tdf#154499 sw spell checking: add 2-word phrase checking

Hunspell dictionaries can contain phrases, i.e. space
separated word sequences, which were used only to reject
compounds and to give better suggestions. Now recognize
2-word phrases in the text, no need to break the phrase
into single words, e.g. "et" and "cetera", which resulted
acceptance of typos (e.g. "et" without "cetera"), also
bad suggestions (e.g. "et" and "cetera" independently from
the context).

More example:

== old .dic file ==
...
et
cetera
von
Neumann
veni
vidi
vici
...

List the 2-word phrases, and break the 3 or more word
into 2-word phrases:

== new .dic file ==
...
et cetera
von Neumann
veni vidi
vidi vici
...

Note: words of the phrase are separated by a space, but
recognized also with punctuation, e.g. in the previous
example, "Veni, vidi, vici."

Note: during typing, the second word of the phrase
will be accepted only at ending the paragraph, i.e.
pressing Enter.

Change-Id: I4a1487abc0e4ab31d09750ee8c817353e6325ca3
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/151487
Tested-by: László Németh <nemeth@numbertext.org>
Reviewed-by: László Németh <nemeth@numbertext.org>

2 files changed, 88 insertions, 39 deletions

diff --git a/sw/inc/swscanner.hxx b/sw/inc/swscanner.hxx
index 5dd118ad0049..88433bfe4bb1 100644
--- a/sw/inc/swscanner.hxx
+++ b/sw/inc/swscanner.hxx
@@ -33,6 +33,7 @@ class SwScanner
 {
     std::function<LanguageType (sal_Int32, sal_Int32, bool)> m_pGetLangOfChar;
     OUString m_aWord;
+    OUString m_aPrevWord;
     const OUString m_aPreDashReplacementText;
     OUString m_aText;
     const LanguageType* m_pLanguage;
@@ -62,6 +63,7 @@ public:
     bool NextWord();
 
     const OUString& GetWord() const    { return m_aWord; }
+    const OUString& GetPrevWord() const { return m_aPrevWord; }
 
     sal_Int32 GetBegin() const         { return m_nBegin; }
     sal_Int32 GetEnd() const           { return m_nBegin + m_nLength; }
diff --git a/sw/source/core/txtnode/txtedt.cxx b/sw/source/core/txtnode/txtedt.cxx
index 528afabed561..1e9082201f7a 100644
--- a/sw/source/core/txtnode/txtedt.cxx
+++ b/sw/source/core/txtnode/txtedt.cxx
@@ -961,6 +961,7 @@ bool SwScanner::NextWord()
     if ( m_nWordType == i18n::WordType::WORD_COUNT )
         m_nLength = forceEachAsianCodePointToWord(m_aText, m_nBegin, m_nLength);
 
+    m_aPrevWord = m_aWord;
     m_aWord = m_aPreDashReplacementText.copy( m_nBegin, m_nLength );
 
     return true;
@@ -1035,8 +1036,11 @@ bool SwTextNode::Spell(SwSpellArgs* pArgs)
         SwScanner aScanner( *this, m_Text, nullptr, ModelToViewHelper(),
                             WordType::DICTIONARY_WORD,
                             nBegin, nEnd );
-        while( !pArgs->xSpellAlt.is() && aScanner.NextWord() )
+        bool bNextWord = aScanner.NextWord();
+        while( !pArgs->xSpellAlt.is() && bNextWord )
         {
+            bool bCalledNextWord = false;
+
             const OUString& rWord = aScanner.GetWord();
 
