(FD, /*shouldClose=*/true));
449    }
450  }
451
452  ~AtomicallyMovedFile() {
453    if (!ok()) return;
454
455    FileStream->flush();
456#ifdef LLVM_ON_WIN32
457    // Win32 does not allow rename/removing opened files.
458    FileStream.reset();
459#endif
460    if (std::error_code ec =
461            llvm::sys::fs::rename(TempFilename.str(), Filename)) {
462      AllWritten = false;
463      Diagnostics.Report(clang::diag::err_unable_to_rename_temp)
464        << TempFilename << Filename << ec.message();
465      // If the remove fails, there's not a lot we can do - this is already an
466      // error.
467      llvm::sys::fs::remove(TempFilename.str());
468    }
469  }
470
471  bool ok() { return (bool)FileStream; }
472  raw_ostream &getStream() { return *FileStream; }
473
474private:
475  DiagnosticsEngine &Diagnostics;
476  StringRef Filename;
477  SmallString<128> TempFilename;
478  std::unique_ptr<llvm::raw_fd_ostream> FileStream;
479  bool &AllWritten;
480};
481} // end anonymous namespace
482
483bool Rewriter::overwriteChangedFiles() {
484  bool AllWritten = true;
485  for (buffer_iterator I = buffer_begin(), E = buffer_end(); I != E; ++I) {
486    const FileEntry *Entry =
487        getSourceMgr().getFileEntryForID(I->first);
488    AtomicallyMovedFile File(getSourceMgr().getDiagnostics(), Entry->getName(),
489                             AllWritten);
490    if (File.ok()) {
491      I->second.write(File.getStream());
492    }
493  }
494  return !AllWritten;
495}
496