Use up to 8 affixes in Hunspell spellcheck suggestions

google.patch

 Index: src/hunspell/affixmgr.cxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/affixmgr.cxx,v
 retrieving revision 1.41
 diff -u -r1.41 affixmgr.cxx
 --- src/hunspell/affixmgr.cxx   16 Dec 2011 09:15:34 -0000      1.41
-+++ src/hunspell/affixmgr.cxx   8 Jan 2013 00:30:25 -0000
++++ src/hunspell/affixmgr.cxx   27 Feb 2013 01:10:35 -0000
 @@ -14,8 +14,14 @@
  
  #include "csutil.hxx"
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +AffixMgr::AffixMgr(hunspell::BDictReader* reader, HashMgr** ptr, int * md)
 +{
 +  bdict_reader = reader;
 +#else
  AffixMgr::AffixMgr(const char * affpath, HashMgr** ptr, int * md, const char * key) 
@@ skipping 170 matching lines @@
 +#endif
                      break; 
                      }
               // piece 3 - is cross product indicator 
 Index: src/hunspell/affixmgr.hxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/affixmgr.hxx,v
 retrieving revision 1.15
 diff -u -r1.15 affixmgr.hxx
 --- src/hunspell/affixmgr.hxx   13 Oct 2011 13:41:54 -0000      1.15
-+++ src/hunspell/affixmgr.hxx   8 Jan 2013 00:30:25 -0000
++++ src/hunspell/affixmgr.hxx   27 Feb 2013 01:10:35 -0000
 @@ -18,6 +18,40 @@
  class PfxEntry;
  class SfxEntry;
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +
 +#include <vector>
 +
 +// This class provides an implementation of the contclasses array in AffixMgr
 +// that is normally a large static array. We should almost never need more than
@@ skipping 64 matching lines @@
  };
  
  #endif
 -
 Index: src/hunspell/filemgr.cxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/filemgr.cxx,v
 retrieving revision 1.5
 diff -u -r1.5 filemgr.cxx
 --- src/hunspell/filemgr.cxx    23 Jun 2011 09:21:50 -0000      1.5
-+++ src/hunspell/filemgr.cxx    8 Jan 2013 00:30:25 -0000
++++ src/hunspell/filemgr.cxx    27 Feb 2013 01:10:35 -0000
 @@ -7,6 +7,32 @@
  
  #include "filemgr.hxx"
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +#include "third_party/hunspell/google/bdict_reader.h"
 +
 +FileMgr::FileMgr(hunspell::LineIterator* iterator) : iterator_(iterator) {
 +}
 +
@@ skipping 24 matching lines @@
  int FileMgr::getlinenum() {
      return linenum;
  }
 +#endif
 Index: src/hunspell/filemgr.hxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/filemgr.hxx,v
 retrieving revision 1.3
 diff -u -r1.3 filemgr.hxx
 --- src/hunspell/filemgr.hxx    15 Apr 2010 11:22:08 -0000      1.3
-+++ src/hunspell/filemgr.hxx    8 Jan 2013 00:30:25 -0000
++++ src/hunspell/filemgr.hxx    27 Feb 2013 01:10:35 -0000
 @@ -7,6 +7,30 @@
  #include "hunzip.hxx"
  #include <stdio.h>
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +namespace hunspell {
 +class LineIterator;
 +}  // namespace hunspell
 +
 +// A class which encapsulates operations of reading a BDICT file.
@@ skipping 22 matching lines @@
      int getlinenum();
  };
  #endif
 +#endif
 Index: src/hunspell/hashmgr.cxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hashmgr.cxx,v
 retrieving revision 1.12
 diff -u -r1.12 hashmgr.cxx
 --- src/hunspell/hashmgr.cxx    23 Jun 2011 09:21:50 -0000      1.12
-+++ src/hunspell/hashmgr.cxx    8 Jan 2013 00:30:25 -0000
++++ src/hunspell/hashmgr.cxx    27 Feb 2013 01:10:35 -0000
 @@ -12,8 +12,14 @@
  