             // get next language for next word, consider language attributes
@@ -1066,25 +1070,45 @@ bool SwTextNode::Spell(SwSpellArgs* pArgs)
                     }
                     else
                     {
-                        // make sure the selection build later from the data
-                        // below does not include "in word" character to the
-                        // left and right in order to preserve those. Therefore
-                        // count those "in words" in order to modify the
-                        // selection accordingly.
-                        const sal_Unicode* pChar = rWord.getStr();
-                        sal_Int32 nLeft = 0;
-                        while (*pChar++ == CH_TXTATR_INWORD)
-                            ++nLeft;
-                        pChar = rWord.getLength() ? rWord.getStr() + rWord.getLength() - 1 : nullptr;
-                        sal_Int32 nRight = 0;
-                        while (pChar && *pChar-- == CH_TXTATR_INWORD)
-                            ++nRight;
-
-                        pArgs->pStartPos->Assign(*this, aScanner.GetEnd() - nRight );
-                        pArgs->pEndPos->Assign(*this, aScanner.GetBegin() + nLeft );
+                        OUString sPrevWord = aScanner.GetPrevWord();
+                        auto nWordBegin = aScanner.GetBegin();
+                        auto nWordEnd = aScanner.GetEnd();
+                        bNextWord = aScanner.NextWord();
+                        const OUString& rActualWord = aScanner.GetPrevWord();
+                        bCalledNextWord = true;
+                        // check space separated word pairs in the dictionary, e.g. "vice versa"
+                        if ( !((bNextWord && pArgs->xSpeller->isValid( rActualWord + " " + aScanner.GetWord(),
+                                static_cast<sal_uInt16>(eActLang), Sequence< PropertyValue >() )) ||
+                           ( !sPrevWord.isEmpty() && pArgs->xSpeller->isValid( sPrevWord + " " + rActualWord,
+                                static_cast<sal_uInt16>(eActLang), Sequence< PropertyValue >() ))) )
+                        {
+                            // make sure the selection build later from the data
+                            // below does not include "in word" character to the
+                            // left and right in order to preserve those. Therefore
+                            // count those "in words" in order to modify the
+                            // selection accordingly.
+                            const sal_Unicode* pChar = aScanner.GetPrevWord().getStr();
+                            sal_Int32 nLeft = 0;
+                            while (*pChar++ == CH_TXTATR_INWORD)
+                                ++nLeft;
+                            pChar = rActualWord.getLength() ? rActualWord.getStr() + rActualWord.getLength() - 1 : nullptr;
+                            sal_Int32 nRight = 0;
+                            while (pChar && *pChar-- == CH_TXTATR_INWORD)
+                                ++nRight;
+
+                            pArgs->pStartPos->Assign(*this, nWordEnd - nRight );
+                            pArgs->pEndPos->Assign(*this, nWordBegin + nLeft );
+                        }
+                        else
+                        {
+                            pArgs->xSpellAlt = nullptr;
+                        }
                     }
                 }
             }
+
+            if ( !bCalledNextWord )
+                bNextWord = aScanner.NextWord();
         }
     }
 
@@ -1342,11 +1366,13 @@ SwRect SwTextFrame::AutoSpell_(SwTextNode & rNode, sal_Int32 nActPos)
         SwScanner aScanner( *pNode, pNode->GetText(), nullptr, ModelToViewHelper(),
                             WordType::DICTIONARY_WORD, nBegin, nEnd);
 
-        while( aScanner.NextWord() )
+        bool bNextWord = aScanner.NextWord();
+        while( bNextWord )
         {
             const OUString& rWord = aScanner.GetWord();
             nBegin = aScanner.GetBegin();
             sal_Int32 nLen = aScanner.GetLen();
+            bool bCalledNextWord = false;
 
