speedup buffer parsing by a factor of 3

git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@4270 a592a061-630c-0410-9148-cb99ea01b6c8

diff --git a/src/ChangeLog b/src/ChangeLog
index ea172d2b6eed4b816b8e1099e18a3e6168ff2179..eaaa1ba9f927a87d1a48bbf20409d64bddcdf301 100644 (file)
--- a/src/ChangeLog
+++ b/src/ChangeLog
@@ -1,10 +1,18 @@
- 
+2002-05-27  Jean-Marc Lasgouttes  <lasgouttes@freesurf.fr>
+
+       * paragraph_pimpl.C (insertChar): do not try to update tables when
+       appending (pos == size())
+
+       * buffer.C (parseSingleLyXformat2Token): reorder a bit the tests
+       in order to reduce drastically the number of comparisons needed to
+       parse a large document
+
 2002-05-29  André Pönitz <poenitz@gmx.net>
 
        * text.C:
        * text2.C:
        * lyxtextclass.C:
-       *       sp_pspell.h:
+       * sp_pspell.h:
        * textclasslist.[Ch]:
        * sp_ispell.h: whitespace change
 

diff --git a/src/buffer.C b/src/buffer.C
index 3da303c63136701d815010840e97f2d198cf8b94..0bf0910091fe12044ebfcc39006fd2ff97ae5ce4 100644 (file)
--- a/src/buffer.C
+++ b/src/buffer.C
@@ -484,6 +484,12 @@ Buffer::parseSingleLyXformat2Token(LyXLex & lex, Paragraph *& par,
        static Paragraph * parBeforeMinipage;
 #endif
 #endif
+
+       // The order of the tags tested may seem unnatural, but this
+       // has been done in order to reduce the number of string
+       // comparisons needed to recognize a given token. This leads
+       // on large documents like UserGuide to a reduction of a
+       // factor 5! (JMarc)
        if (token[0] != '\\') {
 #ifndef NO_COMPABILITY
                if (ert_comp.active) {
@@ -498,11 +504,6 @@ Buffer::parseSingleLyXformat2Token(LyXLex & lex, Paragraph *& par,
 #ifndef NO_COMPABILITY
                }
 #endif
-       } else if (token == "\\i") {
-               Inset * inset = new InsetLatexAccent;
-               inset->read(this, lex);
-               par->insertInset(pos, inset, font);
-               ++pos;
        } else if (token == "\\layout") {
 #ifndef NO_COMPABILITY
                bool old_fromlayout = ert_comp.fromlayout;
@@ -644,6 +645,134 @@ Buffer::parseSingleLyXformat2Token(LyXLex & lex, Paragraph *& par,
                }
 #endif
 