  // build a hash table from a munched word list
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +HashMgr::HashMgr(hunspell::BDictReader* reader)
 +{
 +  bdict_reader = reader;
 +#else
  HashMgr::HashMgr(const char * tpath, const char * apath, const char * key)
@@ skipping 349 matching lines @@
 +
  int HashMgr::is_aliasf() {
      return (aliasf != NULL);
  }
 Index: src/hunspell/hashmgr.hxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hashmgr.hxx,v
 retrieving revision 1.3
 diff -u -r1.3 hashmgr.hxx
 --- src/hunspell/hashmgr.hxx    15 Apr 2010 11:22:08 -0000      1.3
-+++ src/hunspell/hashmgr.hxx    8 Jan 2013 00:30:25 -0000
++++ src/hunspell/hashmgr.hxx    27 Feb 2013 01:10:35 -0000
 @@ -8,10 +8,25 @@
  #include "htypes.hxx"
  #include "filemgr.hxx"
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +#include <string>
 +#include <map>
 +
 +#include "base/stl_util.h"
 +#include "base/string_piece.h"
@@ skipping 77 matching lines @@
 +
    int add_hidden_capitalized_word(char * word, int wbl, int wcl,
      unsigned short * flags, int al, char * dp, int captype);
    int parse_aliasm(char * line, FileMgr * af);
 Index: src/hunspell/htypes.hxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/htypes.hxx,v
 retrieving revision 1.3
 diff -u -r1.3 htypes.hxx
 --- src/hunspell/htypes.hxx     6 Sep 2010 07:58:53 -0000       1.3
-+++ src/hunspell/htypes.hxx     8 Jan 2013 00:30:25 -0000
++++ src/hunspell/htypes.hxx     27 Feb 2013 01:10:35 -0000
 @@ -1,6 +1,16 @@
  #ifndef _HTYPES_HXX_
  #define _HTYPES_HXX_
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +// This is a workaround for preventing errors in parsing Turkish BDICs, which
 +// contain very long AF lines (~ 12,000 chars).
 +// TODO(hbono) change the HashMgr::parse_aliasf() function to be able to parse
 +// longer lines than MAXDELEN.
 +#define MAXDELEN    (8192 * 2)
 +#else
 +#define MAXDELEN    8192
 +#endif  // HUNSPELL_CHROME_CLIENT
 +
  #define ROTATE_LEN   5
  
  #define ROTATE(v,q) \
 Index: src/hunspell/hunspell.cxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hunspell.cxx,v
 retrieving revision 1.29
 diff -u -r1.29 hunspell.cxx
 --- src/hunspell/hunspell.cxx   23 Jun 2011 09:21:50 -0000      1.29
-+++ src/hunspell/hunspell.cxx   8 Jan 2013 00:30:25 -0000
++++ src/hunspell/hunspell.cxx   27 Feb 2013 01:10:35 -0000
 @@ -7,20 +7,37 @@
  
  #include "hunspell.hxx"
  #include "hunspell.h"
 +#ifndef HUNSPELL_CHROME_CLIENT
  #ifndef MOZILLA_CLIENT
  #    include "config.h"
  #endif
 +#endif
  #include "csutil.hxx"
@@ skipping 125 matching lines @@
 +#endif
  }
  
  void Hunspell_destroy(Hunhandle *pHunspell)
 Index: src/hunspell/hunspell.hxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hunspell.hxx,v
 retrieving revision 1.6
 diff -u -r1.6 hunspell.hxx
 --- src/hunspell/hunspell.hxx   21 Jan 2011 17:30:41 -0000      1.6
-+++ src/hunspell/hunspell.hxx   8 Jan 2013 00:30:25 -0000
++++ src/hunspell/hunspell.hxx   27 Feb 2013 01:10:35 -0000
 @@ -5,6 +5,10 @@
  #include "suggestmgr.hxx"
  #include "langnum.hxx"
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +#include "third_party/hunspell/google/bdict_reader.h"
 +#endif
 +
  #define  SPELL_XML "<?xml?>"
  
@@ skipping 36 matching lines @@
 +#endif
  
    /* spell(word) - spellcheck word
     * output: 0 = bad word, not 0 = good word
 Index: src/hunspell/replist.hxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/replist.hxx,v
 retrieving revision 1.2
 diff -u -r1.2 replist.hxx
 --- src/hunspell/replist.hxx    15 Apr 2010 11:22:09 -0000      1.2
-+++ src/hunspell/replist.hxx    8 Jan 2013 00:30:25 -0000
++++ src/hunspell/replist.hxx    27 Feb 2013 01:10:35 -0000
 @@ -2,6 +2,12 @@
  #ifndef _REPLIST_HXX_
  #define _REPLIST_HXX_
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +// Compilation issues in spellchecker.cc think near is a macro, therefore
 +// removing it here solves that problem.
 +#undef near
 +#endif
 +
  #include "hunvisapi.h"
  
