X-Git-Url: https://git.lyx.org/gitweb/?a=blobdiff_plain;f=src%2FCompare.cpp;h=796eb19989f5a45ca3ca662bc8a7c408a877abb9;hb=c7d29be153debac82e3d2e8865fcc849f0a5f40d;hp=cfbfdce45928b4a6d45c195300530338a44fe9b0;hpb=6bb8b55156120c4a061e225212061699e4100abe;p=lyx.git diff --git a/src/Compare.cpp b/src/Compare.cpp index cfbfdce459..796eb19989 100644 --- a/src/Compare.cpp +++ b/src/Compare.cpp @@ -12,14 +12,19 @@ #include "Compare.h" +#include "Author.h" #include "BufferParams.h" #include "Changes.h" +#include "CutAndPaste.h" +#include "ErrorList.h" +#include "Font.h" #include "insets/InsetText.h" -#include "support/lassert.h" - -#include +#include "support/docstream.h" +#include "support/lassert.h" +#include "support/lyxalgo.h" +#include "support/qstring_helpers.h" using namespace std; using namespace lyx::support; @@ -55,14 +60,14 @@ static void step(DocIterator & dit, DocIterator const & end, Direction direction */ class DocRange { public: - DocRange(DocIterator from_, DocIterator to_) + DocRange(DocIterator const & from_, DocIterator const & to_) : from(from_), to(to_) {} - DocRange(Buffer const * buf) + DocRange(Buffer const * buf) : + from(doc_iterator_begin(buf)), + to(doc_iterator_end(buf)) { - from = doc_iterator_begin(buf); - to = doc_iterator_end(buf); to.backwardPos(); } @@ -82,46 +87,32 @@ public: size_t DocRange::length() const { - pit_type startpit = from.pit(); - pit_type endpit = to.pit(); - ParagraphList const & ps_ = from.text()->paragraphs(); - - ParagraphList pars(boost::next(ps_.begin(), startpit), - boost::next(ps_.begin(), endpit + 1)); - - // Remove the end of the last paragraph; afterwards, remove the - // beginning of the first paragraph. - Paragraph & back = pars.back(); - back.eraseChars(to.pos(), back.size(), false); - Paragraph & front = pars.front(); - front.eraseChars(0, from.pos(), false); - - ParagraphList::const_iterator pit = pars.begin(); - ParagraphList::const_iterator end_it = pars.end(); - + ParagraphList const & ps = from.text()->paragraphs(); size_t length = 0; - for (; pit != end_it; ++pit) - length += pit->size() + 1; - - // The last paragraph has no paragraph-end - --length; - return length; + pit_type pit = from.pit(); + pit_type const endpit = to.pit(); + for (; pit < endpit; ++pit) + length += ps[pit].size() + 1; + length += to.pos() - from.pos(); + return length; } class DocPair { public: - DocPair() {} + DocPair() + {} DocPair(DocIterator o_, DocIterator n_) : o(o_), n(n_) {} - bool operator!=(DocPair const & rhs) { + bool operator!=(DocPair const & rhs) + { // this might not be intuitive but correct for our purpose return o != rhs.o && n != rhs.n; } - + DocPair & operator++() { @@ -141,17 +132,17 @@ public: /// DocIterator n; }; - + /** * A pair of two DocRanges. */ class DocRangePair { public: - DocRangePair(DocRange o_, DocRange n_) + DocRangePair(DocRange const & o_, DocRange const & n_) : o(o_), n(n_) {} - - DocRangePair(DocPair from, DocPair to) + + DocRangePair(DocPair const & from, DocPair const & to) : o(from.o, to.o), n(from.n, to.n) {} @@ -160,10 +151,16 @@ public: {} /// Returns the from pair - DocPair from() const { return DocPair(o.from, n.from); } + DocPair from() const + { + return DocPair(o.from, n.from); + } /// Returns the to pair - DocPair to() const { return DocPair(o.to, n.to); } + DocPair to() const + { + return DocPair(o.to, n.to); + } DocRange o; DocRange n; @@ -183,6 +180,49 @@ static DocRangePair stepIntoInset(DocPair const & inset_location) } +/** + * This class is designed to hold a vector that has both positive as + * negative indices. It is internally represented as two vectors, one + * for non-zero