             // get next language for next word, consider language attributes
             // within the word
@@ -1365,31 +1391,49 @@ SwRect SwTextFrame::AutoSpell_(SwTextNode & rNode, sal_Int32 nActPos)
                     ((!bRestoreString && !bContainsComments) || !xSpell->isValid( rWord.replaceAll(OUStringChar(CH_TXTATR_INWORD), ""),
                             static_cast<sal_uInt16>(eActLang), Sequence< PropertyValue >() ) ) )
                 {
-                    sal_Int32 nSmartTagStt = nBegin;
-                    sal_Int32 nDummy = 1;
-                    if ( !pNode->GetSmartTags() || !pNode->GetSmartTags()->InWrongWord( nSmartTagStt, nDummy ) )
+                    OUString sPrevWord = aScanner.GetPrevWord();
+                    bNextWord = aScanner.NextWord();
+                    bCalledNextWord = true;
+                    // check space separated word pairs in the dictionary, e.g. "vice versa"
+                    if ( !((bNextWord && xSpell->isValid( aScanner.GetPrevWord() + " " + aScanner.GetWord(),
+                                static_cast<sal_uInt16>(eActLang), Sequence< PropertyValue >() )) ||
+                           (!sPrevWord.isEmpty() && xSpell->isValid( sPrevWord + " " + aScanner.GetPrevWord(),
+                                static_cast<sal_uInt16>(eActLang), Sequence< PropertyValue >() ))) )
                     {
-                        if( !pNode->GetWrong() )
-                        {
-                            pNode->SetWrong( std::make_unique<SwWrongList>( WRONGLIST_SPELL ) );
-                            pNode->GetWrong()->SetInvalid( 0, nEnd );
-                        }
-                        SwWrongList::FreshState const eState(pNode->GetWrong()->Fresh(
-                            nChgStart, nChgEnd, nBegin, nLen, nInsertPos, nActPos));
-                        switch (eState)
+                        sal_Int32 nSmartTagStt = nBegin;
+                        sal_Int32 nDummy = 1;
+                        if ( !pNode->GetSmartTags() || !pNode->GetSmartTags()->InWrongWord( nSmartTagStt, nDummy ) )
                         {
-                            case SwWrongList::FreshState::FRESH:
-                                pNode->GetWrong()->Insert(OUString(), nullptr, nBegin, nLen, nInsertPos++);
-                                break;
-                            case SwWrongList::FreshState::CURSOR:
-                                bPending = true;
-                                [[fallthrough]]; // to mark as invalid
-                            case SwWrongList::FreshState::NOTHING:
-                                nInvStart = nBegin;
-                                nInvEnd = nBegin + nLen;
-                                break;
+                            if( !pNode->GetWrong() )
+                            {
+                                pNode->SetWrong( std::make_unique<SwWrongList>( WRONGLIST_SPELL ) );
+                                pNode->GetWrong()->SetInvalid( 0, nEnd );
+                            }
+                            SwWrongList::FreshState const eState(pNode->GetWrong()->Fresh(
+                                nChgStart, nChgEnd, nBegin, nLen, nInsertPos, nActPos));
+                            switch (eState)
+                            {
+                                case SwWrongList::FreshState::FRESH:
+                                    pNode->GetWrong()->Insert(OUString(), nullptr, nBegin, nLen, nInsertPos++);
+                                    break;
+                                case SwWrongList::FreshState::CURSOR:
+                                    bPending = true;
+                                    [[fallthrough]]; // to mark as invalid
+                                case SwWrongList::FreshState::NOTHING:
+                                    nInvStart = nBegin;
+                                    nInvEnd = nBegin + nLen;
+                                    break;
+                            }
                         }
                     }
+                    else if( bAddAutoCmpl && rACW.GetMinWordLen() <= aScanner.GetPrevWord().getLength() )
+                    {
+                        // tdf#119695 only add the word if the cursor position is outside the word
+                        // so that the incomplete words are not added as autocomplete candidates
+                        bool bCursorOutsideWord = nActPos > nBegin + nLen || nActPos < nBegin;
+                        if (bCursorOutsideWord)
+                            rACW.InsertWord(aScanner.GetPrevWord(), rDoc);
+                    }
                 }
                 else if( bAddAutoCmpl && rACW.GetMinWordLen() <= rWord.getLength() )
                 {
@@ -1400,6 +1444,9 @@ SwRect SwTextFrame::AutoSpell_(SwTextNode & rNode, sal_Int32 nActPos)
                         rACW.InsertWord(rWord, rDoc);
                 }
             }
+
+            if ( !bCalledNextWord )
+                 bNextWord = aScanner.NextWord();
         }
     }