Fix bug #6058: Change tracking and versioning.

Computes a hash value for the authors when using change tracking.

Text.cpp, BufferParams.h, Author.h:
  Change unsigned int to int because the hash values can be negative.

lyx2lyx:
  Allow to convert negative author_ids.

See also: r30756.

git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@36134 a592a061-630c-0410-9148-cb99ea01b6c8

development/FORMAT

@@ -6,6 +6,13 @@ changes happened in particular if possible.
 The good example would be 2010-01-10 entry.
 
 -----------------------
+2010-10-23 Vincent van Ravesteijn <vfr@lyx.org>
+	* Format incremented to 405: author hash numbers
+	  The authors that are used in change tracking are
+	  now identified in the file by a number that represents
+	  the hash value of the name and email. In this way
+	  collaboration using version control leads to way less
+	  merge conflicts.
 
 2010-10-13 Richard Heck <rgheck@comcast.net>
 	* Format incremented to 404: refstyle support

lib/lyx2lyx/lyx_2_0.py

@@ -771,7 +771,7 @@ def revert_author_id(document):
     " Remove the author_id from the \\author definition "
     i = 0
     anum = 0
-    rx = re.compile(r'(\\author)\s+(\d+)\s+(\".*\")\s*(.*)$')
+    rx = re.compile(r'(\\author)\s+([-\d]+)\s+(\".*\")\s*(.*)$')
     idmap = dict()
 
     while True:
@@ -2013,10 +2013,12 @@ convert = [[346, []],
            [401, []],
            [402, [convert_bibtex_clearpage]],
            [403, [convert_flexnames]],
-           [404, [convert_prettyref]]
+           [404, [convert_prettyref]],
+           [405, []]
 ]
 
-revert =  [[403, [revert_refstyle]],
+revert =  [[404, []],
+           [403, [revert_refstyle]],
            [402, [revert_flexnames]],
            [401, []],
            [400, [revert_diagram]],

src/Author.cpp

@@ -12,9 +12,8 @@
 
 #include "Author.h"
 
-#include "support/lstrings.h"
-
 #include "support/lassert.h"
+#include "support/lstrings.h"
 
 #include <algorithm>
 #include <istream>
@@ -24,6 +23,24 @@ using namespace lyx::support;
 
 namespace lyx {
 
+static int computeHash(docstring const & name,
+	docstring const & email)
+{
+	string const full_author_string = to_ascii(name + email);
+	// Bernstein's hash function
+	unsigned int hash = 5381;
+	for (unsigned int i = 0; i < full_author_string.length(); ++i)
+		hash = ((hash << 5) + hash) + unsigned int(full_author_string[i]);
+	return int(hash);
+}
+
+
+Author::Author(docstring const & name, docstring const & email)
+	: name_(name), email_(email), used_(true)
+{
+	buffer_id_ = computeHash(name_, email_);
+}
+
 
 bool operator==(Author const & l, Author const & r)
 {
@@ -65,18 +82,18 @@ AuthorList::AuthorList()
 
 int AuthorList::record(Author const & a)
 {
+	// If we record an author which equals the current
+	// author, we copy the buffer_id, so that it will
+	// keep the same id in the file.
+	if (authors_.size() > 0 && a == authors_[0])
+		authors_[0].setBufferId(a.buffer_id());
+
 	Authors::const_iterator it(authors_.begin());
 	Authors::const_iterator itend(authors_.end());
-
 	for (int i = 0;  it != itend; ++it, ++i) {
-		if (*it == a) {
-			if (it->buffer_id() == 0)
-				// The current author is internally represented as 
-				// author 0, but it appears he has already an id.
-				it->setBufferId(a.buffer_id());
+		if (*it == a)
 			return i;
 	}
-	}
 	authors_.push_back(a);
 	return last_id_++;
 }
@@ -129,24 +146,6 @@ ostream & operator<<(ostream & os, AuthorList const & a) {
 	AuthorList::Authors::const_iterator a_it = sorted.begin();
 	AuthorList::Authors::const_iterator a_end = sorted.end();
 	
-	// Find the buffer id for the current author (internal id 0),
-	// if he doesn't have a buffer_id yet.
-	if (sorted.get(0).buffer_id() == 0) {
-		unsigned int cur_id = 1;
-		for (; a_it != a_end; ++a_it) {
-			if (a_it->buffer_id() == cur_id)
-				++cur_id;
-			else if (a_it->buffer_id() > cur_id) {
-				break;
-			}
-		}
-		// Set the id in both the original authorlist, 
-		// as in the copy.
-		a.get(0).setBufferId(cur_id);
-		sorted.get(0).setBufferId(cur_id);
-		sorted.sort();
-	}
-	
 	for (a_it = sorted.begin(); a_it != a_end; ++a_it) {
 		if (a_it->used())
 			os << "\\author " << *a_it << "\n";	

src/Author.h

@@ -24,16 +24,15 @@ public:
 	///
 	Author() {}
 	///
-	Author(docstring const & name, docstring const & email)
-		: name_(name), email_(email), used_(true), buffer_id_(0) {}
+	Author(docstring const & name, docstring const & email);
 	///
 	docstring name() const { return name_; }
 	///
 	docstring email() const { return email_; }
 	///
-	unsigned int buffer_id() const { return buffer_id_; }
+	int buffer_id() const { return buffer_id_; }
 	///
-	void setBufferId(unsigned int buffer_id) const { buffer_id_ = buffer_id; }
+	void setBufferId(int buffer_id) const { buffer_id_ = buffer_id; }
 	///
 	void setUsed(bool u) const { used_ = u; }
 	///
@@ -49,7 +48,7 @@ private:
 	///
 	mutable bool used_;
 	/// The id of the author in the lyx-file
-	mutable unsigned int buffer_id_;
+	mutable int buffer_id_;
 };
 
 

src/Buffer.cpp

@@ -128,7 +128,7 @@ namespace {
 
 // Do not remove the comment below, so we get merge conflict in
 // independent branches. Instead add your own.
-int const LYX_FORMAT = 404; // rgh: refstyle
+int const LYX_FORMAT = 405; // vfr: author hash
 
 typedef map<string, bool> DepClean;
 typedef map<docstring, pair<InsetLabel const *, Buffer::References> > RefCache;

src/BufferParams.h

@@ -340,7 +340,8 @@ public:
 	AuthorList const & authors() const;
 
 	/// map of the file's author IDs to AuthorList indexes
-	std::map<unsigned int, int> author_map;
+	typedef std::map<int, int> AuthorMap;
+	AuthorMap author_map;
 	/// the buffer's font encoding
 	std::string const font_encoding() const;
 	///

src/Text.cpp

@@ -456,10 +456,10 @@ void Text::readParToken(Paragraph & par, Lexer & lex,
 	} else if (token == "\\change_inserted" || token == "\\change_deleted") {
 		lex.eatLine();
 		istringstream is(lex.getString());
-		unsigned int aid;
+		int aid;
 		time_t ct;
 		is >> aid >> ct;
-		map<unsigned int, int> const & am = bp.author_map;
+		BufferParams::AuthorMap const & am = bp.author_map;
 		if (am.find(aid) == am.end()) {
 			errorList.push_back(ErrorItem(_("Change tracking error"),
 					    bformat(_("Unknown author index for change: %1$d\n"), aid),