indices and one for negative indices. In this way, the + * vector can grow in both directions. + * If an index is not available in the vector, the default value is + * returned. If an object is put in the vector beyond its size, the + * empty spots in between are also filled with the default value. + */ +template +class compl_vector { +public: + compl_vector() + {} + + void reset(T const & def) + { + default_ = def; + Vp_.clear(); + Vn_.clear(); + } + + /// Gets the value at index. If it is not in the vector + /// the default value is inserted and returned. + T & operator[](int index) { + vector & V = index >= 0 ? Vp_ : Vn_; + unsigned int const ii = index >= 0 ? index : -index - 1; + while (ii >= V.size()) + V.push_back(default_); + return V[ii]; + } + +private: + /// The vector for positive indices + vector Vp_; + /// The vector for negative indices + vector Vn_; + /// The default value that is inserted in the vector + /// if more space is needed + T default_; +}; + + /** * The implementation of the algorithm that does the comparison * between two documents. @@ -190,14 +230,18 @@ static DocRangePair stepIntoInset(DocPair const & inset_location) class Compare::Impl { public: /// - Impl(Compare const & compare) - : abort_(false), compare_(compare) + Impl(Compare const & compare) + : abort_(false), n_(0), m_(0), offset_reverse_diagonal_(0), + odd_offset_(0), compare_(compare), + old_buf_(0), new_buf_(0), dest_buf_(0), dest_pars_(0), + recursion_level_(0), nested_inset_level_(0), D_(0) {} /// - ~Impl() {} + ~Impl() + {} - // Algorithm to find the shortest edit string. This algorithm + // Algorithm to find the shortest edit string. This algorithm // only needs a linear amount of memory (linear with the sum // of the number of characters in the two paragraph-lists). bool diff(Buffer const * new_buf, Buffer const * old_buf, @@ -206,39 +250,73 @@ public: /// Set to true to cancel the algorithm bool abort_; + /// + QString status() + { + QString status; + status += toqstr("recursion level:") + " " + QString::number(recursion_level_) + + " " + toqstr("differences:") + " " + QString::number(D_); + return status; + } + private: /// Finds the middle snake and returns the length of the /// shortest edit script. - int find_middle_snake(DocRangePair const & rp, DocPair & middle_snake); + int findMiddleSnake(DocRangePair const & rp, DocPair & middle_snake); + + enum SnakeResult { + NoSnake, + SingleSnake, + NormalSnake + }; + + /// Retrieve the middle snake when there is overlap between + /// the forward and backward path. + SnakeResult retrieveMiddleSnake(int k, int D, Direction direction, + DocPair & middle_snake); + + /// Find the furthest reaching D-path (number of horizontal + /// and vertical steps; differences between the old and new + /// document) in the k-diagonal (vertical minus horizontal steps). + void furthestDpathKdiagonal(int D, int k, + DocRangePair const & rp, Direction direction); + + /// Is there overlap between the forward and backward path + bool overlap(int k, int D); /// This function is called recursively by a divide and conquer /// algorithm. Each time, the string is divided into two split /// around the middle snake. void diff_i(DocRangePair const & rp); - /// Processes the splitted chunks. It either adds them as deleted, + /// Processes the split chunks. It either adds them as deleted, /// as added, or call diff_i for further processing. - void diff_part(DocRangePair const & rp); + void diffPart(DocRangePair const & rp); - /// Runs the algorithm for the inset located at /c it and /c it_n + /// Runs the algorithm for the inset located at /c it and /c it_n /// and adds the result to /c pars. - void diff_inset(Inset * inset, DocPair const & p); + void diffInset(Inset * inset, DocPair const & p); /// Adds the snake to the destination buffer. The algorithm will /// recursively be applied to any InsetTexts that are within the snake. - void process_snake(DocRangePair const & rp); + void processSnake(DocRangePair const & rp); /// Writes the range to the destination buffer void writeToDestBuffer(DocRange const & range, Change::Type type = Change::UNCHANGED); - + /// Writes the paragraph list to the destination buffer void writeToDestBuffer(ParagraphList const & copy_pars) const; /// The length of the old chunk currently processed - int N_; + int n_; /// The length of the new chunk currently processed - int M_; + int m_; + /// The offset diagonal of the reverse path of the + /// currently processed chunk + int offset_reverse_diagonal_; + /// Is the offset odd or even ? + bool odd_offset_; /// The thread object, used to emit signals to the GUI Compare const & compare_; @@ -258,6 +336,21 @@ private: /// The number of nested insets at this level int nested_inset_level_; + + /// The position/snake in the old/new document + /// of the forward/reverse search + compl_vector ofp; + compl_vector nfp; + compl_vector ofs; + compl_vector nfs; + compl_vector orp; + compl_vector nrp; + compl_vector ors; + compl_vector nrs; + + /// The number of differences in the path the algorithm + /// is currently processing. + int D_; }; ///////////////////////////////////////////////////////////////////// @@ -271,6 +364,15 @@ Compare::Compare(Buffer const * new_buf, Buffer const * old_buf, : new_buffer(new_buf), old_buffer(old_buf), dest_buffer(dest_buf), options_(options), pimpl_(new Impl(*this)) { + connect(&status_timer_, SIGNAL(timeout()), + this, SLOT(doStatusMessage())); + status_timer_.start(1000); +} + + +void Compare::doStatusMessage() +{ + statusMessage(pimpl_->status()); } @@ -279,14 +381,37 @@ void Compare::run() if (!dest_buffer || !new_buffer || !old_buffer) return; - // Copy the buffer params to the new buffer + // Copy the buffer params to the destination buffer dest_buffer->params() = options_.settings_from_new ? new_buffer->params() : old_buffer->params(); - - // do the real work + // Copy extra authors to the destination buffer + AuthorList const & extra_authors = options_.settings_from_new ? + old_buffer->params().authors() : new_buffer->params().authors(); + AuthorList::Authors::const_iterator it = extra_authors.begin(); + for (; it != extra_authors.end(); ++it) + dest_buffer->params().authors().record(*it); + + // We will need this later + DocumentClassConstPtr const olddc = + dest_buffer->params().documentClassPtr(); + // We do not want to share the DocumentClass with the other Buffer. + // See bug #10295. + dest_buffer->params().makeDocumentClass(); + + doStatusMessage(); + // Do the real work if (!doCompare()) return; - + + // The comparison routine simply copies the paragraphs over into the + // new buffer with the document class from wherever they came from. + // So we need to reset the document class of all the paragraphs. + // See bug #10295. + ErrorList el; + cap::switchBetweenClasses( + olddc, dest_buffer->params().documentClassPtr(), + static_cast(dest_buffer->inset()), el); + finished(pimpl_->abort_); return; } @@ -307,15 +432,15 @@ void Compare::abort() } -static void get_paragraph_list(DocRange const & range, +static void getParagraphList(DocRange const & range, ParagraphList & pars) { // Clone the paragraphs within the selection. pit_type startpit = range.from.pit(); pit_type endpit = range.to.pit(); ParagraphList const & ps_ = range.text()->paragraphs(); - ParagraphList tmp_pars(boost::next(ps_.begin(), startpit), - boost::next(ps_.begin(), endpit + 1)); + ParagraphList tmp_pars(lyx::next(ps_.begin(), startpit), + lyx::next(ps_.begin(), endpit + 1)); // Remove the end of the last paragraph; afterwards, remove the // beginning of the first paragraph. Keep this order - there may only @@ -338,7 +463,7 @@ static bool equal(Inset const * i_o, Inset const * i_n) if (i_o->lyxCode() != i_n->lyxCode()) return false; - // Editable insets are assumed to be the same as they are of the + // Editable insets are assumed to be the same as they are of the // same type. If we later on decide that we insert them in the // document as being unchanged, we will run the algorithm on the // contents of the two insets. @@ -357,7 +482,15 @@ static bool equal(Inset const * i_o, Inset const * i_n) } -static bool equal(DocIterator & o, DocIterator & n) { +static bool equal(DocIterator & o, DocIterator & n) +{ + // Explicitly check for this, so we won't call + // Paragraph::getChar for the last pos. + bool const o_lastpos = o.pos() == o.lastpos(); + bool const n_lastpos = n.pos() == n.lastpos(); + if (o_lastpos || n_lastpos) + return o_lastpos && n_lastpos; + Paragraph const & old_par = o.text()->getPar(o.pit()); Paragraph const & new_par = n.text()->getPar(n.pit()); @@ -372,7 +505,7 @@ static bool equal(DocIterator & o, DocIterator & n) { if (i_o && i_n) return equal(i_o, i_n); - } + } Font fo = old_par.getFontSettings(o.buffer()->params(), o.pos()); Font fn = new_par.getFontSettings(n.buffer()->params(), n.pos()); @@ -380,15 +513,15 @@ static bool equal(DocIterator & o, DocIterator & n) { } -/// Traverses a snake in a certain direction. p points to a +/// Traverses a snake in a certain direction. p points to a /// position in the old and new file and they are synchronously /// moved along the snake. The function returns true if a snake /// was found. -static bool traverse_snake(DocPair & p, DocRangePair const & range, +static bool traverseSnake(DocPair & p, DocRangePair const & range, Direction direction) { bool ret = false; - DocPair const & p_end = + DocPair const & p_end = direction == Forward ? range.to() : range.from(); while (p != p_end) { @@ -413,12 +546,165 @@ static bool traverse_snake(DocPair & p, DocRangePair const & range, // ///////////////////////////////////////////////////////////////////// -int Compare::Impl::find_middle_snake(DocRangePair const & rp, - DocPair &) + +void Compare::Impl::furthestDpathKdiagonal(int D, int k, + DocRangePair const & rp, Direction direction) +{ + compl_vector & op = direction == Forward ? ofp : orp; + compl_vector & np = direction == Forward ? nfp : nrp; + compl_vector & os = direction == Forward ? ofs : ors; + compl_vector & ns = direction == Forward ? nfs : nrs; + + // A vertical step means stepping one character in the new document. + bool vertical_step = k == -D; + if (!vertical_step && k != D) { + vertical_step = direction == Forward + ? op[k - 1] < op[k + 1] : op[k - 1] > op[k + 1]; + } + + // Where do we take the step from ? + int const kk = vertical_step ? k + 1 : k - 1; + DocPair p(op[kk], np[kk]); + DocPair const s(os[kk], ns[kk]); + + // If D==0 we simulate a vertical step from (0,-1) by doing nothing. + if (D != 0) { + // Take a step + if (vertical_step && direction == Forward) + step(p.n, rp.n.to, direction); + else if (vertical_step && direction == Backward) + step(p.n, rp.n.from, direction); + else if (!vertical_step && direction == Forward) + step(p.o, rp.o.to, direction); + else if (!vertical_step && direction == Backward) + step(p.o, rp.o.from, direction); + } + + // Traverse