named_enum, itoa, cdec replace --a-b=x with --a_b=x

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@536 ec762483-ff6d-05da-a07a-a48fb63a330f

7 files changed, 87 insertions, 26 deletions

diff --git a/decoder/apply_fsa_models.cc b/decoder/apply_fsa_models.cc
index 1c30eb90..2854b28b 100755
--- a/decoder/apply_fsa_models.cc
+++ b/decoder/apply_fsa_models.cc
@@ -13,6 +13,8 @@
 
 using namespace std;
 
+DEFINE_NAMED_ENUM(FSA_BY)
+
 struct ApplyFsa {
   ApplyFsa(HgCFG &i,
            const SentenceMetadata& smeta,
@@ -74,6 +76,7 @@ void ApplyFsaModels(HgCFG &i,
   a.Compute();
 }
 
+/*
 namespace {
 char const* anames[]={
   "BU_CUBE",
@@ -82,14 +85,18 @@ char const* anames[]={
   0
 };
 }
+*/
 
 //TODO: named enum type in boost?
 
 std::string ApplyFsaBy::name() const {
-  return anames[algorithm];
+//  return anames[algorithm];
+  return GetName(algorithm);
 }
 
 std::string ApplyFsaBy::all_names() {
+  return FsaByNames(" ");
+  /*
   std::ostringstream o;
   for (int i=0;i<N_ALGORITHMS;++i) {
     assert(anames[i]);
@@ -97,19 +104,24 @@ std::string ApplyFsaBy::all_names() {
     o<<anames[i];
   }
   return o.str();
+  */
 }
 
 ApplyFsaBy::ApplyFsaBy(std::string const& n, int pop_limit) : pop_limit(pop_limit) {
-  algorithm=0;
   std::string uname=toupper(n);
+  algorithm=GetFsaBy(uname);
+/*anames=0;
   while(anames[algorithm] && anames[algorithm] != uname) ++algorithm;
   if (!anames[algorithm])
     throw std::runtime_error("Unknown ApplyFsaBy type: "+n+" - legal types: "+all_names());
+*/
 }
 
-ApplyFsaBy::ApplyFsaBy(int i, int pop_limit) : pop_limit(pop_limit) {
-  if (i<0 || i>=N_ALGORITHMS)
+ApplyFsaBy::ApplyFsaBy(FsaBy i, int pop_limit) : pop_limit(pop_limit) {
+/*  if (i<0 || i>=N_ALGORITHMS)
     throw std::runtime_error("Unknown ApplyFsaBy type id: "+itos(i)+" - legal types: "+all_names());
+*/
+  GetName(i); // checks validity
   algorithm=i;
 }
 
diff --git a/decoder/apply_fsa_models.h b/decoder/apply_fsa_models.h
index 5120fb4e..6561c70c 100755
--- a/decoder/apply_fsa_models.h
+++ b/decoder/apply_fsa_models.h
@@ -4,25 +4,36 @@
 #include <string>
 #include <iostream>
 #include "feature_vector.h"
+#include "named_enum.h"
 
 struct FsaFeatureFunction;
 struct Hypergraph;
 struct SentenceMetadata;
 struct HgCFG;
 
+
+#define FSA_BY(X,t)                              \
+  X(t,BU_CUBE,)                                 \
+  X(t,BU_FULL,)                                 \
+  X(t,EARLEY,)                                  \
+
+#define FSA_BY_TYPE FsaBy
+
+DECLARE_NAMED_ENUM(FSA_BY)
+
 struct ApplyFsaBy {
-  enum {
-    BU_CUBE,
-    BU_FULL,
-    EARLEY,
-    N_ALGORITHMS
-  };
+/*enum {
+  BU_CUBE,
+  BU_FULL,
+  EARLEY,
+  N_ALGORITHMS
+  };*/
   int pop_limit; // only applies to BU_FULL so far
   bool IsBottomUp() const {
     return algorithm==BU_FULL || algorithm==BU_CUBE;
   }
   int BottomUpAlgorithm() const;
-  int algorithm;
+  FsaBy algorithm;
   std::string name() const;
   friend inline std::ostream &operator << (std::ostream &o,ApplyFsaBy const& c) {
     o << c.name();
@@ -30,7 +41,7 @@ struct ApplyFsaBy {
       o << "("<<c.pop_limit<<")";
     return o;
   }
-  explicit ApplyFsaBy(int alg, int poplimit=200);
+  explicit ApplyFsaBy(FsaBy alg, int poplimit=200);
   ApplyFsaBy(std::string const& name, int poplimit=200);
   ApplyFsaBy(const ApplyFsaBy &o) : algorithm(o.algorithm) {  }
   static std::string all_names(); // space separated
diff --git a/decoder/cdec.cc b/decoder/cdec.cc
index 3633febd..5898b245 100644
--- a/decoder/cdec.cc
+++ b/decoder/cdec.cc
@@ -193,7 +193,7 @@ void InitCommandLine(int argc, char** argv, OracleBleu &ob, po::variables_map* c
   dconfig_options.add(opts).add(cfgo);
   //add(opts).add(cfgo)
   dcmdline_options.add(dconfig_options).add(clo);
-
+  argv_minus_to_underscore(argc,argv);
   po::store(parse_command_line(argc, argv, dcmdline_options), conf);
   if (conf.count("compgen")) {
     print_options(cout,dcmdline_options);
diff --git a/decoder/cfg.cc b/decoder/cfg.cc
index 81a17355..aa9e5f30 100755
--- a/decoder/cfg.cc
+++ b/decoder/cfg.cc
@@ -11,13 +11,21 @@ using namespace std;
 
 namespace {
 CFG::BinRhs nullrhs(std::numeric_limits<int>::min(),std::numeric_limits<int>::min());
-}
-
 
