Merge pull request #346 from inariksit/hungarian

Hungarian: vowel harmony + handle multiwords

src/hungarian/NounMorphoHun.gf

@@ -407,6 +407,8 @@ oper
 
   front_rounded : pattern Str = #("ö" | "ő" | "ü" | "ű") ;
 
+  i : pattern Str = #("i"|"í") ;
+
   -- front and back rounded
   -- rounded : pattern Str = #("ö" | "ő" | "ü" | "ű" | "o" | "ó" | "u" | "ú")
 
@@ -475,12 +477,11 @@ oper
 
   -- Function to get a harmony from any string
   getHarm : Str -> Harm = \s ->
-   let lastWord : Str = case s of {
+   let lastWord : Str = (splitMultiword s).p2 ; -- only include last word
-                          x + " " + y => y ;
-                          _           => s } ;
     in case lastWord of {
-      _ + #back + _            => H_a ;
+      _ + #back + (#c|#i)* => H_a ;       -- papír, gumi, zokni: back harmony
-      _ + #front_rounded + (#c|"") + (#c|"") => H_o ;
+                                          -- NB. wrong harmony for farmer.
+      _ + #front_rounded + (#c)* => H_o ; -- matches nő, not rövid.
       _ => H_e
     } ;
 
@@ -490,11 +491,8 @@ oper
   harmFromPlNom : Str -> Harm = \férfiak ->
     let ak : Str = dp 2 férfiak ;
      in case ak of {
-          ("ak"|"ek"|"ok") => getHarm ak ;
+          "ik" => getHarm férfiak ;
-
+          _    => getHarm ak
-          -- For any other suffix, the last two letters aren't reliable.
-          -- e.g. gumi-gumik has back harmony.
-          _ => getHarm férfiak
         } ;
 
   -- Even more reliable harmony indicator: singular allative
@@ -632,4 +630,24 @@ oper
       _ => endCase c
     } ;
 
+    Multiword : Type = {p1,p2 : Str} ;
+    splitMultiword : Str -> Multiword = \mw -> case mw of {
+        v + " " + w + " " + x + " " + y + " " + z
+          => splitDash <v ++ w ++ x ++ y+" ", z> ;
+        w + " " + x + " " + y + " " + z
+          => splitDash <w ++ x ++ y+" ", z> ;
+        x + " " + y + " " + z
+          => splitDash <x ++ y+" ", z> ;
+        y + " " + z
+          => splitDash <y + " ", z> ;
+        _ => splitDash <"", mw>
+        } ;
+
+    splitDash : Multiword -> Multiword = \mw ->
+      case mw of {
+        <prefix, x + "-" + y>
+          => <prefix ++ x + "-", y> ;
+        _ => mw
+      } ;
+
 }

src/hungarian/ParadigmsHun.gf

@@ -122,34 +122,56 @@ oper
   harmO = ResHun.H_o ;
 
   mkN = overload {
-    mkN : Str -> N =
+    mkN : Str -> N = \s ->
-      \s -> lin N (regNoun s) ;
+      let mw : Multiword = splitMultiword s ;
+       in case mw of {
+          <pr,n> => multiwordN pr (regNoun n)
+      } ;
 
-    mkN : Str -> Str -> N =
+    mkN : Str -> Str -> N = \n,a ->
-        \n,a-> lin N (regNounNomAcc n a) ;
+      let mwn : Multiword = splitMultiword n ;
+          mwa : Multiword = splitMultiword a ;
+       in multiwordN mwn.p1 (regNounNomAcc mwn.p2 mwa.p2) ;
 
-    mkN : Str -> Str -> Str -> N =
+    mkN : Str -> Str -> Str -> N = \n,a,pln ->
-        \n,a,pln-> lin N (regNounNomAccPl n a pln) ;
+      let mwn : Multiword = splitMultiword n ;
+          mwa : Multiword = splitMultiword a ;
+          mwpln : Multiword = splitMultiword pln ;
+       in multiwordN mwn.p1 (regNounNomAccPl mwn.p2 mwa.p2 mwpln.p2) ;
 
