Re-merge of configuration branch into main branch [277,497].

git-svn-id: https://pet.opendfki.de/repos/pet/main@498 4200e16c-5112-0410-ac55-d7fb557a720a

CHANGELOG

@@ -1,4 +1,8 @@
 v0.99.??
+  - remerge of configuration branch
+  - crash when disabling rule filter fixed (morph.cpp)
+  - clean up of settings file name handling
+  - printing of most data structures now uses C++ streams
   - more harmonization with chart-mapping branch:
     * added dag_nth_element()
     * implemented fs::fs(char *path, type_t type)

TODO

@@ -1,3 +1,44 @@
++- make << operator for tItemPrinters to allow modifier stream like usage
+   - check if making the stream an argument of the print function decreases
+     performance, if not, remove the global pointer to the stream used and
+     replace it with a method argument 
+
+- options to set/configure printers for items/dags/chart
+
+- implement timelimit with signal-based alarm and check the intended
+  functionality (real time?, cpu time?)
+
+- do a uniform treatment of all the resource limits
+
+- More generally: which parts of the processing should be affected by the given
+  limits, especially since we ge a new preprocessing component which may be
+  computationally expensive
+
+- make all _fix_me_'s \todos (so that they are handled by doxygen) and try to
+  settle as many of them as possible
+
+- code/modules that might go away:
+  - unconditional (exhaustive) unpacking
+  - CHIC printing (flop)
+  - extdict??
+
++- finish proper distribution/handling of options, still missing:
+   - `sections' to group the options, e.g., for GUI
+   - default values for when they are set with command line options, but
+     without value, like, e.g., opt_packing
+   - recode options.cpp such that it makes better use of the nice new 
+     functionality, requires renaming of the options
+   - integrate management of settings files into the config system and get rid
+     of the old settings, this needs more elaborate values (maps and such)
+   - is there a way of turning the run-time type errors, at least for the 
+     static code, into compile-time or link-time errors?
+     1. transform the string keys into #define symbols. That at least prevents
+        the use of unknown keys (requires proper include file use, which isn't
+        such a bad idea anyway). That does not resolve the problem of 
+        key<->type correlation errors.
+
+-! fix the problem with multiple coreferences in one TDL conjunction in flop
+
 - check if the dumping of cyclic structures works correctly in flop
 
 - proper handling of upcase/lowercase handling in a) tokenizers and b) lexicon
@@ -10,8 +51,6 @@
   introduced that controls if conversion to lower case is done at all.
   (ticket #5) (The Ueber- bzw. Aegypten bug)  
 
-- reasonable splitting of logging and output, ESSENTIAL FOR API IMPLEMENTATION!
-
 - unpacking of edges does not terminate in some cases with unary rules
   fix: separate the local unary edges and do a fix point computation. This must
   terminate for correct grammars
@@ -26,7 +65,7 @@
   - foreign character encoding input and output
   - automated tests for single modules ??
 
-- finish lexdb: lexical database (postgres) integration
+- finish lexdb: lexical database (sqlite instead of postgres?) integration
 
 - unreleased memory? (see valgrind-errors-15-apr-04)
   Memory leak seems to be not ignoreable (berthold et al. Jan 2006)
@@ -40,8 +79,6 @@
         used for unification and mark it with a special value. After some 
         thinking, i would say that this might not be feasible.
 
-- 64bit version does not work properly
-
 - remove all uses of negative values as `marker'-values, especially where casts
   from integers to pointers and back are involved. 
 
@@ -197,4 +234,12 @@ refactoring:
   represented by its type and no features, or all appropriate features are
   present. THOUGHT ABOUT THIS AND I DON'T THINK IT WILL WORK. Would require 
   quite complicated refilling using appropriate type fs's
-
++ reasonable splitting of logging and output, ESSENTIAL FOR API IMPLEMENTATION!
++- 64bit version does not work properly (current tests suggest that this seems
+  to be fixed)
++ all functionality for quick checks (opt_nqc...) into one place (fs.cpp/.h)
++ parsing of properties file for simple (built-in) logging
++ checked: config cheap does the same thing as main cheap. There seems to be a
+  reordering of tasks, which is why packed parses report different number of
+  passive items and parses that hit the limits report different numbers of 
+  tasks and items