  #include "w_char.hxx"
 Index: src/hunspell/suggestmgr.cxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/suggestmgr.cxx,v
 retrieving revision 1.24
 diff -u -r1.24 suggestmgr.cxx
 --- src/hunspell/suggestmgr.cxx 14 Feb 2011 21:47:24 -0000      1.24
-+++ src/hunspell/suggestmgr.cxx 8 Jan 2013 00:30:26 -0000
-@@ -12,9 +12,110 @@
++++ src/hunspell/suggestmgr.cxx 27 Feb 2013 01:10:36 -0000
+@@ -12,9 +12,114 @@
  
  const w_char W_VLINE = { '\0', '|' };
  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +namespace {
 +// A simple class which creates temporary hentry objects which are available
 +// only in a scope. To conceal memory operations from SuggestMgr functions,
 +// this object automatically deletes all hentry objects created through
 +// CreateScopedHashEntry() calls in its destructor. So, the following snippet
 +// raises a memory error.
@@ skipping 25 matching lines @@
 +  hentry* CreateScopedHashEntry(int index, const hentry* source);
 +
 + private:
 +  // A struct which encapsulates the new hentry struct introduced in hunspell
 +  // 1.2.8. For a pointer to an hentry struct 'h', hunspell 1.2.8 stores a word
 +  // (including a NUL character) into 'h->word[0]',...,'h->word[h->blen]' even
 +  // though arraysize(h->word[]) is 1. Also, it changed 'astr' to a pointer so
 +  // it can store affix flags into 'h->astr[0]',...,'h->astr[alen-1]'. To handle
 +  // this new hentry struct, we define a struct which combines three values: an
 +  // hentry struct 'hentry'; a char array 'word[kMaxWordLen]', and; an unsigned
-+  // short value 'astr' so a hentry struct 'h' returned from
++  // short array 'astr' so a hentry struct 'h' returned from
 +  // CreateScopedHashEntry() satisfies the following equations:
 +  //   hentry* h = factory.CreateScopedHashEntry(0, source);
 +  //   h->word[0] == ((HashEntryItem*)h)->entry.word[0].
 +  //   h->word[1] == ((HashEntryItem*)h)->word[0].
 +  //   ...
 +  //   h->word[h->blen] == ((HashEntryItem*)h)->word[h->blen-1].
-+  //   h->astr[0] == ((HashEntryItem*)h)->astr.
-+  // Our BDICT does not use affix flags longer than one for now since they are
-+  // discarded by convert_dict, i.e. 'h->astr' is always <= 1. Therefore, this
-+  // struct does not use an array for 'astr'.
++  //   h->astr[0] == ((HashEntryItem*)h)->astr[0].
++  //   h->astr[1] == ((HashEntryItem*)h)->astr[1].
++  //   ...
++  //   h->astr[h->alen-1] == ((HashEntryItem*)h)->astr[h->alen-1].
 +  enum {
 +    kMaxWordLen = 128,
++    kMaxAffixLen = 8,
 +  };
 +  struct HashEntryItem {
 +    hentry entry;
 +    char word[kMaxWordLen];
-+    unsigned short astr;
++    unsigned short astr[kMaxAffixLen];
 +  };
 +
 +  HashEntryItem hash_items_[MAX_ROOTS];
 +};
 +
 +ScopedHashEntryFactory::ScopedHashEntryFactory() {
 +  memset(&hash_items_[0], 0, sizeof(hash_items_));
 +}
 +
 +ScopedHashEntryFactory::~ScopedHashEntryFactory() {
 +}
 +
 +hentry* ScopedHashEntryFactory::CreateScopedHashEntry(int index,
 +                                                      const hentry* source) {
-+  if (index >= MAX_ROOTS || source->blen >= kMaxWordLen || source->alen > 1)
++  if (index >= MAX_ROOTS || source->blen >= kMaxWordLen)
 +    return NULL;
 +
 +  // Retrieve a HashEntryItem struct from our spool, initialize it, and
 +  // returns the address of its 'hentry' member.
 +  size_t source_size = sizeof(hentry) + source->blen + 1;
 +  HashEntryItem* hash_item = &hash_items_[index];
 +  memcpy(&hash_item->entry, source, source_size);
 +  if (source->astr) {
-+    hash_item->astr = *source->astr;
-+    hash_item->entry.astr = &hash_item->astr;
++    hash_item->entry.alen = source->alen;
++    if (hash_item->entry.alen > kMaxAffixLen)
++      hash_item->entry.alen = kMaxAffixLen;
++    memcpy(hash_item->astr, source->astr, hash_item->entry.alen * sizeof(hash_item->astr[0]));
++    hash_item->entry.astr = &hash_item->astr[0];
 +  }
 +  return &hash_item->entry;
 +}
 +
 +}  // namespace
 +#endif
 +
 +
 +#ifdef HUNSPELL_CHROME_CLIENT
 +SuggestMgr::SuggestMgr(hunspell::BDictReader* reader,
 +                       const char * tryme, int maxn, 
 +                       AffixMgr * aptr)
 +{
 +  bdict_reader = reader;
 +#else
  SuggestMgr::SuggestMgr(const char * tryme, int maxn, 
                         AffixMgr * aptr)
  {
 +#endif
  