+       } else if (token == "\\end_inset") {
+               lyxerr << "Solitary \\end_inset. Missing \\begin_inset?.\n"
+                      << "Last inset read was: " << last_inset_read
+                      << endl;
+               // Simply ignore this. The insets do not have
+               // to read this.
+               // But insets should read it, it is a part of
+               // the inset isn't it? Lgb.
+       } else if (token == "\\begin_inset") {
+#ifndef NO_COMPABILITY
+               insertErtContents(par, pos, false);
+               ert_stack.push(ert_comp);
+               ert_comp = ErtComp();
+#endif
+               readInset(lex, par, pos, font);
+#ifndef NO_COMPABILITY
+               ert_comp = ert_stack.top();
+               ert_stack.pop();
+               insertErtContents(par, pos);
+#endif
+       } else if (token == "\\family") {
+               lex.next();
+               font.setLyXFamily(lex.getString());
+       } else if (token == "\\series") {
+               lex.next();
+               font.setLyXSeries(lex.getString());
+       } else if (token == "\\shape") {
+               lex.next();
+               font.setLyXShape(lex.getString());
+       } else if (token == "\\size") {
+               lex.next();
+               font.setLyXSize(lex.getString());
+#ifndef NO_COMPABILITY
+       } else if (token == "\\latex") {
+               lex.next();
+               string const tok = lex.getString();
+               if (tok == "no_latex") {
+                       // Do the insetert.
+                       insertErtContents(par, pos);
+               } else if (tok == "latex") {
+                       ert_comp.active = true;
+                       ert_comp.font = font;
+               } else if (tok == "default") {
+                       // Do the insetert.
+                       insertErtContents(par, pos);
+               } else {
+                       lex.printError("Unknown LaTeX font flag "
+                                      "`$$Token'");
+               }
+#endif
+       } else if (token == "\\lang") {
+               lex.next();
+               string const tok = lex.getString();
+               Language const * lang = languages.getLanguage(tok);
+               if (lang) {
+                       font.setLanguage(lang);
+               } else {
+                       font.setLanguage(params.language);
+                       lex.printError("Unknown language `$$Token'");
+               }
+       } else if (token == "\\numeric") {
+               lex.next();
+               font.setNumber(font.setLyXMisc(lex.getString()));
+       } else if (token == "\\emph") {
+               lex.next();
+               font.setEmph(font.setLyXMisc(lex.getString()));
+       } else if (token == "\\bar") {
+               lex.next();
+               string const tok = lex.getString();
+               // This is dirty, but gone with LyX3. (Asger)
+               if (tok == "under")
+                       font.setUnderbar(LyXFont::ON);
+               else if (tok == "no")
+                       font.setUnderbar(LyXFont::OFF);
+               else if (tok == "default")
+                       font.setUnderbar(LyXFont::INHERIT);
+               else
+                       lex.printError("Unknown bar font flag "
+                                      "`$$Token'");
+       } else if (token == "\\noun") {
+               lex.next();
+               font.setNoun(font.setLyXMisc(lex.getString()));
+       } else if (token == "\\color") {
+               lex.next();
+               font.setLyXColor(lex.getString());
+       } else if (token == "\\SpecialChar") {
+               LyXLayout const & layout =
+                       textclasslist[params.textclass][par->layout()];
+
+               // Insets don't make sense in a free-spacing context! ---Kayvan
+               if (layout.free_spacing || par->isFreeSpacing()) {
+                       if (lex.isOK()) {
+                               lex.next();
+                               string next_token = lex.getString();
+                               if (next_token == "\\-") {
+                                       par->insertChar(pos, '-', font);
+                               } else if (next_token == "\\protected_separator"
+                                       || next_token == "~") {
+                                       par->insertChar(pos, ' ', font);
+                               } else {
+                                       lex.printError("Token `$$Token' "
+                                                      "is in free space "
+                                                      "paragraph layout!");
+                                       --pos;
+                               }
+                       }
+               } else {
+                       Inset * inset = new InsetSpecialChar;
+                       inset->read(this, lex);
+                       par->insertInset(pos, inset, font);
+               }
+               ++pos;
+       } else if (token == "\\i") {
+               Inset * inset = new InsetLatexAccent;
+               inset->read(this, lex);
+               par->insertInset(pos, inset, font);
+               ++pos;
+       } else if (token == "\\backslash") {
+#ifndef NO_COMPABILITY
+               if (ert_comp.active) {
+                       ert_comp.contents += "\\";
+               } else {
+#endif
+               par->insertChar(pos, '\\', font);
+               ++pos;
+#ifndef NO_COMPABILITY
+               }
+#endif
 #ifndef NO_COMPABILITY
        } else if (token == "\\begin_float") {
                insertErtContents(par, pos);
@@ -1040,79 +1169,6 @@ Buffer::parseSingleLyXformat2Token(LyXLex & lex, Paragraph *& par,
        } else if (token == "\\float_placement") {
                lex.nextToken();
                params.float_placement = lex.getString();
-       } else if (token == "\\family") {
-               lex.next();
-               font.setLyXFamily(lex.getString());
-       } else if (token == "\\series") {
-               lex.next();
-               font.setLyXSeries(lex.getString());
-       } else if (token == "\\shape") {
-               lex.next();
-               font.setLyXShape(lex.getString());
-       } else if (token == "\\size") {
-               lex.next();
-               font.setLyXSize(lex.getString());
-#ifndef NO_COMPABILITY
-       } else if (token == "\\latex") {
-               lex.next();
-               string const tok = lex.getString();
-               if (tok == "no_latex") {
-                       // Do the insetert.