borland/l2lib.h

@@ -17,7 +17,7 @@
 */
 
 /* Revision History:
-   01/10/05	UC	created
+   01/10/05  UC  created
 */
 
 // Error conditions are communicated to the caller by throwing an exception

cheap/Makefile.am

@@ -82,6 +82,7 @@ libcheap_a_SOURCES = \
 	yy-tokenizer.cpp yy-tokenizer.h \
 	$(top_srcdir)/common/bitcode.cpp \
 	$(top_srcdir)/common/chunk-alloc.cpp \
+	$(top_srcdir)/common/configs.cpp \
 	$(top_srcdir)/common/dag-alloc.cpp \
 	$(top_srcdir)/common/dag-arced.cpp \
 	$(top_srcdir)/common/dag-common.cpp \
@@ -92,6 +93,7 @@ libcheap_a_SOURCES = \
 	$(top_srcdir)/common/hash.cpp \
 	$(top_srcdir)/common/lex-io.cpp \
 	$(top_srcdir)/common/lex-tdl.cpp \
+	$(top_srcdir)/common/logging.cpp \
 	$(top_srcdir)/common/mfile.c \
 	$(top_srcdir)/common/settings.cpp \
 	$(top_srcdir)/common/types.cpp \

cheap/chart.cpp

@@ -24,13 +24,13 @@
 #include "tsdb++.h"
 #include "item-printer.h"
 #include "hashing.h"
-
+#include "options.h" // for verbosity
 #include <ostream>
 
 using namespace std;
 using namespace HASH_SPACE;
 
-//#define DEBUG
+//#define PETDEBUG
 
 chart::chart(int len, auto_ptr<item_owner> owner)
     : _Chart(), _trees(), _readings(), _pedges(0),
@@ -51,7 +51,7 @@ chart::~chart()
 
 void chart::add(tItem *it)
 {
-#ifdef DEBUG
+#ifdef PETDEBUG
   it->print(DEBUGLOGGER); DEBUGLOGGER << endl;
 #endif
 
@@ -89,7 +89,7 @@ void chart::remove(hash_set<tItem *> &to_delete)
                  , _Chart.end());
     for(hash_set<tItem *>::const_iterator hit = to_delete.begin()
           ; hit != to_delete.end(); hit++) {
-#ifdef DEBUG
+#ifdef PETDEBUG
       it->print(DEBUGLOGGER); DEBUGLOGGER << "removed " << endl;
 #endif
 
@@ -112,15 +112,15 @@ void chart::remove(hash_set<tItem *> &to_delete)
 }
 
 void chart::print(std::ostream &out, tAbstractItemPrinter *pr,
-                  bool passives, bool actives) {
+                  bool passives, bool actives) const {
   tItemPrinter def_print(out, verbosity > 2, verbosity > 10);
   if (pr == NULL) {
     pr = &def_print;
   }
   for(chart_iter pos(this); pos.valid(); pos++) {
     tItem *curr = pos.current();
     if ((curr->passive() && passives) || (! curr->passive() && actives)) {
-      pr->print(curr); out << endl;
+      pr->print(curr);// out << endl;
     }
   }
 }
@@ -212,3 +212,8 @@ chart::connected(item_predicate &valid) {
   }
   return reached[rightmost()];
 }
+
+std::ostream &operator<<(std::ostream &out, const chart &ch) {
+  ch.print(out);
+  return out;
+}

cheap/chart.h

@@ -53,7 +53,7 @@ class chart {
    *  passive items with \a passives and \a actives.
    */
   void print(std::ostream &out, tAbstractItemPrinter *aip = NULL,
-             bool passives = true, bool actives = true);
+             bool passives = true, bool actives = true) const;
 
   /** Get statistics from the chart, like nr. of active/passive edges, average
    *  feature structure size, items contributing to a reading etc.
@@ -123,6 +123,7 @@ class chart {
   friend class chart_iter_adj_passive;
 };
 
+std::ostream &operator<<(std::ostream &out, const chart &ch) ;
 
 /** Return all items from the chart.
  * \attention iterators must return items in order of `stamp', so the
@@ -131,12 +132,12 @@ class chart {
 class chart_iter {
 public:
   /** Create a new iterator for \a C */
-  inline chart_iter(chart *C) : _LI(C->_Chart) {
+  inline chart_iter(const chart *C) : _LI(C->_Chart) {
     _curr = _LI.begin();
   }
 
   /** Create a new iterator for \a C */
-  inline chart_iter(chart &C) : _LI(C._Chart) {
+  inline chart_iter(const chart &C) : _LI(C._Chart) {
     _curr = _LI.begin();
   }
 
@@ -162,8 +163,8 @@ class chart_iter {
 private:
   friend class chart;
 
-  std::vector<class tItem *> &_LI;
-  std::vector<class tItem *>::iterator _curr;
+  const std::vector<class tItem *> &_LI;
+  std::vector<class tItem *>::const_iterator _curr;
 };
 
 /** Return all passive items having a specified span.