-WordID CFG::BinName(BinRhs const& b)
+// index i >= N.size()?  then it's in M[i-N.size()]
+WordID BinName(CFG::BinRhs const& b,CFG::NTs const& N,CFG::NTs const& M)
 {
+  int nn=N.size();
   ostringstream o;
-#define BinNameOWORD(w) do { int n=w; if (n>0) o << TD::Convert(n); else { o << 'V' << -n; } } while(0)
+#define BinNameOWORD(w) \
+  do { \
+    int n=w; if (n>0) o << TD::Convert(n); \
+    else {                                   \
+      int i=-n; \
+      CFG::NT const&nt = i<nn?N[i]:M[i-nn];    \
+      o << nt.from << i; }                     \
+  } while(0)
+
   BinNameOWORD(b.first);
   o<<'+';
   BinNameOWORD(b.second);
@@ -25,6 +33,10 @@ WordID CFG::BinName(BinRhs const& b)
   return TD::Convert(o.str());
 }
 
+}
+
+
+
 void CFG::Binarize(CFGBinarize const& b) {
   if (!b.Binarizing()) return;
   if (!b.bin_l2r) {
@@ -57,7 +69,7 @@ void CFG::Binarize(CFGBinarize const& b) {
           new_nts.back().ruleids.push_back(newruleid);
           new_rules.push_back(Rule(newnt,bin));
           if (b.bin_name_nts)
-            new_nts.back().from.nt=BinName(bin);
+            new_nts.back().from.nt=BinName(bin,nts,new_nts);
           ++newnt;++newruleid;
         }
       }
diff --git a/decoder/cfg.h b/decoder/cfg.h
index 64924f14..e1f818e8 100755
--- a/decoder/cfg.h
+++ b/decoder/cfg.h
@@ -40,11 +40,10 @@ struct CFG {
   typedef std::vector<RuleHandle> Ruleids;
 
   void print_nt_name(std::ostream &o,NTHandle n) const {
-    o << nts[n].from;
+    o << nts[n].from << n;
   }
 
   typedef std::pair<int,int> BinRhs;
-  WordID BinName(BinRhs const& b);
 
   struct Rule {
     // for binarizing - no costs/probs
@@ -106,16 +105,17 @@ struct CFG {
     swap(goal_nt,o.goal_nt);
   }
   void Binarize(CFGBinarize const& binarize_options);
+
+  typedef std::vector<NT> NTs;
+  NTs nts;
+  typedef std::vector<Rule> Rules;
+  Rules rules;
+  int goal_nt;
+  prob_t goal_inside,pushed_inside; // when we push viterbi weights to goal, we store the removed probability in pushed_inside
 protected:
   bool uninit;
   Hypergraph const* hg_; // shouldn't be used for anything, esp. after binarization
-  prob_t goal_inside,pushed_inside; // when we push viterbi weights to goal, we store the removed probability in pushed_inside
   // rules/nts will have same index as hg edges/nodes
-  typedef std::vector<Rule> Rules;
-  Rules rules;
-  typedef std::vector<NT> NTs;
-  NTs nts;
-  int goal_nt;
 };
 
 inline void swap(CFG &a,CFG &b) {
diff --git a/decoder/cfg_options.h b/decoder/cfg_options.h
index 956586f0..acd8d05b 100755
--- a/decoder/cfg_options.h
+++ b/decoder/cfg_options.h
@@ -28,6 +28,7 @@ struct CFGOptions {
   void Validate() {
     format.Validate();
     binarize.Validate();
+//    if (cfg_output.empty()) binarize.bin_name_nts=false;
   }
   char const* description() const {
     return "CFG output options";
diff --git a/decoder/program_options.h b/decoder/program_options.h
index 251f5680..87afb320 100755
--- a/decoder/program_options.h
+++ b/decoder/program_options.h
@@ -13,6 +13,31 @@
 #include <iosfwd>
 
 
+// change --opt-name=x --opt_name=x for all strings x.  danger: probably the argv from int main isn't supposed to be modified?
+inline int arg_minusto_underscore(char *s) {
+  if (!*s || *s++ != '-') return 0;
+  if (!*s || *s++ != '-') return 0;
+  int chars_replaced=0;
+  for(;*s;++s) {
+    if (*s=='=')
+      break;
+    if (*s=='-') {
+      *s='_';
+      ++chars_replaced;
+    }
+  }
+  return chars_replaced;
+}
+
+inline
+int argv_minus_to_underscore(int argc, char **argv) {
+  int chars_replaced=0;
+  for (int i=1;i<argc;++i) {
+    chars_replaced+=arg_minusto_underscore(argv[i]);
+  }
+  return chars_replaced;
+}
+
 template <class T>
 boost::program_options::typed_value<T>*
 defaulted_value(T *v)