-    mkN : (x1,_,_,x4 : Str) -> N =
+    mkN : (x1,_,_,x4 : Str) -> N = \n,a,pln,possd ->
-        \n,a,pln,possd -> lin N (regNoun4 n a pln possd) ;
+      let mwn : Multiword = splitMultiword n ;
+          mwa : Multiword = splitMultiword a ;
+          mwpln : Multiword = splitMultiword pln ;
+          mwpossd : Multiword = splitMultiword possd ;
+       in multiwordN mwn.p1 (regNoun4 mwn.p2 mwa.p2 mwpln.p2 mwpossd.p2) ;
 
-    mkN : (unoka : Str) -> (testvér : N) -> N = -- Compound noun: e.g. `mkN "unoka" (mkN "testvér")`.
+    mkN : (unoka : Str) -> (testvér : N) -> N = compoundN ;
-        \prefix,n -> n ** {s = \\x => prefix + n.s ! x} ;
 
+    ---------------------------------------------------------------------
+    -- Not in the visible API. TODO remove, improve or document better --
+    ---------------------------------------------------------------------
+
+    -- Worst case with 9 strings.
     mkN : (x1,_,_,_,_,_,_,_,x9 : Str) -> N =
       \nomsg,accsg,supsg,allsg,nompl,f,g,h,i ->
       lin N (worstCaseNoun nomsg accsg supsg allsg nompl
                            f g h i (harmFromSgAll allsg)) ;
 
-    -- mkN : (férfi : Str) -> (harm : Harmony) -> (ak : Str) -> N ; -- Noun with unpredictable vowel harmony and plural allomorph
+    -- Noun with unpredictable vowel harmony and plural allomorph
-    mkN : Str -> Harmony -> N =
-      \s,h -> lin N (mkNounHarm h (pluralAllomorph s) s) ;
-
     mkN : Str -> (plural : Str) -> Harmony -> N =
-      \s,pl,h -> lin N (mkNounHarm h pl s) ;
+      \s,pl,h -> lin N (mkNounHarm h pl s)
+
     } ;
 
+   multiwordN : Str -> Noun -> N =
+     \prefix,n -> case prefix of {
+       _ + "-" => compoundN prefix n ;
+       _       => lin N (n ** {s = \\x => prefix ++ n.s ! x})
+     } ;
+
+   compoundN : (unoka : Str) -> (testvér : Noun) -> N = -- Compound noun: e.g. `mkN "unoka" (mkN "testvér")`.
+     \prefix,n -> lin N (n ** {s = \\x => prefix + n.s ! x}) ;
 
   mkN2 = overload {
     mkN2 : Str -> N2 = \s -> lin N2 (regNoun s) ;

src/hungarian/ResHun.gf

@@ -97,6 +97,12 @@ oper
     objdef = Def ;
     } ;
 
+  defNPPrefix : (p,n : Str) -> Number -> NounPhrase = \vala,mi,n -> emptyNP ** {
+    s = \\p,c => vala + mkCaseNoun mi ! n ! c ;
+    n = n ;
+    objdef = Def ;
+    } ;
+
   linCN : CNoun -> Str = \cn -> cn.s ! SgNom ++ cn.compl ! Sg ! Nom ++ cn.postmod ;
   linNP' : Possessor -> Case -> NounPhrase -> Str = \p,c,np -> np.s ! p ! c ++ np.postmod ;
   linNP : NounPhrase -> Str = linNP' NoPoss Nom ;

src/hungarian/StructuralHun.gf

@@ -87,7 +87,7 @@ lin nobody_NP = mkVerb; ""
 lin nothing_NP = defNP "" N.NumSg ;
 lin somebody_NP = defNP "" N.NumSg ;
 -}
-lin something_NP = defNP "valami" Sg ;
+lin something_NP = defNPPrefix "vala" "mi" Sg ; -- vowel harmony according to mi
 
 -------
 -- Prep