    // register affix manager and check in string of chars to 
    // try when building candidate suggestions
-@@ -407,6 +508,49 @@
+@@ -407,6 +512,49 @@
    int lenr, lenp;
    int wl = strlen(word);
    if (wl < 2 || ! pAMgr) return ns;
 +  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +  const char *pattern, *pattern2;
 +  hunspell::ReplacementIterator iterator = bdict_reader->GetReplacementIterator();
 +  while (iterator.GetNext(&pattern, &pattern2)) {
 +      r = word;
 +      lenr = strlen(pattern2);
@@ skipping 29 matching lines @@
 +              sp = strchr(prev, ' ');
 +            }
 +          }
 +          r++; // search for the next letter
 +    }
 +  }
 +#else
    int numrep = pAMgr->get_numrep();
    struct replentry* reptable = pAMgr->get_reptable();
    if (reptable==NULL) return ns;
-@@ -448,6 +592,7 @@
+@@ -448,6 +596,7 @@
            r++; // search for the next letter
        }
     }
 +#endif
     return ns;
  }
  
-@@ -678,7 +823,9 @@
+@@ -678,7 +827,9 @@
  // error is missing a letter it needs
  int SuggestMgr::forgotchar(char ** wlst, const char * word, int ns, int cpdsuggest)
  {
 -   char candidate[MAXSWUTF8L];
 +   // TODO(rouslan): Remove the interim change below when this patch lands:
 +   // http://sf.net/tracker/?func=detail&aid=3595024&group_id=143754&atid=756395
 +   char candidate[MAXSWUTF8L + 4];
     char * p;
     clock_t timelimit = clock();
     int timer = MINTIMER;
-@@ -700,8 +847,10 @@
+@@ -700,8 +851,10 @@
  // error is missing a letter it needs
  int SuggestMgr::forgotchar_utf(char ** wlst, const w_char * word, int wl, int ns, int cpdsuggest)
  {
 -   w_char  candidate_utf[MAXSWL];
 -   char    candidate[MAXSWUTF8L];
 +   // TODO(rouslan): Remove the interim change below when this patch lands:
 +   // http://sf.net/tracker/?func=detail&aid=3595024&group_id=143754&atid=756395
 +   w_char  candidate_utf[MAXSWL + 1];
 +   char    candidate[MAXSWUTF8L + 4];
     w_char * p;
     clock_t timelimit = clock();
     int timer = MINTIMER;
-@@ -1057,6 +1206,9 @@
+@@ -1057,6 +1210,9 @@
  
    struct hentry* hp = NULL;
    int col = -1;
 +#ifdef HUNSPELL_CHROME_CLIENT
 +  ScopedHashEntryFactory hash_entry_factory;
 +#endif
    phonetable * ph = (pAMgr) ? pAMgr->get_phonetable() : NULL;
    char target[MAXSWUTF8L];
    char candidate[MAXSWUTF8L];
-@@ -1115,7 +1267,11 @@
+@@ -1115,7 +1271,11 @@
  
      if (sc > scores[lp]) {
        scores[lp] = sc;  
 +#ifdef HUNSPELL_CHROME_CLIENT
 +      roots[lp] = hash_entry_factory.CreateScopedHashEntry(lp, hp);
 +#else
        roots[lp] = hp;
 +#endif
        lval = sc;
        for (j=0; j < MAX_ROOTS; j++)
          if (scores[j] < lval) {
 Index: src/hunspell/suggestmgr.hxx
 ===================================================================
 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/suggestmgr.hxx,v
 retrieving revision 1.5
 diff -u -r1.5 suggestmgr.hxx
 --- src/hunspell/suggestmgr.hxx 21 Jan 2011 22:10:24 -0000      1.5
-+++ src/hunspell/suggestmgr.hxx 8 Jan 2013 00:30:26 -0000
++++ src/hunspell/suggestmgr.hxx 27 Feb 2013 01:10:36 -0000
 @@ -52,7 +52,11 @@
  
  
  public:
 +#ifdef HUNSPELL_CHROME_CLIENT
 +  SuggestMgr(hunspell::BDictReader* reader, const char * tryme, int maxn, AffixMgr *aptr);
 +#else
    SuggestMgr(const char * tryme, int maxn, AffixMgr *aptr);
 +#endif
    ~SuggestMgr();
  
    int suggest(char*** slst, const char * word, int nsug, int * onlycmpdsug);
 @@ -66,6 +70,10 @@
    char * suggest_morph_for_spelling_error(const char * word);
  
  private:
 +#ifdef HUNSPELL_CHROME_CLIENT
 +   // Not owned by us, owned by the Hunspell object.
 +   hunspell::BDictReader* bdict_reader;
 +#endif
     int testsug(char** wlst, const char * candidate, int wl, int ns, int cpdsuggest,
       int * timer, clock_t * timelimit);
     int checkword(const char *, int, int, int *, clock_t *);