snake + if (traverseSnake(p, rp, direction)) { + // Record last snake + os[k] = p.o; + ns[k] = p.n; + } else { + // Copy last snake from the previous step + os[k] = s.o; + ns[k] = s.n; + } + + //Record new position + op[k] = p.o; + np[k] = p.n; +} + + +bool Compare::Impl::overlap(int k, int D) +{ + // To generalize for the forward and reverse checks + int kk = offset_reverse_diagonal_ - k; + + // Can we have overlap ? + if (kk <= D && kk >= -D) { + // Do we have overlap ? + if (odd_offset_) + return ofp[k] >= orp[kk] && nfp[k] >= nrp[kk]; + else + return ofp[kk] >= orp[k] && nfp[kk] >= nrp[k]; + } + return false; +} + + +Compare::Impl::SnakeResult Compare::Impl::retrieveMiddleSnake( + int k, int D, Direction direction, DocPair & middle_snake) { - N_ = rp.o.length(); - M_ = rp.n.length(); - return M_ + N_; + compl_vector & os = direction == Forward ? ofs : ors; + compl_vector & ns = direction == Forward ? nfs : nrs; + compl_vector & os_r = direction == Forward ? ors : ofs; + compl_vector & ns_r = direction == Forward ? nrs : nfs; + + // The diagonal while doing the backward search + int kk = -k + offset_reverse_diagonal_; + + // Did we find a snake ? + if (os[k].empty() && os_r[kk].empty()) { + // No, there is no snake at all, in which case + // the length of the shortest edit script is M+N. + LATTEST(2 * D - odd_offset_ == m_ + n_); + return NoSnake; + } + + if (os[k].empty()) { + // Yes, but there is only 1 snake and we found it in the + // reverse path. + middle_snake.o = os_r[kk]; + middle_snake.n = ns_r[kk]; + return SingleSnake; + } + + middle_snake.o = os[k]; + middle_snake.n = ns[k]; + return NormalSnake; +} + + +int Compare::Impl::findMiddleSnake(DocRangePair const & rp, + DocPair & middle_snake) +{ + // The lengths of the old and new chunks. + n_ = rp.o.length(); + m_ = rp.n.length(); + + // Forward paths are centered around the 0-diagonal; reverse paths + // are centered around the diagonal N - M. (Delta in the article) + offset_reverse_diagonal_ = n_ - m_; + + // If the offset is odd, only check for overlap while extending forward + // paths, otherwise only check while extending reverse paths. + odd_offset_ = (offset_reverse_diagonal_ % 2 != 0); + + ofp.reset(rp.o.from); + nfp.reset(rp.n.from); + ofs.reset(DocIterator()); + nfs.reset(DocIterator()); + orp.reset(rp.o.to); + nrp.reset(rp.n.to); + ors.reset(DocIterator()); + nrs.reset(DocIterator()); + + // In the formula below, the "+ 1" ensures we round like ceil() + int const D_max = (m_ + n_ + 1)/2; + // D is the number of horizontal and vertical steps, i.e. + // different characters in the old and new chunk. + for (int D = 0; D <= D_max; ++D) { + // to be used in the status messages + D_ = D; + + // Forward and reverse paths + for (int f = 0; f < 2; ++f) { + Direction direction = f == 0 ? Forward : Backward; + + // Diagonals between -D and D can be reached by a D-path + for (int k = -D; k <= D; k += 2) { + // Find the furthest reaching D-path on this diagonal + furthestDpathKdiagonal(D, k, rp, direction); + + // Only check for overlap for forward paths if the offset is odd + // and only for reverse paths if the offset is even. + if (odd_offset_ == (direction == Forward)) { + + // Do the forward and backward paths overlap ? + if (overlap(k, D - odd_offset_)) { + retrieveMiddleSnake(k, D, direction, middle_snake); + return 2 * D - odd_offset_; + } + } + if (abort_) + return 0; + } + } + } + // This should never be reached + return -2; } @@ -440,12 +726,14 @@ bool Compare::Impl::diff(Buffer const * new_buf, Buffer const * old_buf, DocRangePair rp(old_buf_, new_buf_); DocPair from = rp.from(); - traverse_snake(from, rp, Forward); + traverseSnake(from, rp, Forward); DocRangePair const snake(rp.from(), from); - process_snake(snake); - + processSnake(snake); + // Start the recursive algorithm - diff_i(rp); + DocRangePair rp_new(from, rp.to()); + if (!rp_new.o.empty() || !rp_new.n.empty()) + diff_i(rp_new); for (pit_type p = 0; p < (pit_type)dest_pars_->size(); ++p) { (*dest_pars_)[p].setBuffer(const_cast(*dest_buf)); @@ -458,12 +746,15 @@ bool Compare::Impl::diff(Buffer const * new_buf, Buffer const * old_buf, void Compare::Impl::diff_i(DocRangePair const & rp) { + if (abort_) + return; + // The middle snake DocPair middle_snake; // Divides the problem into two smaller problems, split around // the snake in the middle. - int const L_ses = find_middle_snake(rp, middle_snake); + int const L_ses = findMiddleSnake(rp, middle_snake); // Set maximum of progress bar if (++recursion_level_ == 1) @@ -485,35 +776,35 @@ void Compare::Impl::diff_i(DocRangePair const & rp) } else { // Retrieve the complete snake DocPair first_part_end = middle_snake; - traverse_snake(first_part_end, rp, Backward); + traverseSnake(first_part_end, rp, Backward); DocRangePair first_part(rp.from(), first_part_end); - + DocPair second_part_begin = middle_snake; - traverse_snake(second_part_begin, rp, Forward); + traverseSnake(second_part_begin, rp, Forward); DocRangePair second_part(second_part_begin, rp.to()); - + // Split the string in three parts: // 1. in front of the snake - diff_part(first_part); + diffPart(first_part); // 2. the snake itself, and DocRangePair const snake(first_part.to(), second_part.from()); - process_snake(snake); + processSnake(snake); // 3. behind the snake. - diff_part(second_part); + diffPart(second_part); } --recursion_level_; } -void Compare::Impl::diff_part(DocRangePair const & rp) +void Compare::Impl::diffPart(DocRangePair const & rp) { // Is there a finite length string in both buffers, if not there // is an empty string and we write the other one to the buffer. if (!rp.o.empty() && !rp.n.empty()) diff_i(rp); - + else if (!rp.o.empty()) writeToDestBuffer(rp.o, Change::DELETED); @@ -522,7 +813,7 @@ void Compare::Impl::diff_part(DocRangePair const & rp) } -void Compare::Impl::diff_inset(Inset * inset, DocPair const & p) +void Compare::Impl::diffInset(Inset * inset, DocPair const & p) { // Find the dociterators for the beginning and the // end of the inset, for the old and new document. @@ -533,7 +824,7 @@ void Compare::Impl::diff_inset(Inset * inset, DocPair const & p) ParagraphList * backup_dest_pars = dest_pars_; dest_pars_ = &inset->asInsetText()->text().paragraphs(); dest_pars_->clear(); - + ++nested_inset_level_; diff_i(rp); --nested_inset_level_; @@ -542,10 +833,10 @@ void Compare::Impl::diff_inset(Inset * inset, DocPair const & p) } -void Compare::Impl::process_snake(DocRangePair const & rp) +void Compare::Impl::processSnake(DocRangePair const & rp) { ParagraphList pars; - get_paragraph_list(rp.o, pars); + getParagraphList(rp.o, pars); // Find insets in this paragaph list DocPair it = rp.from(); @@ -558,8 +849,8 @@ void Compare::Impl::process_snake(DocRangePair const & rp) pit_type const pit = it.o.pit() - rp.o.from.pit(); pos_type const pos = pit ? it.o.pos() : it.o.pos() - rp.o.from.pos(); inset = pars[pit].getInset(pos); - LASSERT(inset, /**/); - diff_inset(inset, it); + LASSERT(inset, continue); + diffInset(inset, it); } } writeToDestBuffer(pars); @@ -570,7 +861,7 @@ void Compare::Impl::writeToDestBuffer(DocRange const & range, Change::Type type) { ParagraphList pars; - get_paragraph_list(range, pars); + getParagraphList(range, pars); pos_type size = 0;