-                       insertErtContents(par, pos);
-               } else if (tok == "latex") {
-                       ert_comp.active = true;
-                       ert_comp.font = font;
-               } else if (tok == "default") {
-                       // Do the insetert.
-                       insertErtContents(par, pos);
-               } else {
-                       lex.printError("Unknown LaTeX font flag "
-                                      "`$$Token'");
-               }
-#endif
-       } else if (token == "\\lang") {
-               lex.next();
-               string const tok = lex.getString();
-               Language const * lang = languages.getLanguage(tok);
-               if (lang) {
-                       font.setLanguage(lang);
-               } else {
-                       font.setLanguage(params.language);
-                       lex.printError("Unknown language `$$Token'");
-               }
-#ifndef NO_COMPABILITY
-               // if the contents is still empty we're just behind the
-               // latex font change and so this inset should also be
-               // inserted with that language tag!
-               if (ert_comp.active && ert_comp.contents.empty()) {
-                       ert_comp.font.setLanguage(font.language());
-               }
-#endif
-       } else if (token == "\\numeric") {
-               lex.next();
-               font.setNumber(font.setLyXMisc(lex.getString()));
-       } else if (token == "\\emph") {
-               lex.next();
-               font.setEmph(font.setLyXMisc(lex.getString()));
-       } else if (token == "\\bar") {
-               lex.next();
-               string const tok = lex.getString();
-               // This is dirty, but gone with LyX3. (Asger)
-               if (tok == "under")
-                       font.setUnderbar(LyXFont::ON);
-               else if (tok == "no")
-                       font.setUnderbar(LyXFont::OFF);
-               else if (tok == "default")
-                       font.setUnderbar(LyXFont::INHERIT);
-               else
-                       lex.printError("Unknown bar font flag "
-                                      "`$$Token'");
-       } else if (token == "\\noun") {
-               lex.next();
-               font.setNoun(font.setLyXMisc(lex.getString()));
-       } else if (token == "\\color") {
-               lex.next();
-               font.setLyXColor(lex.getString());
        } else if (token == "\\align") {
                int tmpret = lex.findToken(string_align);
                if (tmpret == -1) ++tmpret;
@@ -1159,52 +1215,6 @@ Buffer::parseSingleLyXformat2Token(LyXLex & lex, Paragraph *& par,
                lex.eatLine();
                par->params().labelWidthString(lex.getString());
                // do not delete this token, it is still needed!
-       } else if (token == "\\end_inset") {
-               lyxerr << "Solitary \\end_inset. Missing \\begin_inset?.\n"
-                      << "Last inset read was: " << last_inset_read
-                      << endl;
-               // Simply ignore this. The insets do not have
-               // to read this.
-               // But insets should read it, it is a part of
-               // the inset isn't it? Lgb.
-       } else if (token == "\\begin_inset") {
-#ifndef NO_COMPABILITY
-               insertErtContents(par, pos, false);
-               ert_stack.push(ert_comp);
-               ert_comp = ErtComp();
-#endif
-               readInset(lex, par, pos, font);
-#ifndef NO_COMPABILITY
-               ert_comp = ert_stack.top();
-               ert_stack.pop();
-#endif
-       } else if (token == "\\SpecialChar") {
-               LyXLayout const & layout =
-                       textclasslist[params.textclass][par->layout()];
-
-               // Insets don't make sense in a free-spacing context! ---Kayvan
-               if (layout.free_spacing || par->isFreeSpacing()) {
-                       if (lex.isOK()) {
-                               lex.next();
-                               string next_token = lex.getString();
-                               if (next_token == "\\-") {
-                                       par->insertChar(pos, '-', font);
-                               } else if (next_token == "\\protected_separator"
-                                       || next_token == "~") {
-                                       par->insertChar(pos, ' ', font);
-                               } else {
-                                       lex.printError("Token `$$Token' "
-                                                      "is in free space "
-                                                      "paragraph layout!");
-                                       --pos;
-                               }
-                       }
-               } else {
-                       Inset * inset = new InsetSpecialChar;
-                       inset->read(this, lex);
-                       par->insertInset(pos, inset, font);
-               }
-               ++pos;
        } else if (token == "\\newline") {
 #ifndef NO_COMPABILITY
                if (!ert_comp.in_tabular && ert_comp.active) {
@@ -1253,17 +1263,6 @@ Buffer::parseSingleLyXformat2Token(LyXLex & lex, Paragraph *& par,
                        par->bibkey = new InsetBibKey(p);
                }
                par->bibkey->read(this, lex);
-       } else if (token == "\\backslash") {
-#ifndef NO_COMPABILITY
-               if (ert_comp.active) {
-                       ert_comp.contents += "\\";
-               } else {
-#endif
-               par->insertChar(pos, '\\', font);
-               ++pos;
-#ifndef NO_COMPABILITY
-               }
-#endif
        } else if (token == "\\the_end") {
 #ifndef NO_COMPABILITY
                // If we still have some ert active here we have to insert

diff --git a/src/paragraph_pimpl.C b/src/paragraph_pimpl.C
index 6f311fe2373301342de0f0fdf9f59bc166436814..30de048e1d60479cc9a046b66f821e09b051195a 100644 (file)
--- a/src/paragraph_pimpl.C
+++ b/src/paragraph_pimpl.C
@@ -112,6 +112,15 @@ void Paragraph::Pimpl::insertChar(pos_type pos, value_type c,
 {
        lyx::Assert(pos <= size());
 
+       // This is actually very common when parsing buffers (and
+       // maybe inserting ascii text)
+       if (pos == size()) {
+               // when appending characters, no need to update tables
+               text.push_back(c);
+               owner_->setFont(pos, font);
+               return;
+       }
+
        text.insert(text.begin() + pos, c);
 
        // Update the font table.